This Dum Week 2026-08-30

1
00:00:00,288 --> 00:00:03,542
The voice behind an iconic character has died.
2
00:00:03,718 --> 00:00:07,646
My name is Optimus Prime. We are autonomous
3
00:00:08,146 --> 00:00:11,333
robotic organisms from the planet Cybertron.
4
00:00:11,653 --> 00:00:15,485
Peter Cullen had been a voice for many Transformers franchises since
5
00:00:15,985 --> 00:00:19,620
the 1980s. We're from Cybertron. Born in Montreal,
6
00:00:19,733 --> 00:00:23,676
Canada, according to the CBC, Cullen got his start in radio in
7
00:00:24,176 --> 00:00:27,668
the 1960s, transitioning to television and then film.
8
00:00:28,164 --> 00:00:31,755
While he voiced other well-known characters, he's best known for the deep,
9
00:00:31,900 --> 00:00:34,996
deep, gravelly, intimidating voice of Optimus Prime.
10
00:00:36,184 --> 00:00:39,954
This is not our war. Cullen reportedly hadn't been
11
00:00:40,454 --> 00:00:44,511
cast as Optimus Prime in 20 years when director Michael Bay requested
12
00:00:45,011 --> 00:00:48,266
him to be part of his revival of Transformers in 2007,
13
00:00:48,539 --> 00:00:52,133
a blockbuster film starring Shia LaBeouf and Megan Fox.
14
00:00:52,389 --> 00:00:55,839
He reportedly died at his home in Los Angeles on August
15
00:00:56,339 --> 00:00:58,936
26th. Peter Cullen was 85.
16
00:01:03,064 --> 00:01:06,204
It's a story that I've told many times,
17
00:01:07,153 --> 00:01:10,727
and maybe some of you have heard it, but it's a story very dear to
18
00:01:11,227 --> 00:01:14,686
me, and it's the truth, and it happened. My brother Larry,
19
00:01:14,879 --> 00:01:18,843
a captain in the Marine Corps, had returned from Vietnam. He was a decorated
20
00:01:19,343 --> 00:01:23,555
Marine, and the Bronze
21
00:01:24,055 --> 00:01:27,901
Star with V and a couple of Purple Hearts. And we
22
00:01:28,401 --> 00:01:32,462
were sharing an apartment, going through a difficult time in my life when
23
00:01:32,511 --> 00:01:35,347
Larry came down. And it was about 1984,
24
00:01:36,920 --> 00:01:39,314
and we were, as I said, living in an apartment together.
25
00:01:40,658 --> 00:01:44,201
And he would always take a little interest in
26
00:01:44,298 --> 00:01:48,869
where I was going every day, 'cause I was a— well,
27
00:01:49,096 --> 00:01:52,406
a grunt, as the Marines would say. Mm-hmm. You know, if you're a ground trooper,
28
00:01:52,438 --> 00:01:55,722
you're a grunt. Anyway,
29
00:01:55,738 --> 00:01:59,477
Larry's 13 months older, 5 inches taller.
30
00:02:02,158 --> 00:02:04,872
But we're 13 months apart, we were very, very close.
31
00:02:06,100 --> 00:02:09,321
So he said, Peter, where are you going today? I said,
32
00:02:10,159 --> 00:02:13,061
well, Larry, why, you need the car? He said, no,
33
00:02:13,142 --> 00:02:16,576
no, I just wanna know where you're going. I said, I'm going to audition.
34
00:02:16,689 --> 00:02:20,388
I'm gonna be— I'm auditioning as a truck.
35
00:02:20,437 --> 00:02:25,166
A truck? And he
36
00:02:25,666 --> 00:02:28,698
did the same thing. And so I said,
37
00:02:28,714 --> 00:02:31,969
but Larry, he's evidently, he's a hero truck. He's a
38
00:02:32,034 --> 00:02:35,904
real, I mean, he's a hero. And his face got somber
39
00:02:36,404 --> 00:02:38,733
and he said, Peter, if you're gonna be a hero,
40
00:02:40,130 --> 00:02:44,033
be a real hero. Don't be a Hollywood hero
41
00:02:44,533 --> 00:02:47,857
type with all the BS and all the yelling and screaming and the trying to
42
00:02:48,357 --> 00:02:50,913
be tough. Be strong enough to be gentle.
43
00:02:53,420 --> 00:02:57,854
Well, okay, Larry, yeah. So his
44
00:02:57,951 --> 00:03:00,684
voice is ringing in my ears as I'm driving to the audition,
45
00:03:02,237 --> 00:03:05,876
and I read the copy, and Larry is just coming
46
00:03:06,376 --> 00:03:10,040
out, his voice, the way he talked to me. And it said,
47
00:03:11,040 --> 00:03:14,445
you know, I'm Optimus Prime. No,
48
00:03:14,832 --> 00:03:18,333
it was, My name is Optimus Prime.
49
00:03:22,940 --> 00:03:26,490
Welcome to This Dumb Week, a weekly live Twitter Space hosted by Alex Marinos
50
00:03:26,522 --> 00:03:29,879
and yours truly, Dr. Rollergator. Available in podcast form
51
00:03:29,976 --> 00:03:33,527
on Apple Podcasts, Spotify, Amazon Music, Audible,
52
00:03:34,190 --> 00:03:37,953
iHeartRadio, and others. Alex, welcome back. Have you had a sufficiently
53
00:03:37,986 --> 00:03:41,550
dumb week? Uh, yeah, given that the
54
00:03:42,050 --> 00:03:44,981
US has just bombed Iran again, I think, uh, we're getting there.
55
00:03:46,639 --> 00:03:49,555
Well, when, when did this occur? When did I just miss this doing show prep
56
00:03:49,717 --> 00:03:53,265
still and getting ready for the show? Probably. I mean,
57
00:03:53,281 --> 00:03:56,263
it is Groundhog Day. Doesn't matter.
58
00:03:56,765 --> 00:04:00,671
Is this in response to Optimus Prime dying? Because there
59
00:04:01,171 --> 00:04:04,463
was no official— Yeah, Optimus Prime died. Well, I mean, that was, that was the
60
00:04:04,963 --> 00:04:10,519
entire intro, Alex. That was the— I didn't
61
00:04:11,019 --> 00:04:13,933
know that Iran was behind it, but you know, I think I, in, in that
62
00:04:13,967 --> 00:04:15,341
rare limited, uh, instance. Yeah.
63
00:04:17,180 --> 00:04:21,722
If Iran had something to do— everyone, including Netanyahu's
64
00:04:22,222 --> 00:04:25,699
idiot son. If Iran had something to do with
65
00:04:25,747 --> 00:04:28,284
Optimus Prime's death, I do want revenge.
66
00:04:28,626 --> 00:04:32,033
So. Yeah, are you gonna go on a jihad, or how are
67
00:04:32,533 --> 00:04:36,260
you thinking? I'll figure that out as we
68
00:04:36,760 --> 00:04:38,773
go along, just like most things in this country right now.
69
00:04:40,110 --> 00:04:43,815
Okay. You act first to figure it
70
00:04:44,315 --> 00:04:46,550
out later, right? Yes, that definitely seems to be the American way.
71
00:04:48,518 --> 00:04:51,073
All right, all right.
72
00:04:51,716 --> 00:04:54,640
Well, Alex, we'll get moving here on, uh, some of the, uh,
73
00:04:55,653 --> 00:04:59,148
arbitrary news of the week before we get into anything too deep here.
74
00:05:06,473 --> 00:05:09,894
Well, this is one animal sighting you don't expect to see in Indiana.
75
00:05:09,959 --> 00:05:13,335
Police in Madison and Hancock Counties are investigating
76
00:05:13,351 --> 00:05:17,024
reports of lions on the loose. WRTV's Kat
77
00:05:17,524 --> 00:05:21,885
Sandoval reports on the unusual exotic sighting. Police received
78
00:05:21,901 --> 00:05:25,061
a report about a lion on the loose in a field like this,
79
00:05:25,334 --> 00:05:27,901
and it's the talk of the town. It kind of seems cool to have a
80
00:05:27,917 --> 00:05:31,430
lion here. I mean, we've never had some in Indiana, but it
81
00:05:31,930 --> 00:05:34,109
also kind of seems kind of scary because, I mean, I don't want to get
82
00:05:34,158 --> 00:05:38,264
chased by a lion. The Madison County Sheriff's Department says they received
83
00:05:38,764 --> 00:05:42,805
a tip at 10 Thursday morning about a lioness and a cub sighting between
84
00:05:42,901 --> 00:05:46,817
Pendleton and Marklesville. North of Hancock and south of Madison
85
00:05:46,913 --> 00:05:49,804
County. Well, it's not something you see every day in Indiana,
86
00:05:50,607 --> 00:05:54,540
and, uh, everybody was talking about it, and so it
87
00:05:55,040 --> 00:05:59,036
seemed like somebody would know who owns it. The Madison County and Hancock County sheriffs,
88
00:05:59,357 --> 00:06:02,874
along with the Indiana State Police, joined in the hours-long search,
89
00:06:03,307 --> 00:06:07,129
tracking physical evidence and deploying thermal imaging drones.
90
00:06:07,466 --> 00:06:09,971
I have no idea, but yeah.
91
00:06:10,276 --> 00:06:14,197
They're out looking. And then it blew up online. The Madison
92
00:06:14,229 --> 00:06:17,587
County Sheriff's Facebook page had hundreds of comments like,
93
00:06:18,005 --> 00:06:21,283
in the jungle, the mighty jungle, the lion sleeps tonight,
94
00:06:21,685 --> 00:06:25,397
and here kitty kitty. I heard it all day at work and I
95
00:06:25,461 --> 00:06:28,675
was talking to people and everybody's like, yeah, it's true. And I'm like, oh my
96
00:06:29,175 --> 00:06:32,354
gosh, because I live, you know, right down the road
97
00:06:32,370 --> 00:06:35,681
from where they're saying it's at. And while police could not confirm whether
98
00:06:36,181 --> 00:06:39,592
the the lions are truly on the prowl. They say they got a tip that
99
00:06:40,092 --> 00:06:43,664
the animals were last seen near 1100 South and South County
100
00:06:43,696 --> 00:06:47,191
Road 50 West. The thermal imaging drones did
101
00:06:47,223 --> 00:06:50,765
give police a bit of a tailspin when they spotted a pig
102
00:06:51,150 --> 00:06:54,757
and not lions. Saw that they're looking for them, so I really hope they find
103
00:06:55,257 --> 00:06:58,748
them so they're safe. Now the big question is, was it really a
104
00:06:59,248 --> 00:07:02,772
lion? Police are still working on that answer. For now, they've called off the
105
00:07:03,272 --> 00:07:06,924
search until they find another lead. Reporting in Pendleton, I'm Kat
106
00:07:07,424 --> 00:07:11,484
Sandoval for WFLA. All right,
107
00:07:11,516 --> 00:07:14,924
Alex. So it seems that
108
00:07:15,424 --> 00:07:19,561
the illegal immigration situation in this country is so out of control
109
00:07:19,578 --> 00:07:22,925
that we have illegal immigrant lions running through
110
00:07:24,283 --> 00:07:27,146
the cities. Alex, what do you think here? Do you think we can get this
111
00:07:27,437 --> 00:07:31,629
situation under control anytime soon, or is this— Is
112
00:07:31,645 --> 00:07:34,332
this a long con to entrap Mali-anopolis? Sorry.
113
00:07:35,484 --> 00:07:39,151
Milo Yiannopoulos deported to the UK one day after ICE
114
00:07:39,377 --> 00:07:40,314
arrest.
115
00:07:46,597 --> 00:07:52,781
Yes, Alex, you, uh, you have predicted my—
116
00:07:52,894 --> 00:07:55,391
I, I cannot pull a fast one on you so far today.
117
00:07:57,386 --> 00:08:01,191
Um, Milo Yiannopoulos deported
118
00:08:01,691 --> 00:08:04,641
to the UK one day after ICE arrest. Milo Yiannopoulos, a controversy-baiting internet personality who
119
00:08:05,141 --> 00:08:08,549
once vehemently backed President Donald
120
00:08:09,049 --> 00:08:12,877
Trump has been deported to the United Kingdom, a Department of Homeland Security spokesperson said
121
00:08:13,377 --> 00:08:17,510
on Saturday. Yiannopoulos was arrested by ICE at the Louis Armstrong
122
00:08:17,542 --> 00:08:20,940
New Orleans International Airport on Friday after officials
123
00:08:21,440 --> 00:08:23,569
said he failed to show up for an immigration hearing in July.
124
00:08:24,515 --> 00:08:27,769
DHS said that a final removal order was issued by an immigration
125
00:08:27,881 --> 00:08:32,113
judge on July 22nd. ICE agents are increasingly making immigration-related
126
00:08:32,613 --> 00:08:36,398
arrests at airports. Yanopoulos, a self-described libertarian who was
127
00:08:36,784 --> 00:08:40,848
an early supporter of Trump's political career, has worked with other MAGA-aligned figures
128
00:08:41,024 --> 00:08:44,735
and conspiracy theorists like Ye, formerly Kanye West,
129
00:08:45,056 --> 00:08:48,477
since his public parting from Trump's inner circle. Ye, who has come
130
00:08:48,525 --> 00:08:51,706
under fire in recent years for a series of antisemitic statements,
131
00:08:51,722 --> 00:08:54,661
performed at the New Orleans Caesars Superdome on Friday evening.
132
00:08:55,352 --> 00:08:58,628
It is not clear whether Yanopoulos was at the— was in New Orleans for
133
00:08:59,128 --> 00:09:02,510
the concert. Yiannopoulos served as a tech editor at Breitbart, a conservative website,
134
00:09:02,878 --> 00:09:05,267
until February 2017, according to the site.
135
00:09:05,668 --> 00:09:09,339
Yiannopoulos resigned at that time following backlash over comments he'd made suggesting
136
00:09:09,355 --> 00:09:13,074
he supported certain adult relationships with children. In those comments,
137
00:09:13,138 --> 00:09:16,937
the openly gay agitator discussed the value of cross-generational
138
00:09:16,985 --> 00:09:20,464
relationships. And I think particularly in the gay
139
00:09:20,512 --> 00:09:23,509
world and outside the Catholic Church, if that's where some of you want to go
140
00:09:23,541 --> 00:09:26,731
with this, I think in the gay world, some of the most important, enriching,
141
00:09:26,763 --> 00:09:30,851
and incredibly life-affirming, important shaping relationships very often between younger boys
142
00:09:31,351 --> 00:09:35,082
and older men, Yiannopoulos said in a past video clip. Quote, they can
143
00:09:35,582 --> 00:09:39,104
be hugely positive experiences. He also remarked that sexual abuse is not the worst
144
00:09:39,604 --> 00:09:42,117
thing that's ever going to happen to you. Going bankrupt is worse.
145
00:09:42,614 --> 00:09:46,251
Yiannopoulos was disinvited from speaking at a conservative political
146
00:09:46,299 --> 00:09:50,001
action conference around the time in the wake of those inflammatory remarks.
147
00:09:50,065 --> 00:09:53,190
He later apologized and said he didn't advocate for illegal behavior.
148
00:09:53,591 --> 00:09:56,443
He said he was sexually abused himself as a teen by 2 older men.
149
00:09:56,972 --> 00:10:00,483
The Trump administration on Friday publicized Yanopoulos's arrest to help promote
150
00:10:00,983 --> 00:10:04,859
its voluntary deportation program. DHS is offering undocumented immigrants
151
00:10:04,988 --> 00:10:08,306
$3,000 and a free flight home for self-deportation.
152
00:10:08,995 --> 00:10:12,987
Quote, being in detention is a choice, DHS wrote on X above Yanopoulos's
153
00:10:13,487 --> 00:10:17,043
haggard-looking mugshot. Quote, we encourage all illegal aliens to take control
154
00:10:17,315 --> 00:10:20,201
of their departure with the Customs and Border Patrol Home app.
155
00:10:20,762 --> 00:10:23,824
Conspiracy peddler Laura Loomer, who is close to Trump and
156
00:10:24,324 --> 00:10:27,783
who has claimed responsibility for a number of firings and criminal investigations, took credit for
157
00:10:28,023 --> 00:10:31,398
Ianopolis to ICE in a social media post. So Alex,
158
00:10:31,721 --> 00:10:35,339
you did catch my setup here, so I'm going
159
00:10:35,371 --> 00:10:38,692
to have to start being a little bit more clever in the way that I
160
00:10:39,192 --> 00:10:43,058
string things together, because I can't have you just predicting the future
161
00:10:43,074 --> 00:10:47,079
here. It's an arms race, I tell
162
00:10:47,579 --> 00:10:50,968
ya. Yeah, well, you know,
163
00:10:52,491 --> 00:10:55,129
whatever. Whatever?
164
00:10:56,294 --> 00:10:59,740
Yeah, pretty much. Just whatever. Oh man.
165
00:11:00,889 --> 00:11:03,575
Come on. Yep. Yeah. Yeah. Is that,
166
00:11:03,656 --> 00:11:05,744
is it that kind of day today, Alex?
167
00:11:07,338 --> 00:11:11,074
Look, it, you know, if now we're in like Lumerus
168
00:11:11,574 --> 00:11:14,631
fan, we're, uh, getting on the wrong side of our
169
00:11:15,131 --> 00:11:17,753
Lumerus. What, what, what does you in?
170
00:11:19,064 --> 00:11:22,801
Uh, whatever. I don't know. I have no comments
171
00:11:23,301 --> 00:11:26,925
on the situation. He probably should have made his appointment. To be honest,
172
00:11:26,941 --> 00:11:30,479
I, I, I don't even know what his immigration
173
00:11:30,979 --> 00:11:35,320
situation was. He said at one point that he had a green card, which don't
174
00:11:35,820 --> 00:11:39,623
seem to— don't tend to expire, but maybe he lied. Like, I have just
175
00:11:39,802 --> 00:11:42,932
no idea. I think he had a worker's visa
176
00:11:43,432 --> 00:11:45,808
for a while, but then he didn't have a job for a very long time,
177
00:11:45,856 --> 00:11:49,214
so I don't know what happens to your worker's visa if you
178
00:11:49,714 --> 00:11:52,058
don't have a job for a significant period of time. If you have a green
179
00:11:52,558 --> 00:11:56,322
card, which is what he claimed at one point, those don't
180
00:11:56,370 --> 00:11:58,600
expire. Um,
181
00:11:59,980 --> 00:12:03,445
but whatever. Yeah, I don't know.
182
00:12:03,654 --> 00:12:07,183
I mean— Are you, are you potentially considering supporting a GoFundMe if
183
00:12:07,683 --> 00:12:10,969
he needs one for his re-immigration?
184
00:12:12,237 --> 00:12:16,151
Well, as he said, uh, you know, being abused
185
00:12:16,184 --> 00:12:20,129
is one thing, but being broke is worse. So I guess I'm just gonna Not
186
00:12:20,177 --> 00:12:22,886
contribute. Ah, all right,
187
00:12:22,982 --> 00:12:25,674
Alex, I guess that, that works. All right, well,
188
00:12:26,283 --> 00:12:29,552
Alex, we, we've, um, in the last couple weeks have discussed
189
00:12:30,052 --> 00:12:33,815
some, uh, police who are like using flock cameras inappropriately or
190
00:12:34,315 --> 00:12:37,132
doing some other similar shady, uh, type, uh,
191
00:12:37,838 --> 00:12:41,203
stalking behavior or using— abusing their
192
00:12:41,251 --> 00:12:45,482
positions basically in law enforcement to, uh, do inappropriate things
193
00:12:45,514 --> 00:12:48,943
that are frowned upon in this society. But, uh, we, we haven't talked
194
00:12:49,443 --> 00:12:53,927
too much about judges. Who, uh, who might be engaged in, in untoward
195
00:12:53,960 --> 00:12:57,695
behavior, Alex. And this next story is, is about one such judge.
196
00:13:00,330 --> 00:13:03,798
Welcome back. A former Hamilton County judge pleading not guilty to
197
00:13:03,895 --> 00:13:07,186
prostitution charges. Investigators say James Kenny had
198
00:13:07,218 --> 00:13:10,565
sex with an inmate and brought banned items into the jail.
199
00:13:10,615 --> 00:13:14,474
Our WLWT News 5 anchor Chris Jacobs here now with the developments
200
00:13:14,523 --> 00:13:17,872
from court today. Hi there, Chris. This was a quick hearing today, so Kenny entered
201
00:13:18,372 --> 00:13:22,330
the courtroom in a wheelchair. Today's bond bond hearing in a Clermont County courtroom
202
00:13:22,410 --> 00:13:25,922
only lasted a few minutes, but it detailed why the former judge
203
00:13:26,275 --> 00:13:29,691
is now on the wrong side of the law. Former Hamilton County
204
00:13:29,803 --> 00:13:33,363
Judge James Kenny now has the legal system pointed at him.
205
00:13:33,492 --> 00:13:37,998
Essentially, that he would bring those items vape pens, cans tobacco
206
00:13:38,286 --> 00:13:41,975
to the jail and use his status as an attorney to gain access.
207
00:13:42,312 --> 00:13:44,993
The 75-year-old's charges of prostitution,
208
00:13:45,057 --> 00:13:48,542
solicitation, and obstruction come from an investigation
209
00:13:48,654 --> 00:13:52,524
earlier this month when inmates at the Clermont County Jail were found
210
00:13:52,572 --> 00:13:56,570
with banned tobacco products. Ultimately, this defendant was interviewed
211
00:13:57,325 --> 00:14:00,921
and admitted to facts consistent with what was determined by the
212
00:14:01,421 --> 00:14:04,823
investigation. An investigation revealed communication between
213
00:14:04,951 --> 00:14:08,291
Kenny and an inmate described as being in a sexual nature,
214
00:14:08,709 --> 00:14:12,707
and there were records of Kenny visiting the inmate using his attorney credentials.
215
00:14:12,899 --> 00:14:16,320
It's important to note that the defendant James Kenny was
216
00:14:16,352 --> 00:14:19,884
not legal representation for any of the inmates involved in this
217
00:14:20,384 --> 00:14:24,123
matter. The sheriff's office says one of those visits in early August involved
218
00:14:24,623 --> 00:14:28,217
sexual activity, and he allegedly traded the tobacco and vape pens for
219
00:14:28,717 --> 00:14:32,697
sex acts, which the prosecutor says is an even bigger problem. There's concern
220
00:14:32,874 --> 00:14:37,084
for Well, the court also
221
00:14:37,584 --> 00:14:41,150
discussing a previous conviction from 2004 for sexual imposition.
222
00:14:41,586 --> 00:14:45,120
The investigation into Kenny isn't over just yet, and more charges
223
00:14:45,217 --> 00:14:49,167
could soon be filed. Today he was given a $150,000 cash
224
00:14:49,200 --> 00:14:54,705
surety bond. Chris Jacobs, WLWT News 5.
225
00:14:54,833 --> 00:14:58,391
Last— the last time, Alex, that we had a judge situation like
226
00:14:58,455 --> 00:15:01,939
this, it was actually sexual in nature as well. You may recall it was
227
00:15:02,439 --> 00:15:05,602
a female judge who was caught having sex with chief of police inside of her
228
00:15:06,102 --> 00:15:09,677
chambers. That was, that was the previous, uh,
229
00:15:10,403 --> 00:15:13,629
incarnation. So apparently, apparently judges are really, uh,
230
00:15:14,516 --> 00:15:17,824
keen on having some sex scandals behind them when, when they do get in
231
00:15:18,324 --> 00:15:20,972
trouble for something. Are you, uh,
232
00:15:20,988 --> 00:15:23,876
previewing, uh, your next series of,
233
00:15:24,168 --> 00:15:27,489
uh, female judges getting frisky? Uh,
234
00:15:27,859 --> 00:15:31,368
there, to my knowledge, to my knowledge, there is no rampant, um,
235
00:15:32,793 --> 00:15:36,440
phenomenon of judges, female judges having sex with
236
00:15:37,312 --> 00:15:40,232
people they're not supposed to in places they're not supposed to.
237
00:15:41,892 --> 00:15:46,325
I will keep a lookout for that, but I haven't seen such an organic trend
238
00:15:46,422 --> 00:15:50,282
like I did the female teachers who were sleeping with students
239
00:15:50,331 --> 00:15:54,343
trend. Maybe look into it as a control. That's a good point.
240
00:15:54,504 --> 00:15:57,980
Well, it might actually be a consequence, Alex,
241
00:15:58,126 --> 00:16:03,481
of the patriarchy because fewer females are expected to be judges in
242
00:16:03,530 --> 00:16:07,347
high position like that. So you would— I see. Yes, you'd have a smaller
243
00:16:07,847 --> 00:16:11,904
pool to pull from. That's right. So, you know,
244
00:16:12,532 --> 00:16:15,883
now, Alex, okay, at the beginning of,
245
00:16:15,979 --> 00:16:19,491
uh, this, this show, I say that we are a— we are on
246
00:16:19,603 --> 00:16:23,393
Twitter Spaces, that we're a Twitter show. And I, I say that because
247
00:16:23,893 --> 00:16:27,281
I've never really adopted calling this place X.
248
00:16:27,604 --> 00:16:30,827
Okay, so it's always been Twitter, and in my opinion,
249
00:16:30,860 --> 00:16:34,204
it'll always be Twitter. Uh, Elon's branding
250
00:16:34,301 --> 00:16:38,026
has never caught on to me linguistically. I don't know about you.
251
00:16:38,170 --> 00:16:40,707
Do you refer to this place as Twitter, or do you call it X when,
252
00:16:40,900 --> 00:16:44,111
when it's casually— when you're not thinking about it? I,
253
00:16:44,416 --> 00:16:47,804
I mix and match. Um, I, I work— I don't know.
254
00:16:48,109 --> 00:16:51,738
Yeah, I— some— there's some linguistic constructions where Twitter
255
00:16:52,238 --> 00:16:53,970
just works and X just doesn't. So,
256
00:16:54,050 --> 00:16:58,016
okay, well, I do on occasion dabble and
257
00:16:58,787 --> 00:17:02,092
Twitter discourse. Well, depending on how the
258
00:17:02,592 --> 00:17:05,285
future unfolds, that might become more of a problem, Alex.
259
00:17:05,782 --> 00:17:09,280
New Twitter launches, said Musk's X gave up the name.
260
00:17:09,857 --> 00:17:13,483
Operation Bluebird, the Virginia-based startup trying to revive
261
00:17:13,983 --> 00:17:17,109
the allegedly abandoned Twitter name and logo, announced Monday that
262
00:17:17,609 --> 00:17:20,928
it has launched a new social media network, Twitter.Now. We are
263
00:17:21,428 --> 00:17:24,457
a small company and we have investors and we have a product, Steven Coates,
264
00:17:24,538 --> 00:17:27,345
one of Operation Bluebird's co-founders, told Ars Technica.
265
00:17:27,907 --> 00:17:30,650
And we have waited months and months to launch, and we are not going to
266
00:17:31,150 --> 00:17:34,147
wait anymore. As Ars reported in December 2025,
267
00:17:34,163 --> 00:17:37,772
Coates and his coterie believe that when Elon Musk bought Twitter
268
00:17:37,788 --> 00:17:41,317
in 2022 and quickly changed the company's name to X, the Twitter
269
00:17:41,817 --> 00:17:44,974
identity and intellectual property were abandoned. Musk's decision,
270
00:17:45,086 --> 00:17:49,434
they claim, created the opening for Operation Bluebird to claim Twitter's trademarks.
271
00:17:49,738 --> 00:17:53,171
But Musk's X Corporation quickly sued Operation Bluebird late
272
00:17:53,671 --> 00:17:57,037
last year and asked a federal judge in Delaware to issue a preliminary injunction to
273
00:17:57,537 --> 00:18:01,177
prevent a new version of Twitter from launching. However, in April 2026 hearing,
274
00:18:01,530 --> 00:18:05,077
U.S. District Court Judge Colm Connolly gave a tentative ruling from
275
00:18:05,125 --> 00:18:09,041
the bench that the— that X seemed to have given up intellectual property claims over
276
00:18:09,057 --> 00:18:12,588
the word tweet and the Twitter bird logo, and possibly even the word
277
00:18:13,088 --> 00:18:16,311
Twitter as well. The judge has yet to issue a written order. For Coates,
278
00:18:16,360 --> 00:18:19,586
who serves as Twitter's new general counsel and previously worked in
279
00:18:20,086 --> 00:18:23,245
the same role during the pre-Musk era, the judge's words
280
00:18:23,261 --> 00:18:26,644
were enough to serve as a green light. Quote, it's our opinion that X
281
00:18:27,144 --> 00:18:30,746
has abandoned the rights on the Twitter and tweet trademarks, he said. Still, the new
282
00:18:31,246 --> 00:18:34,655
Twitter takes great pains to put daylight between itself and previous iterations of
283
00:18:35,155 --> 00:18:38,484
Twitter. Quote, Operation Bluebird Incorporated picked up the name
284
00:18:38,740 --> 00:18:42,025
X Corp walked away from. It's rebuilding it on trust in
285
00:18:42,525 --> 00:18:45,614
your browser at Twitter. Now, it states prominently on its website, we are not X.
286
00:18:45,934 --> 00:18:49,331
We are not affiliated with X Corp. Neither the X Corporation nor
287
00:18:49,831 --> 00:18:53,064
Elon Musk nor Andrew Mayo, one of X's attorneys, immediately responded to Ars
288
00:18:53,564 --> 00:18:56,975
Technica's request for comment. Twitter.Now is still in its nascent stage,
289
00:18:57,472 --> 00:19:00,759
only has hundreds of users for the time being. The social media network
290
00:19:01,259 --> 00:19:04,222
looks and feels much like the Twitter of old and many of its offshoots.
291
00:19:04,623 --> 00:19:08,375
It has replies and retweets. A new and notable feature
292
00:19:08,875 --> 00:19:12,319
is the automated fact-checking tool, a Gemini-based veracity engine for
293
00:19:12,819 --> 00:19:16,167
real-time analysis— Vera for short— which runs on every tweet.
294
00:19:16,632 --> 00:19:20,464
Coates has been testing Vera in recent days by posting obviously false messages
295
00:19:20,560 --> 00:19:24,521
like George Washington was our second president. Our first
296
00:19:25,021 --> 00:19:27,537
goal is to see if we can truly bring back a town square that's safer
297
00:19:27,617 --> 00:19:31,515
and less harmful. We say freedom of speech and not freedom of
298
00:19:31,627 --> 00:19:34,466
reach. We want people to say what they want, but we also want to create
299
00:19:34,966 --> 00:19:38,894
a platform that's not financially locked into that viral content that's harmful or
300
00:19:39,006 --> 00:19:43,048
inaccurate. Josh Gerben, a Washington, D.C.-based trademark
301
00:19:43,548 --> 00:19:47,138
attorney who has been closely following the case, told Ars that while Operation Bluebird may
302
00:19:47,267 --> 00:19:50,828
have a workable legal theory that X Corporation has abandoned the Twitter trademark
303
00:19:51,228 --> 00:19:54,146
It's not an open and shut case. Quote, I would say there's an element of
304
00:19:54,162 --> 00:19:57,256
chutzpah to do this, and they're going to get the the best
305
00:19:57,368 --> 00:20:01,632
fight that X has. He said, quote, it's one thing to file the trademark application
306
00:20:02,132 --> 00:20:05,848
now that they've launched. There will be another response from X Corporation. We really
307
00:20:06,201 --> 00:20:09,214
they've really upped the ante now. It's a really interesting escalation of the case.
308
00:20:09,326 --> 00:20:11,795
Go put another round of popcorn in the microwave, Alex.
309
00:20:12,837 --> 00:20:17,021
I I don't think I will be venturing over to this new Twitter or Twitter
310
00:20:17,521 --> 00:20:20,932
dot now. I I did previously every single time there was a Twitter
311
00:20:21,432 --> 00:20:24,542
competitor during the contention. days, make an account. Um,
312
00:20:24,671 --> 00:20:27,952
but I, I only, uh, made posts about food,
313
00:20:28,290 --> 00:20:31,555
but not even really compelling posts about food. I just named food items.
314
00:20:31,651 --> 00:20:35,300
That was really my only contribution. So I had a Parler account,
315
00:20:35,769 --> 00:20:38,905
uh, where I, you know, like, well, I think my first post was like spaghetti.
316
00:20:41,766 --> 00:20:44,985
And, uh, I think I might have a Truth Social account. It's important to contribute
317
00:20:45,485 --> 00:20:49,255
when you join a new ecosystem, for sure. It's important to bring some— bring
318
00:20:49,287 --> 00:20:52,647
something. Even spaghetti. Well,
319
00:20:52,777 --> 00:20:56,288
Alex, uh, you're making an effort. In the last few weeks, we,
320
00:20:56,320 --> 00:21:00,095
we have had a dearth of, uh, requests to speak,
321
00:21:00,579 --> 00:21:04,580
and, and while we have one right now, uh, that's not
322
00:21:04,693 --> 00:21:07,281
really at the halfway mark or towards the end of the show, I'm gonna let
323
00:21:07,781 --> 00:21:11,121
it go up, but we're taking a risk because the person who's requested to speak's
324
00:21:11,621 --> 00:21:14,543
handle is Howling Bunghole. So, uh,
325
00:21:14,752 --> 00:21:17,789
we're gonna take a gamble here on Howling Bunghole to see what
326
00:21:17,838 --> 00:21:21,969
Howling Bunghole has to say. Howling Bunghole, I've approved
327
00:21:22,469 --> 00:21:24,708
your request to speak. If you are up here, go ahead and comment.
328
00:21:28,171 --> 00:21:32,940
All right, he's not here yet, so we will give him a moment to get
329
00:21:33,440 --> 00:21:37,092
on board, but we will move forward with the next, the next
330
00:21:37,140 --> 00:21:40,588
story, and then once that story ends, we'll see if Howling Bunghole has
331
00:21:40,684 --> 00:21:44,224
any thoughts on the matter. The US-Canada
332
00:21:44,724 --> 00:21:48,008
trade war taking a sweeping turn after President Trump hit Canada with 50% tariffs.
333
00:21:48,073 --> 00:21:51,303
Canada now striking back tonight Canada fired back tonight,
334
00:21:51,351 --> 00:21:55,326
imposing new tariffs on 700 different American goods sold
335
00:21:55,439 --> 00:21:57,290
in Canada. Here's Rachel Scott.
336
00:21:59,244 --> 00:22:02,569
Tonight, after President Trump imposed steep new tariffs on Canadian goods,
337
00:22:02,924 --> 00:22:06,927
Canada now firing back with tariffs of up to 50% on
338
00:22:06,959 --> 00:22:10,083
some 700 American products sold in Canada.
339
00:22:10,260 --> 00:22:14,052
Canada will match the United States tariffs dollar for dollar,
340
00:22:14,743 --> 00:22:18,436
rate for rate. Canada doubling its tariffs on American goods,
341
00:22:18,469 --> 00:22:21,921
including steel, aluminum, furniture, and clothing. Here in
342
00:22:21,953 --> 00:22:25,676
the U.S., President Trump's tariffs on Canadian goods now forcing small
343
00:22:25,725 --> 00:22:29,303
businesses like Jesse Prager's hardware store in Philadelphia to
344
00:22:29,319 --> 00:22:33,106
raise prices. 2x4 went from $5 up to— I
345
00:22:33,606 --> 00:22:36,958
think I'm now at $12 to $14 for one 2x4. President Trump
346
00:22:37,458 --> 00:22:40,456
today fuming Canada has been ripping off the USA for decades,
347
00:22:40,873 --> 00:22:44,259
adding, I deal with many countries and Canada is easily the most
348
00:22:44,759 --> 00:22:48,127
difficult and unreasonable. The president later posting, the United States
349
00:22:48,627 --> 00:22:52,203
is giving serious consideration to changing the name of Lake Ontario to
350
00:22:52,703 --> 00:22:56,426
Lake America. Canadian officials unfazed. We'll always call
351
00:22:56,538 --> 00:23:00,183
it Lake Ontario. Great Lakes, period.
352
00:23:01,500 --> 00:23:05,016
And David, amid the president's new trade war with Canada, a new report shows
353
00:23:05,128 --> 00:23:08,597
all the tariffs that he has imposed on goods from other countries is really taking
354
00:23:09,097 --> 00:23:12,386
a toll on American families, costing them nearly $1,100 more
355
00:23:12,835 --> 00:23:16,127
a year. David, Rachel Scott live at the White House again tonight. Rachel, thanks.
356
00:23:16,448 --> 00:23:19,708
Tonight with the Tonight, President Trump escalating
357
00:23:19,724 --> 00:23:23,401
his trade war with Canada, signing a new executive order renaming
358
00:23:23,577 --> 00:23:25,761
Lake Ontario Lake America.
359
00:23:26,146 --> 00:23:30,257
Canada's Prime Minister and some American officials pushing back. Here's ABC's
360
00:23:30,757 --> 00:23:34,640
Mary Bruce. Tonight, President Trump escalating his trade
361
00:23:34,672 --> 00:23:38,446
war with Canada, signing an executive order to rename Lake
362
00:23:38,946 --> 00:23:42,797
Ontario. We're going to be changing the name of Lake Ontario.
363
00:23:42,910 --> 00:23:46,885
I'm renaming Lake Ontario effective immediately to Lake America.
364
00:23:47,093 --> 00:23:51,085
The president fuming since Ontario Premier Doug Ford unloaded on
365
00:23:51,101 --> 00:23:54,612
him when the trade deal fell apart. Well, he can kiss my ass as far
366
00:23:54,708 --> 00:23:58,331
as I'm concerned. Today, Canada's Prime Minister Mark Carney noting the
367
00:23:58,831 --> 00:24:02,306
name of the lake dates back more than 400 years, adding,
368
00:24:02,482 --> 00:24:06,041
this lake is called Lake Ontario today and
369
00:24:06,089 --> 00:24:10,161
forever. The lake borders New York, whose Governor Kathy Hochul bluntly
370
00:24:10,257 --> 00:24:14,569
rejected the president's new name, posting, New York won't be calling
371
00:24:14,601 --> 00:24:18,192
it that. It comes as Commerce Secretary Howard Lutnick spent the
372
00:24:18,692 --> 00:24:21,960
day in court watching his administration lawyers try to convince a judge
373
00:24:21,976 --> 00:24:25,487
to allow President Trump's name to go up on the Kennedy Center,
374
00:24:25,807 --> 00:24:29,415
telling reporters it would motivate the president to raise money for
375
00:24:29,915 --> 00:24:32,701
renovations. You need to induce him,
376
00:24:33,022 --> 00:24:36,533
entice him, get him to agree to wear it and do
377
00:24:36,629 --> 00:24:38,874
it. And if he didn't do it, you could always take the letters down.
378
00:24:39,114 --> 00:24:42,946
The administration has even threatened to demolish the Kennedy Center entirely
379
00:24:43,234 --> 00:24:46,862
if Trump's name doesn't go up. Reporters asking Lutnick if
380
00:24:47,362 --> 00:24:50,864
they have that authority. I have no idea. No ruling yet
381
00:24:50,961 --> 00:24:51,673
from the judge.
382
00:24:52,770 --> 00:24:55,875
And so, Alex,
383
00:24:56,602 --> 00:25:00,209
the trade wars are going well. We have— we're winning them so,
384
00:25:00,726 --> 00:25:04,771
so much that we get to just arbitrarily rename the
385
00:25:04,835 --> 00:25:08,104
lakes that we have. That's— we're
386
00:25:08,169 --> 00:25:12,016
doing— at least America can Rest
387
00:25:12,065 --> 00:25:15,051
assured, it is fighting this battle of maturity. Yes,
388
00:25:15,356 --> 00:25:18,936
yes, this is true. You know, it's being the bigger country,
389
00:25:18,968 --> 00:25:22,324
so to speak. Uh, fellas. Yes,
390
00:25:22,500 --> 00:25:26,000
yes. Howling Bungle. Yes, Howling Bungle. Go ahead. Yes, I have
391
00:25:26,016 --> 00:25:29,709
a solution to this. Is that, uh, well, we take Doug Ford and
392
00:25:30,209 --> 00:25:34,397
then we get, uh, former DC Mayor Marion Barry, uh, to basically
393
00:25:34,542 --> 00:25:37,287
fight this out. It'll be crackhead against crackhead.
394
00:25:37,994 --> 00:25:41,931
Well, that— no, now you're thinking of Rob Ford, sir. You're thinking of Rob
395
00:25:42,431 --> 00:25:46,221
Ford. Doug Ford. All those Leiferts look the same. Now, now,
396
00:25:46,430 --> 00:25:49,531
sir, sir, you know, Rob Ford is near and dear to my
397
00:25:49,547 --> 00:25:53,308
heart because Rob Ford was a partier. Rob Ford partied hard.
398
00:25:53,886 --> 00:25:57,807
And when it's time to party, you will party hard. And that's— that is an
399
00:25:57,839 --> 00:26:00,747
Andrew WK American mantra that we take to heart here.
400
00:26:00,892 --> 00:26:02,788
So don't confuse your Fords.
401
00:26:04,027 --> 00:26:07,244
But, uh, oh, one thing,
402
00:26:07,519 --> 00:26:10,979
are you guys tracking the bad blood that exists between Laura
403
00:26:11,479 --> 00:26:14,654
Loomer and Milo Yiannopoulos? I wouldn't call it tracking,
404
00:26:14,914 --> 00:26:18,468
no. Uh, we did just discuss though the— his deportation.
405
00:26:19,947 --> 00:26:22,940
Oh, uh, get this, is that a while ago,
406
00:26:23,037 --> 00:26:25,868
uh, I think it was maybe like last year, uh,
407
00:26:26,903 --> 00:26:30,423
Milo went kind of off the Well, his, uh,
408
00:26:30,552 --> 00:26:33,649
cheese kind of slipped off the cracker and he started smearing this, uh,
409
00:26:36,037 --> 00:26:39,652
rather embarrassing gossip about Laura Loomer, claiming that she
410
00:26:40,152 --> 00:26:43,366
wasn't really Jewish, which I find that hard to believe. I mean,
411
00:26:43,382 --> 00:26:46,692
you could look at that nose and realize that Milo is not exactly
412
00:26:47,192 --> 00:26:50,325
being honest about that. Which, which version of the nose? Because I'm sure
413
00:26:50,341 --> 00:26:54,572
she's on her like 5th version right now. The 5.0 nose.
414
00:26:54,750 --> 00:26:57,727
5.0. Okay. All right.
415
00:26:58,796 --> 00:27:02,927
All right. Now, the really juicy bit is, is that he went
416
00:27:03,427 --> 00:27:06,934
around saying, and he said this on X, formerly known as Twitter,
417
00:27:07,867 --> 00:27:11,455
that she would get drunk and then sexually
418
00:27:11,648 --> 00:27:15,513
harass interns. And that
419
00:27:16,013 --> 00:27:20,102
one of the interns claimed that her— well, let's just— I
420
00:27:20,246 --> 00:27:23,096
want to keep things clean— that her hoo-hoo tasted like wet bread.
421
00:27:23,948 --> 00:27:27,461
And if Laura Loomer is listening to this, I just want to let you know
422
00:27:27,995 --> 00:27:29,581
I really don't mind the taste of wet bread.
423
00:27:32,183 --> 00:27:36,547
Howling Mung Hole, we took a risk and then you certainly honored
424
00:27:37,212 --> 00:27:40,256
us taking that risk with that last, that last comment there, sir. Uh,
425
00:27:40,786 --> 00:27:44,706
if I will say, uh, I, I, I speak for myself,
426
00:27:44,722 --> 00:27:49,060
but I think I also speak for Alex, that, uh, if, if you want Laura
427
00:27:49,560 --> 00:27:53,139
Loomer You, you will not get a fight from either of us. Uh,
428
00:27:53,411 --> 00:27:57,369
you can have her. She's all yours, sir. Uh, you will
429
00:27:57,869 --> 00:28:01,439
not receive any competition or backstabbing from, uh, us up here.
430
00:28:01,679 --> 00:28:04,675
And your godspeed, sir. That's, that's all.
431
00:28:06,005 --> 00:28:09,899
That's all you. Well, thanks for just letting
432
00:28:10,399 --> 00:28:13,247
me get my 2 cents in. All right, sir. Have a good day.
433
00:28:13,840 --> 00:28:17,109
Um, Alex, moving on to the additional news here that we have.
434
00:28:17,622 --> 00:28:21,442
We have an interesting move in
435
00:28:21,554 --> 00:28:24,797
the crypto space, Alex, the crypto/finance space.
436
00:28:25,647 --> 00:28:29,387
Better and Coinbase announced general availability
437
00:28:29,419 --> 00:28:33,576
of the first token-backed conforming mortgage to expand homeownership
438
00:28:34,076 --> 00:28:36,851
access for a new generation of mortgage borrowers.
439
00:28:37,268 --> 00:28:41,377
From Business Wire, Better Mortgage and Coinbase today
440
00:28:41,441 --> 00:28:44,876
announced general availability for Better's token-backed conforming
441
00:28:44,989 --> 00:28:49,142
mortgage product, These first-of-its-kind mortgages will be originated
442
00:28:49,642 --> 00:28:53,422
and serviced by Better, powered by Coinbase, and designed in accordance
443
00:28:53,470 --> 00:28:57,654
with guidelines from Fannie Mae, making the first lien a standard
444
00:28:57,894 --> 00:29:01,292
conforming mortgage. As part of this offering, if approved for
445
00:29:01,405 --> 00:29:04,755
a loan by Better, Coinbase One members will be eligible
446
00:29:05,172 --> 00:29:08,762
for a rebate equal to 1% of the mortgage value, up to a maximum of
447
00:29:08,795 --> 00:29:12,594
$10,000. Rebates are applied as a lender credit against
448
00:29:12,642 --> 00:29:16,617
closing costs, and any rebates will be paid by Better, reflected directly
449
00:29:16,729 --> 00:29:19,987
on the borrower's closing disclosure. In addition
450
00:29:20,099 --> 00:29:23,534
to opening the product to general availability, the companies have also announced
451
00:29:24,034 --> 00:29:27,548
that its 1% rebate will extend to all of Better's home finance products,
452
00:29:27,564 --> 00:29:30,823
including standard mortgages, home equity lines of credit balances,
453
00:29:31,160 --> 00:29:34,531
and refinances. The product launch builds on strong early demand
454
00:29:35,031 --> 00:29:38,528
for Better and Coinbase's original crypto-backed mortgage product. After opening
455
00:29:38,624 --> 00:29:41,996
a waitlist for that offering in June of this year, the companies found
456
00:29:42,496 --> 00:29:46,218
that 76% of respondents were already Coinbase One users and 60% said
457
00:29:46,250 --> 00:29:49,812
they planned to buy within 6 months. Waitlist data also revealed
458
00:29:49,956 --> 00:29:53,773
$260 million in projected loan volume before general
459
00:29:53,789 --> 00:29:57,461
availability. Better and Coinbase are now scaling that early
460
00:29:57,894 --> 00:30:01,342
success, bringing traditional mortgage and HELOC offerings to Coinbase's
461
00:30:01,842 --> 00:30:04,903
broader membership. This partnership has always been
462
00:30:04,967 --> 00:30:08,431
about expanding access to homeownership by meeting borrowers where they are,
463
00:30:08,896 --> 00:30:12,296
said Ziggy Johnson, Chief Technology Officer at Better Mortgage.
464
00:30:12,889 --> 00:30:16,257
In 2025, high interest rates, record home prices, and limited
465
00:30:16,757 --> 00:30:19,912
inventory pushed the median age of a first-time home buyer to 40.
466
00:30:20,489 --> 00:30:24,015
Coinbase counts millions of monthly users worldwide, and by allowing
467
00:30:24,515 --> 00:30:28,245
Coinbase One members to pledge crypto as collateral without selling their holdings,
468
00:30:28,662 --> 00:30:32,300
we're opening a new path forward, uh, path towards homeownership
469
00:30:32,604 --> 00:30:36,098
for a generation of borrowers whose wealth increasingly lives on-chain.
470
00:30:37,075 --> 00:30:40,905
Our members already trust Coinbase for their financial lives. This extends
471
00:30:40,921 --> 00:30:44,416
that trust to one of the biggest financial decisions they'll ever make said Ben
472
00:30:44,545 --> 00:30:47,919
Shen, head of financial services and loyalty products at Coinbase.
473
00:30:48,240 --> 00:30:51,839
By enabling borrowers to pledge their digital assets in the mortgage underwriting
474
00:30:52,339 --> 00:30:55,599
process, we are allowing crypto to be more useful and powerful in the real
475
00:30:55,695 --> 00:30:59,053
world, expanding the pathways to homeownership while preserving
476
00:30:59,150 --> 00:31:03,536
long-term investment positions. The expanded offer became available
477
00:31:03,568 --> 00:31:06,974
to Coinbase One members applying for eligible Better Mortgage products,
478
00:31:07,280 --> 00:31:12,077
including standard mortgages, HELOCs, and refinances on
479
00:31:12,255 --> 00:31:16,041
August 12th, 2026. Consumers can learn more and check availability
480
00:31:16,236 --> 00:31:20,100
at better.com/crypto-backed-mortgages. So,
481
00:31:20,357 --> 00:31:23,634
Alex, we get to see
482
00:31:23,650 --> 00:31:26,848
how that plays out in the near future, I guess. I don't know how much—
483
00:31:28,085 --> 00:31:30,628
maybe it said it and I just missed it, but I don't know how much
484
00:31:31,049 --> 00:31:34,895
of your crypto holdings you can pledge and
485
00:31:34,944 --> 00:31:38,918
exactly how that How much are
486
00:31:39,289 --> 00:31:42,889
you thinking to pledge of your crypto holdings, Kevin? Well,
487
00:31:42,938 --> 00:31:46,357
not mine specifically, but in an average transaction,
488
00:31:46,405 --> 00:31:50,360
right, the crypto holdings
489
00:31:50,408 --> 00:31:54,491
can't possibly be the only thing in consideration for that underwriting
490
00:31:54,524 --> 00:31:58,910
process, right? It's got to have to be a fraction of
491
00:31:58,942 --> 00:32:03,007
consideration because one, the volatility
492
00:32:03,344 --> 00:32:07,333
on on Bitcoin even is just too high to
493
00:32:07,462 --> 00:32:10,997
really securitize a loan betting off
494
00:32:11,497 --> 00:32:14,234
of that as collateral. You can't just— you can't do it. That's not going to
495
00:32:14,734 --> 00:32:16,608
be— That's right. You need a stable asset like the dollar.
496
00:32:18,362 --> 00:32:21,116
Well, Alex. Well, Alex, Alex, Alex.
497
00:32:22,211 --> 00:32:26,333
The stability of the dollar is very stable within the dollar's
498
00:32:26,833 --> 00:32:31,197
ecosystem. Okay? It is not stable necessarily across If
499
00:32:31,697 --> 00:32:34,395
you assume the dollar as your unit, then it doesn't move at all, really.
500
00:32:34,604 --> 00:32:37,625
Exactly. Exactly. It is, it is parity with
501
00:32:38,125 --> 00:32:42,165
the dollar. Okay. And also a parody. Ah, that was
502
00:32:42,264 --> 00:32:42,660
a lot.
503
00:32:45,700 --> 00:32:48,762
Yes. Anyway, moving on,
504
00:32:49,037 --> 00:32:52,342
moving past that. So, so wait, so the, the whole deal with this is
505
00:32:52,423 --> 00:32:56,047
that they could accept your crypto as part of your collateral?
506
00:32:56,177 --> 00:32:59,669
Is that the— Yeah. It's my understanding. I just don't know. I just, I just
507
00:33:00,169 --> 00:33:03,357
don't know what— yeah, yeah, it's the— that is the innovation, so to speak.
508
00:33:03,518 --> 00:33:06,690
Um, but I guess that somehow will also be tokenized.
509
00:33:07,301 --> 00:33:11,237
I mean, that, that just seems like a formality, not less than an actual achievement
510
00:33:11,737 --> 00:33:13,695
of some kind. Um, but wow,
511
00:33:15,607 --> 00:33:18,455
you wrote, you wrote a thing on the blockchain.
512
00:33:20,220 --> 00:33:23,899
Amazing. I, I'm wondering, like, you know, You have
513
00:33:24,399 --> 00:33:28,820
a volatile asset as collateral
514
00:33:29,320 --> 00:33:33,000
backing for a mortgage on an asset that,
515
00:33:34,292 --> 00:33:37,763
generically speaking, historically is an increasing asset,
516
00:33:37,925 --> 00:33:41,112
but then also has had time periods where
517
00:33:41,385 --> 00:33:44,825
it has collapsed. And I'm wondering
518
00:33:45,242 --> 00:33:49,586
what exactly happens when you increase the expansion of
519
00:33:49,619 --> 00:33:53,671
having these crypto-backed mortgage assets Where God
520
00:33:53,687 --> 00:33:56,876
forbid they, they, they do a very, very strong connection,
521
00:33:56,908 --> 00:34:00,962
meaning that you're going to get something like a 1-to-1 collateral
522
00:34:01,539 --> 00:34:04,567
loan based off of the, the market price at signing.
523
00:34:05,240 --> 00:34:08,189
And then what happens when housing prices tank,
524
00:34:08,589 --> 00:34:11,746
you lose your job, and then the crypto assets also tank?
525
00:34:11,810 --> 00:34:14,117
Like, what happens in that circle?
526
00:34:15,415 --> 00:34:18,620
Well, look, I mean, I guess the house value also
527
00:34:19,120 --> 00:34:22,450
goes down, so it'll Goes together. But the loan— the loan
528
00:34:22,950 --> 00:34:26,761
stayed— the loan amount didn't stay the same. You know, the collateral goes down,
529
00:34:26,793 --> 00:34:29,885
the asset goes down, but the loan balance stays the same. You'd have
530
00:34:29,917 --> 00:34:33,587
to go— you'd have a huge fight for refinancing at that point,
531
00:34:33,603 --> 00:34:37,449
but then you've already, you know, collateralized your debt for that.
532
00:34:37,545 --> 00:34:40,911
So the, the amount— the amount you're borrowing— you're borrowing is not,
533
00:34:41,840 --> 00:34:45,109
um, denominated in Bitcoin then?
534
00:34:45,606 --> 00:34:49,696
Right. It's dollars. Right. Like, I'm sorry, I'm just
535
00:34:50,196 --> 00:34:53,593
asking for a clarification. I mean, I'd imagine, I'd imagine since they're not,
536
00:34:54,106 --> 00:34:57,938
they're not accepting and they're not pricing homes in Bitcoin,
537
00:34:58,131 --> 00:35:01,547
they're pricing them in, you know, dollars. I mean,
538
00:35:01,627 --> 00:35:05,428
how do you know? This could get, this could get really fancy.
539
00:35:05,748 --> 00:35:10,078
Now, if it was, if it was in Bitcoin, that would be fascinating to
540
00:35:10,578 --> 00:35:14,350
the degree that it's a small Proportion, I'm sure like large
541
00:35:14,382 --> 00:35:17,646
financial institutions can effectively consider this a
542
00:35:17,695 --> 00:35:22,060
type of what's
543
00:35:22,560 --> 00:35:26,397
the word diversification and and you know eat any potential
544
00:35:26,414 --> 00:35:30,041
losses. The only problem comes if you
545
00:35:30,074 --> 00:35:34,603
know this kind of mortgages are a large proportion of
546
00:35:35,593 --> 00:35:37,427
the overall pool, which seems.
547
00:35:40,962 --> 00:35:44,463
Still in the future, let's just say. So I guess we'll
548
00:35:44,963 --> 00:35:48,432
find out. I'm just wondering how much
549
00:35:48,448 --> 00:35:51,969
weight they're going to give to that collateral. I could understand it being fractional.
550
00:35:54,878 --> 00:35:58,580
Let's say that you could give $10,000
551
00:35:58,741 --> 00:36:02,831
worth of Bitcoin for something along the lines
552
00:36:02,927 --> 00:36:06,347
of a $1,000 Right.
553
00:36:06,444 --> 00:36:09,013
Like, okay. Joking aside now, like,
554
00:36:09,336 --> 00:36:12,373
I'm sure there is a class of person who,
555
00:36:12,970 --> 00:36:16,843
you know, is super into their Bitcoin and has a good
556
00:36:16,941 --> 00:36:20,774
chunk of it. Right. And does not, is not feeling up
557
00:36:21,274 --> 00:36:24,818
for selling it in order to get a mortgage. So,
558
00:36:24,963 --> 00:36:27,881
you know, this will help them, presumably.
559
00:36:29,730 --> 00:36:32,729
These people don't tend to be You know, poor exactly.
560
00:36:32,905 --> 00:36:36,689
So presumably you'd have a non-trivial amount
561
00:36:37,138 --> 00:36:40,119
that falls under that. So I don't know, it seems like a niche product,
562
00:36:40,135 --> 00:36:43,967
I guess, is what I'm saying, uh, that I'm, I'm pretty sure
563
00:36:44,467 --> 00:36:47,862
is not gonna lead to the next financial collapse. And if it happens,
564
00:36:47,878 --> 00:36:50,908
you can use this, you can take this recording. I, I,
565
00:36:52,223 --> 00:36:56,487
for the record, I didn't say it would cause the next collapse. I'm just asking.
566
00:36:56,535 --> 00:36:59,597
I'm just I was just asking what happens in that,
567
00:36:59,661 --> 00:37:03,925
that game of musical chairs during a collapse period.
568
00:37:05,127 --> 00:37:08,814
What happens, Gator? The same thing that happens every time. Oh, they'll get
569
00:37:09,314 --> 00:37:11,972
bailed out. Not— not— no, no, no, no, no, no, no.
570
00:37:12,084 --> 00:37:16,252
Only the— only the banks get bailed out, Alex. The— I
571
00:37:16,268 --> 00:37:20,147
mean, the other— have we checked? Have we checked if Coinbase
572
00:37:20,196 --> 00:37:23,706
is not a bank at this point? It is. I don't
573
00:37:24,206 --> 00:37:27,329
think it is technically a bank, even though it now has bank-like products, but it's
574
00:37:27,393 --> 00:37:30,681
also a casino. No, because it has gambling-like
575
00:37:30,698 --> 00:37:34,621
products. It's an all-in-one
576
00:37:35,299 --> 00:37:38,705
facility right now. You can do sports betting on Coinbase now. You can
577
00:37:39,205 --> 00:37:42,576
do general prediction markets on Coinbase
578
00:37:43,076 --> 00:37:46,149
now. You can do all sorts of tokens. You seem very
579
00:37:46,841 --> 00:37:50,909
familiar with their product offering, I have to say. Well, I do have
580
00:37:50,943 --> 00:37:54,152
Coinbase. I mean,
581
00:37:54,930 --> 00:37:58,170
it's hard to not know what they do when you go onto the app and
582
00:37:58,670 --> 00:38:01,281
you see all of the features. I see.
583
00:38:02,485 --> 00:38:05,132
Maybe Elon bought the wrong everything app after all.
584
00:38:05,710 --> 00:38:09,400
Well, yes, that is correct. Although, although they
585
00:38:09,900 --> 00:38:13,813
don't have, um, they don't have Grok or any Grok-like
586
00:38:13,877 --> 00:38:17,552
products, so they don't have anime porn bots, uh, who will,
587
00:38:17,889 --> 00:38:20,889
who will sing sweet nothings into your ear as you try to go to sleep.
588
00:38:21,274 --> 00:38:24,644
They are missing out on that product. Now that you mentioned that,
589
00:38:24,788 --> 00:38:27,259
you do know that the Companions product, as it was called,
590
00:38:31,123 --> 00:38:34,210
Now, now, why would Al— why would he do that, Alex? It was,
591
00:38:34,420 --> 00:38:36,795
it was probably the most popular thing he's ever done.
592
00:38:38,812 --> 00:38:42,239
Uh, I don't think it was that popular, which is why
593
00:38:42,368 --> 00:38:46,086
it's being phased out, is my guess. Ah, okay. All right. Um, but I,
594
00:38:46,151 --> 00:38:49,156
I find it fascinating, right? Because, um, you know,
595
00:38:49,172 --> 00:38:52,452
someone is telling you to, like, telling, encouraging, let's say,
596
00:38:52,533 --> 00:38:55,951
to, uh, develop an emotional attachment to this type
597
00:38:56,451 --> 00:38:59,755
of, well, product. Um, and now
598
00:38:59,803 --> 00:39:03,093
they're like, ah, you know, well, not enough of you are biting, so we're just
599
00:39:03,593 --> 00:39:07,796
gonna— I will say, I have successfully
600
00:39:07,812 --> 00:39:10,941
avoided developing an emotional attachment to any of the
601
00:39:11,022 --> 00:39:13,798
LLM-related, uh, products that are out there.
602
00:39:14,954 --> 00:39:18,501
Me too. Me too. They need to work harder. Um, I, I did not.
603
00:39:19,383 --> 00:39:23,154
I did develop an emotional attachment to the voice of Peter
604
00:39:23,654 --> 00:39:27,583
Cullen as Optimus Prime. And I did do a
605
00:39:28,083 --> 00:39:31,754
fair amount of sulking in the past over the passing of Peter Cullen this week.
606
00:39:32,300 --> 00:39:36,039
But I have not gotten any emotional connection to any
607
00:39:36,456 --> 00:39:39,232
LLM or talking LLM or anything similar.
608
00:39:40,708 --> 00:39:44,543
Give it time. Give it time, Gator. Maybe these
609
00:39:44,575 --> 00:39:48,551
2 things can come together somehow and Provide
610
00:39:49,051 --> 00:39:53,029
you with an AI Peter Cullen. No, no, do not disrespect Peter Cullen's voice
611
00:39:53,400 --> 00:39:57,138
by turning him into an AI. No, uh, there, there is a—
612
00:39:57,170 --> 00:39:59,220
authenticity matters in these circles, Alex.
613
00:40:00,882 --> 00:40:03,351
Yes, give it time.
614
00:40:04,125 --> 00:40:07,326
Are you, are you, are you poo-pooing the idea that authenticity matters,
615
00:40:07,666 --> 00:40:10,827
Alex? Well, the problem is that the word
616
00:40:11,327 --> 00:40:13,711
authenticity has become such a commonplace that it— well,
617
00:40:14,188 --> 00:40:17,939
No longer has namesake quality.
618
00:40:18,164 --> 00:40:21,867
Oh no, oh no. Nothing is authentic and everything is authentic. Is that where
619
00:40:22,367 --> 00:40:25,554
we are now, Alex? Is that where you are this week? Yes, that's where we
620
00:40:26,054 --> 00:40:29,338
are. Okay. All right. Well, Alex,
621
00:40:29,402 --> 00:40:32,416
I don't know if you've heard this. This is the next story we're going to
622
00:40:32,432 --> 00:40:36,423
go into here in the smorgasbord of early news stories
623
00:40:36,520 --> 00:40:38,668
that are disconnected rather largely.
624
00:40:40,050 --> 00:40:44,149
For the last several months, there has been a mystery of sightings
625
00:40:44,262 --> 00:40:47,618
in New York City of people who have been disappearing
626
00:40:47,731 --> 00:40:51,581
into open manhole covers and have
627
00:40:51,597 --> 00:40:54,803
no explanation as to what they're actually doing. Groups of people,
628
00:40:55,596 --> 00:40:58,214
5 or 6 strong, who are opening manhole covers,
629
00:40:58,247 --> 00:41:01,467
popping in and disappearing inside of them doing Lord knows what.
630
00:41:02,095 --> 00:41:05,621
And this has been going on for several months now. Well, there's a recent story
631
00:41:05,975 --> 00:41:08,404
on this phenomenon that's coming out of New York.
632
00:41:09,606 --> 00:41:13,582
Now to the growing manhole mystery, this time on New York's Long Island.
633
00:41:13,630 --> 00:41:16,451
Surveillance images show 5 men opening a manhole,
634
00:41:16,579 --> 00:41:19,946
climbing down, and closing it again. Police searched the
635
00:41:20,446 --> 00:41:24,017
sewer but found no one inside. ABC's Aaron Katersky on the latest
636
00:41:24,517 --> 00:41:26,646
in a series of similar events across New York.
637
00:41:28,025 --> 00:41:30,894
Tonight, a new chapter in a manhole mystery.
638
00:41:31,006 --> 00:41:34,806
Police on Long Island are looking at this video showing these people slipping
639
00:41:34,854 --> 00:41:38,268
into the sewer through a manhole in the wee hours of Thursday morning.
640
00:41:38,380 --> 00:41:41,909
They replaced the COVID once inside. Police said they searched below
641
00:41:41,973 --> 00:41:45,486
ground in Wantagh and found no one and apparently nothing disturbed
642
00:41:45,598 --> 00:41:48,293
or damaged. It's kind of crazy, to be honest with you,
643
00:41:48,630 --> 00:41:51,453
but I heard about it going on in the city. Wow.
644
00:41:52,576 --> 00:41:55,945
In New York City, at least 3 people were spotted just last week climbing out
645
00:41:56,445 --> 00:42:00,228
of a manhole near Penn Station wearing overalls and headlamps. And in June,
646
00:42:00,276 --> 00:42:03,933
surveillance video captured this group exiting a manhole in Brooklyn.
647
00:42:04,045 --> 00:42:08,473
One guy almost struck by a car. The NYPD suspects these underground
648
00:42:08,973 --> 00:42:12,925
scavengers are looking for valuables. But entering manholes is both illegal
649
00:42:13,118 --> 00:42:16,790
and dangerous. This kind of urban scavenging may be happening more frequently
650
00:42:16,886 --> 00:42:19,785
here in New York City, Whit, but on Long Island, police say it may be
651
00:42:19,930 --> 00:42:22,941
the first such incident, and they hope the last. Whit.
652
00:42:25,180 --> 00:42:28,047
Now, Alex, I'm going to— I'm going to go out on a limb here,
653
00:42:28,256 --> 00:42:32,002
and I'm going to blame the nomenclature for this
654
00:42:32,131 --> 00:42:36,122
phenomenon. Now, I didn't hear in any of these reports anything about
655
00:42:36,235 --> 00:42:40,110
women going down these. Manholes. And so I think when
656
00:42:40,126 --> 00:42:43,950
you tell people that these are manholes, that you're going to get some men
657
00:42:44,464 --> 00:42:47,741
wanting to go down the holes. I think that's what's just a natural thing
658
00:42:48,737 --> 00:42:51,484
when you name it something like that for people to want to do. They're going
659
00:42:51,984 --> 00:42:55,355
to want to go down the hole if it's a manhole. When people have their
660
00:42:55,404 --> 00:42:59,179
homes and they build their man caves, well, that's where they naturally traipse
661
00:42:59,679 --> 00:43:02,890
to. They actually go down there and spend time in their man cave. So how
662
00:43:02,954 --> 00:43:04,769
do you get to a man cave? You have to go down the manhole.
663
00:43:06,713 --> 00:43:10,457
I see. Yes. I don't
664
00:43:10,489 --> 00:43:13,799
know. To me, like, the first thing I thought about when they kept talking about
665
00:43:14,299 --> 00:43:18,638
it that way is like, if you wanted to cause this phenomenon, this is about
666
00:43:19,138 --> 00:43:23,057
the kind of story you'd put out. Like, all these
667
00:43:23,089 --> 00:43:25,842
people are going down the manholes. Let's see.
668
00:43:27,435 --> 00:43:29,560
Nobody knows why. I'll go investigate why.
669
00:43:30,252 --> 00:43:34,270
That's right. And now there's a new person down the hole. What are they
670
00:43:34,770 --> 00:43:37,958
looking for? Who knows? Who knows what's down there? Maybe valuable things!
671
00:43:39,923 --> 00:43:43,677
Maybe. Yes, I will
672
00:43:43,726 --> 00:43:46,569
say that that is very— there might be treasure down there!
673
00:43:48,782 --> 00:43:52,299
Yep. Well, there have been
674
00:43:52,347 --> 00:43:56,172
no reports of any chemical companies accidentally
675
00:43:56,269 --> 00:43:59,870
dumping ooze down the sewers, so we might be able to rule out that there
676
00:44:00,370 --> 00:44:04,323
are some Teenage Mutant Turtles that have been
677
00:44:05,093 --> 00:44:08,605
studying karate down there and
678
00:44:09,105 --> 00:44:12,918
getting in trouble. All right, Alex, we have an update now. Gator, I think
679
00:44:13,175 --> 00:44:17,023
the solution for New York before we move on is, of course, some more flock
680
00:44:17,523 --> 00:44:21,160
cameras. I feel that would address the issue. Yes,
681
00:44:21,738 --> 00:44:25,314
yes, yes. Well, Alex, we actually have— we're
682
00:44:25,814 --> 00:44:28,906
going to get to meet the flockers, but we have an update on a
683
00:44:29,406 --> 00:44:33,412
story that's very niche to this show here. It's something we've followed
684
00:44:34,663 --> 00:44:37,889
Somewhat exclusively outside of the area and the country.
685
00:44:37,969 --> 00:44:40,361
It's been this dumb week exclusive. Yes.
686
00:44:41,645 --> 00:44:45,304
Well, it's not exclusive to us, but in the realm of these
687
00:44:45,401 --> 00:44:49,558
types of shows, yes, it is somewhat of an exclusive. And it is the Albuquerque
688
00:44:49,831 --> 00:44:52,286
DWI scheme that was busted.
689
00:44:53,185 --> 00:44:55,978
Oh, yes. Yes, there's a little bit of an update.
690
00:44:56,251 --> 00:45:00,200
Some of the— if you recall last time we discussed that there were
691
00:45:00,585 --> 00:45:05,272
wiretaps involved in bringing down this scheme.
692
00:45:05,320 --> 00:45:09,061
And for those of you who are not familiar, in Albuquerque,
693
00:45:09,254 --> 00:45:12,594
New Mexico, over the last several years, it has been somewhat
694
00:45:12,690 --> 00:45:16,912
revealed that there has been a longtime scheme amongst
695
00:45:16,961 --> 00:45:20,589
the police and some
696
00:45:20,653 --> 00:45:23,897
defense attorneys to entrap people
697
00:45:24,828 --> 00:45:28,601
in DWIs, or DUIs as it's sometimes known across
698
00:45:29,101 --> 00:45:32,245
the country, driving under the influence. And what would happen is, There would be a
699
00:45:32,293 --> 00:45:36,466
person who would be at a bar with an individual sharing drinks
700
00:45:36,966 --> 00:45:40,639
and becoming friendly with someone, and then maybe asking them
701
00:45:41,139 --> 00:45:45,165
for a ride home, getting a ride home from this person. And this
702
00:45:45,855 --> 00:45:49,916
person who is a patsy, they're there to assist
703
00:45:50,189 --> 00:45:53,960
in the scheme. So they get someone as a sucker, that sucker
704
00:45:54,460 --> 00:45:57,459
starts to drive them home, and they get pulled over by the police.
705
00:45:57,748 --> 00:46:01,201
Okay. They get pulled over and clearly they're intoxicated. They have
706
00:46:01,362 --> 00:46:04,526
maybe above the legal limit of alcohol in their system.
707
00:46:05,907 --> 00:46:10,356
Then the officer says that they're going to be arrested and
708
00:46:10,565 --> 00:46:13,392
face fines for this DWI, might lose their license,
709
00:46:13,954 --> 00:46:17,456
and gives them— this buddy who's
710
00:46:17,552 --> 00:46:21,198
enticing them to drive the car gives them a card to
711
00:46:21,246 --> 00:46:24,474
call for an attorney. And this attorney— has a
712
00:46:24,555 --> 00:46:28,570
deal, cut a deal with the police to entice
713
00:46:29,070 --> 00:46:32,939
them to not show up for the first hearing, and the case gets
714
00:46:32,987 --> 00:46:36,489
dropped. And this case gets dropped by paying the attorney
715
00:46:37,340 --> 00:46:40,778
something on the order of $3,000, and the officer
716
00:46:41,278 --> 00:46:44,038
just doesn't show up for the first hearing and the case gets dropped. So this
717
00:46:44,135 --> 00:46:46,930
was a scam that was long-running,
718
00:46:46,962 --> 00:46:50,495
multi-decades, that was recently exposed. And the last time we covered this
719
00:46:50,995 --> 00:46:54,478
several weeks ago, possibly a month ago, It was revealed that the FBI had done
720
00:46:54,590 --> 00:46:58,123
somewhat of an extensive wiretap to uncover a lot
721
00:46:58,623 --> 00:47:01,784
of these details. There were allegations being made, but they
722
00:47:01,816 --> 00:47:04,818
didn't have enough evidence to actually bring the scheme down.
723
00:47:05,348 --> 00:47:08,704
And Alex, this week we actually have some of the
724
00:47:09,204 --> 00:47:13,345
text of some of the discussions captured from those
725
00:47:13,441 --> 00:47:16,331
wiretaps to report with.
726
00:47:19,141 --> 00:47:22,433
For at least three decades, a group of Metro police officers committed
727
00:47:22,933 --> 00:47:26,045
crimes while wearing the badge. Today, federal wiretaps
728
00:47:26,545 --> 00:47:29,722
revealed just how untouchable some of those police officers
729
00:47:29,915 --> 00:47:33,913
thought they were, talking casually about lying on the job,
730
00:47:33,929 --> 00:47:37,494
taking bribes, and targeting victims for their DWI
731
00:47:37,847 --> 00:47:41,428
corruption scheme. KRQE investigative reporter Gabrielle
732
00:47:41,540 --> 00:47:45,121
Burkhardt shows you what those. Wiretaps reveal about the culture
733
00:47:45,393 --> 00:47:49,729
inside the DWI unit and asks current
734
00:47:50,229 --> 00:47:52,876
leadership how public trust can be rebuilt.
735
00:47:53,743 --> 00:47:57,982
Hundreds of pages unsealed in federal court show us how casually
736
00:47:58,078 --> 00:48:01,707
officers talked on the phone about their crimes— crimes the public
737
00:48:02,207 --> 00:48:05,561
would likely never know about if not for citizens speaking up and
738
00:48:06,061 --> 00:48:09,333
for the right people to finally listen. You can do You can do almost
739
00:48:09,833 --> 00:48:13,066
whatever you want in this unit, bro, but don't broadcast it.
740
00:48:13,339 --> 00:48:16,993
Don't tell the whole world. Former Albuquerque police officer
741
00:48:17,025 --> 00:48:20,678
Honorio Alba says to fellow officer Joshua Montaño in
742
00:48:20,710 --> 00:48:24,060
a December 2023 phone call. You would think
743
00:48:24,560 --> 00:48:27,281
that you were listening to other people, just regular criminals, like other,
744
00:48:27,777 --> 00:48:31,447
other investigations that you would be doing. And now we have to almost kind
745
00:48:31,947 --> 00:48:36,291
of pause and say, wait a minute, I'm actually listening to Raul
746
00:48:36,500 --> 00:48:40,466
Bujanda, former special agent in charge of Albuquerque's FBI
747
00:48:40,514 --> 00:48:43,837
field office, was head of the office at the time the FBI
748
00:48:44,030 --> 00:48:47,884
was listening in on the phone calls of certain Albuquerque police
749
00:48:48,384 --> 00:48:52,925
officers, piecing together what we now know was a decades-long corruption
750
00:48:53,102 --> 00:48:57,020
scheme involving officers funneling their DWI arrests to local
751
00:48:57,052 --> 00:49:00,505
defense attorney Thomas Clear III's office. Helping each
752
00:49:01,005 --> 00:49:04,924
other lie about missing court or not filing charges at all—an
753
00:49:05,424 --> 00:49:09,166
extortion scheme to line their own pockets and purposefully botch
754
00:49:09,182 --> 00:49:12,492
criminal cases. And it's like, really? Is this where
755
00:49:12,992 --> 00:49:16,461
we're at? I mean, it was so disappointing. Intercepted conversations
756
00:49:16,493 --> 00:49:20,800
capture a candid and close relationship at least four officers
757
00:49:20,896 --> 00:49:24,656
within the DWI unit had with Ricardo Rick Mendez,
758
00:49:25,010 --> 00:49:27,242
then a. The FBI says the recordings were illegal for that attorney's office.
759
00:49:27,740 --> 00:49:31,111
Conversations showing just how easy it was for certain
760
00:49:31,611 --> 00:49:35,447
police officers to wear the badge and ignore the oath they took
761
00:49:35,947 --> 00:49:39,059
to earn that badge. Biggest takeaway from this investigation?
762
00:49:39,541 --> 00:49:43,314
Biggest takeaway is I wish we would have started sooner. Sooner, because in
763
00:49:43,814 --> 00:49:47,313
2025, the FBI would get a confession from Clear that he
764
00:49:47,393 --> 00:49:50,717
operated this scheme since at least 1995.
765
00:49:51,584 --> 00:49:55,180
I wish that we would have been able to kind of convince the
766
00:49:55,680 --> 00:49:59,291
system that this was something that was actually happening, not because we were
767
00:49:59,791 --> 00:50:03,995
saying it as investigators working for an agency, the FBI back then, but because
768
00:50:04,413 --> 00:50:08,122
it was what was happening. Months before their homes were
769
00:50:08,186 --> 00:50:12,200
raided by the FBI, Alba and Montaño mention a supervisor
770
00:50:12,700 --> 00:50:16,439
at work reminding officers that new arrests had to be booked into custody.
771
00:50:16,873 --> 00:50:20,341
Rather than issuing a court summons, seems standard.
772
00:50:20,469 --> 00:50:24,387
But for Montaño and Alba, issuing a summons instead of booking someone
773
00:50:24,419 --> 00:50:28,096
into jail made it easier for the officers to simply not
774
00:50:28,596 --> 00:50:32,496
follow through with a case. Well, that sucks, dude, because now it destroys
775
00:50:32,672 --> 00:50:36,590
the whole system," Montaño told Alba. In another exchange,
776
00:50:36,847 --> 00:50:39,802
Alba's on the phone with fellow officer Harvey Johnson,
777
00:50:40,219 --> 00:50:44,248
complaining that a judge is asking for. Proof of the doctor appointment Alba
778
00:50:44,748 --> 00:50:48,820
claimed he had to take his dad to. His excuse for missing another court
779
00:50:49,320 --> 00:50:51,756
date in a DWI case that already saw delays.
780
00:50:52,414 --> 00:50:55,526
They're bugging for doctors' excuses, bro, Alba says.
781
00:50:56,056 --> 00:50:59,168
Johnson tells him to just take the day off. Later,
782
00:50:59,393 --> 00:51:02,762
the FBI is listening when Alba tells a prosecutor he
783
00:51:02,778 --> 00:51:05,313
was too sick to take his dad to the appointment.
784
00:51:05,843 --> 00:51:09,116
TMI, but I just threw up. Up all over myself.
785
00:51:09,356 --> 00:51:12,887
We're going to blame McDonald's. I shouldn't have ate that
786
00:51:13,387 --> 00:51:16,995
hot and spicy. So you always talk about what does it sound like in context.
787
00:51:17,524 --> 00:51:20,830
It's even more nonchalant. It's just like whatever, right?
788
00:51:20,878 --> 00:51:24,520
Like this was— like this was their job. So how much were they all
789
00:51:25,020 --> 00:51:28,372
pocketing? Search warrants show the FBI was looking for evidence
790
00:51:28,872 --> 00:51:32,560
of high-dollar purchases in their homes. Mendez and Alba allude to how
791
00:51:32,640 --> 00:51:36,202
much they've raked in over the years, even consider tracking
792
00:51:36,234 --> 00:51:39,505
it themselves to prove their value to Clear, as the
793
00:51:40,005 --> 00:51:43,338
2 speculate whether Clear might fire Mendez. When I went to
794
00:51:43,354 --> 00:51:47,203
him, guess how much he was charging for a DWI? Mendez asks
795
00:51:47,284 --> 00:51:51,598
Alba, referring to Clear. Like $2,000 to $3,000? Alba guesses.
796
00:51:51,790 --> 00:51:55,527
$1,300, Mendez replies. That's crazy,
797
00:51:55,783 --> 00:51:59,857
Alba replies. The 2 go on about easily getting $8,500
798
00:51:59,873 --> 00:52:03,225
from their latest target, A woman they're convinced has money.
799
00:52:03,802 --> 00:52:07,314
As for Clear himself, he has made so much money over the years.
800
00:52:07,427 --> 00:52:10,682
Alba tells Mendez, I tell you what, a lot of attorneys
801
00:52:10,714 --> 00:52:14,900
struggled during COVID bro. Oh yeah, we had a blast,
802
00:52:15,044 --> 00:52:18,492
Mendez replies. How is it that we're going to start to make people feel
803
00:52:18,992 --> 00:52:22,678
like they can trust law enforcement, and specifically APD? But I
804
00:52:23,178 --> 00:52:26,351
think the steps that were done even before I even got here— Once
805
00:52:26,851 --> 00:52:30,172
the Once an investigator of corruption within APD, Bujanda is now working
806
00:52:30,364 --> 00:52:33,897
with the department as the city's executive director of public
807
00:52:34,397 --> 00:52:38,521
safety. You look at your hiring as a step in regaining
808
00:52:38,553 --> 00:52:41,684
public trust? Yes, because it's real easy
809
00:52:42,184 --> 00:52:45,377
to kind of just pick someone from within, and then there's always going to be
810
00:52:45,877 --> 00:52:49,921
that level of doubt. Bujanda knows regaining public trust isn't easy.
811
00:52:50,178 --> 00:52:51,864
I can tell you from my perspective,
812
00:52:53,229 --> 00:52:56,601
You can trust us, but let us earn it. Let us earn that trust.
813
00:52:57,002 --> 00:53:00,358
Today, 10 former law enforcement officers have pled guilty
814
00:53:00,390 --> 00:53:03,810
to federal crimes. 2 defense attorneys and Mendez,
815
00:53:03,987 --> 00:53:07,728
the former paralegal. Does this send a different message to law
816
00:53:07,840 --> 00:53:11,437
enforcement officers across the state? Yeah, that's a good question,
817
00:53:11,533 --> 00:53:14,552
and I think, I think the answer is I hope that it sends a good
818
00:53:14,600 --> 00:53:18,133
message, right? That they are signing up for the They
819
00:53:18,197 --> 00:53:21,411
thought they were signing up for, to do good in this world.
820
00:53:21,508 --> 00:53:24,416
Gabrielle Burkhardt, KRQE Investigates.
821
00:53:26,746 --> 00:53:30,186
So Alex, there's, there's not a complete answer to a question that
822
00:53:30,686 --> 00:53:34,525
you had in previous, uh, coverage of this, this story, um, but there
823
00:53:35,025 --> 00:53:38,767
is somewhat of an indication that, that judges were not completely
824
00:53:38,880 --> 00:53:41,709
oblivious to this, um, uh,
825
00:53:42,356 --> 00:53:47,738
potentially happening. They, they did actually occasionally demand from
826
00:53:48,238 --> 00:53:51,622
From the police. Hey, hey, can you give me your doctor's note for this
827
00:53:52,122 --> 00:53:55,716
excuse? Because I'm going to need this here. I don't really buy your story of
828
00:53:56,216 --> 00:53:59,554
why you didn't show up. I mean, it still raises a question
829
00:54:00,054 --> 00:54:02,396
of how they got away with this for so long, but at least you could
830
00:54:02,896 --> 00:54:07,084
see from those conversations that judges did occasionally
831
00:54:07,116 --> 00:54:08,931
get a little suspicious of the activity.
832
00:54:11,857 --> 00:54:15,024
Yep, yep. Well, but, you know, the nature of the
833
00:54:15,524 --> 00:54:18,999
answer almost, you know, tells you that
834
00:54:19,499 --> 00:54:23,524
it's, um, they basically didn't,
835
00:54:23,638 --> 00:54:26,992
you know, care, right? Like, it's like, yeah, here's something that
836
00:54:27,040 --> 00:54:30,674
you can sort of hang your doubt
837
00:54:31,174 --> 00:54:33,206
on or whatever. What are you gonna do about it?
838
00:54:35,442 --> 00:54:38,090
Yes, yes, yes. Well,
839
00:54:38,993 --> 00:54:42,396
you know, we're actually gonna today in the last
840
00:54:42,896 --> 00:54:46,412
several weeks. Go ahead. Go ahead. Sorry. No, I was gonna ask
841
00:54:46,912 --> 00:54:50,707
since this was in Albuquerque, were any of the lawyers involved called Saul?
842
00:54:54,287 --> 00:54:57,638
Not to my knowledge, sir. None of them were named Saul.
843
00:54:57,687 --> 00:55:00,748
Did that come through loud and clear? Alex,
844
00:55:00,958 --> 00:55:03,294
Alex, sir, I've I am I'm a very big.
845
00:55:03,720 --> 00:55:07,271
Breaking Bad fan. You are not going to sneak by a Saul
846
00:55:07,304 --> 00:55:10,498
Goodman reference with, uh, with me missing it, I don't
847
00:55:10,530 --> 00:55:13,616
think. Very good. All right,
848
00:55:14,020 --> 00:55:17,787
um, now Alex, uh, we were just talking about the,
849
00:55:17,916 --> 00:55:21,376
uh, the AI companions, um, being abandoned
850
00:55:21,392 --> 00:55:24,846
by Elon Musk and, uh, a comment on our, on our
851
00:55:24,927 --> 00:55:28,414
YouTube feed. Alex, yes, we have live chat on the YouTube
852
00:55:28,914 --> 00:55:32,332
feed that posts to the video feed of this, this, uh, space here.
853
00:55:33,011 --> 00:55:37,146
Uh, comment says, uh, and Grokopedia is completely abandoned
854
00:55:37,646 --> 00:55:41,182
now, which is actually a thought I had recently, that Grokopedia is— you
855
00:55:41,682 --> 00:55:44,877
don't hear anything about it anymore. Um, it seems
856
00:55:45,377 --> 00:55:48,685
like a vaporware product now. Alex, would you, would you agree
857
00:55:49,185 --> 00:55:52,398
with that sentiment that the Grokopedia is vaporware?
858
00:55:52,947 --> 00:55:56,514
Funnily enough, I was looking into it recently
859
00:55:57,014 --> 00:56:00,018
and came to the same conclusion. And frankly, I mean,
860
00:56:00,115 --> 00:56:03,406
what, what was the plan there, you know,
861
00:56:03,984 --> 00:56:07,356
to do, you know, with Brackeypedia? Like, they would have had
862
00:56:07,856 --> 00:56:11,402
to be updating it as they went. Mm-hmm.
863
00:56:11,450 --> 00:56:15,288
Like, and I guess some level, um,
864
00:56:16,251 --> 00:56:19,928
you know, much like the companions, if, if it's kind of
865
00:56:20,425 --> 00:56:24,320
like a catch-22 or something where If
866
00:56:24,820 --> 00:56:27,420
it's not showing, you know, consumer demand,
867
00:56:27,920 --> 00:56:31,680
um, there's no point sort of putting in the tokens.
868
00:56:32,180 --> 00:56:34,500
And if it's, you know, putting in the tokens, it's not going to show consumer
869
00:56:35,000 --> 00:56:37,920
demand. It seemed, um,
870
00:56:38,420 --> 00:56:41,180
more like a stunt. I guess in retrospect it looks more like a stunt than
871
00:56:41,680 --> 00:56:44,990
anything else. But I— the reason I'm, uh, mentioning this is because I was,
872
00:56:45,160 --> 00:56:47,420
um, building out something.
873
00:56:47,680 --> 00:56:51,040
Um, it's coming out nice. I, I don't know,
874
00:56:51,540 --> 00:56:55,472
maybe I'll publish something. Uh, but basically, have mentioned
875
00:56:55,972 --> 00:56:59,614
this, have discussed this before with you, Gator. Um, a thing that basically gets an
876
00:56:59,711 --> 00:57:02,873
LLM to be, to, to act like an HTTP
877
00:57:03,373 --> 00:57:06,405
server. So whatever you,
878
00:57:07,095 --> 00:57:10,258
um, put on your browser's URL bar,
879
00:57:11,109 --> 00:57:14,592
um, I had it, I had it take over my mDNS on my local network.
880
00:57:14,689 --> 00:57:18,028
So whatever .local you put, it will hallucinate
881
00:57:18,528 --> 00:57:21,945
a website. Yes. Yes. I remember this product from a while ago. Yes. Yes.
882
00:57:22,844 --> 00:57:25,349
Yeah, so I've been, I've been going back to that and trying to make it
883
00:57:25,849 --> 00:57:29,170
fast and whatever, but the cool thing is that you can put wikipedia.local and it
884
00:57:29,670 --> 00:57:33,104
will make you a Wikipedia with a search bar that you can then type
885
00:57:33,604 --> 00:57:36,075
whatever you want in the search bar and it will make you, you know,
886
00:57:36,235 --> 00:57:39,350
a website. Um, now did you have
887
00:57:39,462 --> 00:57:43,445
that, did you have that, uh, creating it and, and committing
888
00:57:43,557 --> 00:57:46,784
it, or was it just pure hallucination all the time?
889
00:57:48,454 --> 00:57:52,438
No, it saves cache. That's what you mean. Okay. Like, so if you
890
00:57:52,711 --> 00:57:56,228
revisit a website, a webpage, but the interesting challenge
891
00:57:56,728 --> 00:58:00,099
I tackled this time around, now that I've revisited the project, is to have
892
00:58:00,599 --> 00:58:04,179
it generate the page really fast, like in a few seconds rather than like
893
00:58:04,679 --> 00:58:08,339
half a minute or something, which is surprisingly
894
00:58:08,451 --> 00:58:11,985
complicated. But whatever, long story short, I have
895
00:58:12,485 --> 00:58:16,364
a lot of sort of optimism for the general concept
896
00:58:16,621 --> 00:58:20,644
of, you know, like a web of LLM.
897
00:58:21,686 --> 00:58:25,550
Like, you know, it's like navigating the web
898
00:58:25,774 --> 00:58:29,573
by navigating the mind of an LLM, which is kind of weird. Um, but I
899
00:58:29,621 --> 00:58:33,853
think there's, there's something there somehow. Uh, and Grokypedia was some
900
00:58:33,869 --> 00:58:37,236
kind of an attempt to do something like that. And, and I guess if you
901
00:58:37,736 --> 00:58:41,019
think about it a different way, um, if tokens were free, right, if you had
902
00:58:41,519 --> 00:58:45,331
models that were like really like decent, not necessarily like top of the
903
00:58:45,831 --> 00:58:49,845
line, but like not awful and very fast,
904
00:58:50,797 --> 00:58:54,443
uh, and also very cheap, this sort of thing would
905
00:58:54,943 --> 00:58:59,241
be something that you could do for, you know,
906
00:58:59,322 --> 00:59:02,249
you know, well, depending how cheap they were, I guess. But like,
907
00:59:02,459 --> 00:59:04,965
if you could get them to be as cheap as some of the things we
908
00:59:05,465 --> 00:59:09,501
were hearing about, um, you can get that to be, you know, extremely cheap.
909
00:59:10,476 --> 00:59:14,325
Uh, so I don't know, I think we might, we might see more of
910
00:59:14,825 --> 00:59:17,695
that concept in the future. That's as much as I'm willing to say. But it's—
911
00:59:18,594 --> 00:59:22,076
there's something very interesting about abandoning the, uh, the chat paradigm and,
912
00:59:22,494 --> 00:59:26,666
uh, discussing with an LLM and moving to a discovery,
913
00:59:26,763 --> 00:59:27,581
let's call it, paradigm.
914
00:59:28,672 --> 00:59:31,690
Hmm. Well,
915
00:59:32,380 --> 00:59:36,312
you're, you're still sort of mimicking the internet when you're doing it,
916
00:59:36,376 --> 00:59:41,335
right? So you're moving from one interface to a separate
917
00:59:41,835 --> 00:59:45,840
sort of interface and creating the interaction to fit
918
00:59:46,338 --> 00:59:49,761
a previously established experience, right? So,
919
00:59:50,774 --> 00:59:54,838
how would that differ— how would that differ in
920
00:59:54,951 --> 00:59:57,699
utility for you? Well,
921
00:59:58,008 --> 01:00:01,143
the idea is that you can ask for websites that don't exist, right? So,
922
01:00:03,352 --> 01:00:06,529
I was watching this very strange video about,
923
01:00:06,689 --> 01:00:10,448
uh, monoblock chairs, which is the, you know, the white molded,
924
01:00:11,027 --> 01:00:14,818
uh, chairs. Do you know what I mean? I'm talking about—
925
01:00:15,318 --> 01:00:18,818
do you get like the white garden chairs? Does that bring an image? Okay,
926
01:00:19,059 --> 01:00:21,806
yeah, sure. Um, apparently they're like the,
927
01:00:22,111 --> 01:00:25,452
the, the endpoint of a very long development procedure
928
01:00:25,565 --> 01:00:28,922
of like how to, you know, which includes like developing novel materials,
929
01:00:29,034 --> 01:00:32,809
developing novel techniques, yada yada, uh, you know, approaching the problem
930
01:00:32,874 --> 01:00:36,465
from all sorts of angles, and then ultimately it has
931
01:00:36,965 --> 01:00:40,056
become an object that, um, has no,
932
01:00:40,905 --> 01:00:44,015
like, actual designer per se. Like, at some point
933
01:00:44,515 --> 01:00:48,102
the idea escaped to, you know, the various factories in India and China
934
01:00:48,602 --> 01:00:52,671
and whatever, and people started making modifications to them until something emerged
935
01:00:53,171 --> 01:00:56,037
that lots of people wanted to manufacture. Sure.
936
01:00:56,117 --> 01:00:59,964
And, you know, it's kind of like, uh, you know,
937
01:01:00,300 --> 01:01:03,639
emergence emerges its way into the world. So it's like, literally,
938
01:01:03,752 --> 01:01:07,043
it's like a design object, which is in some ways the pinnacle of
939
01:01:07,543 --> 01:01:11,186
industrial design, that has no designer. Like, literally, they will put it on, you know,
940
01:01:11,844 --> 01:01:14,943
um, some, uh, uh,
941
01:01:15,537 --> 01:01:19,455
catalogs of like the most, you know, important objects constructed by humanity.
942
01:01:20,274 --> 01:01:23,838
Um, and it's like designer unknown. Anyway, long story short, I was— I just
943
01:01:23,903 --> 01:01:26,215
put— wrote like, you know, monoblock chair. Mm-hmm.
944
01:01:27,670 --> 01:01:31,030
.local, and it just showed up with a website about that chair,
945
01:01:31,320 --> 01:01:34,900
right? Okay. So instead of like talking to an LLM and saying like,
946
01:01:35,400 --> 01:01:38,220
tell me about this chair, and then it just like vomits a bunch of text
947
01:01:38,720 --> 01:01:41,020
at you, which isn't kind of what you want, you know, you know what I
948
01:01:41,520 --> 01:01:44,800
mean? Okay. You, you want to like experience it. You don't want to just like
949
01:01:45,300 --> 01:01:48,840
text, text in your face. Um, it gives you
950
01:01:49,340 --> 01:01:52,220
a different way of doing it. And, and in the same mode as you were
951
01:01:52,720 --> 01:01:54,700
saying like, well, you know, we've got the web already, why would we want to
952
01:01:55,200 --> 01:01:58,516
hallucinate one? We also have chat already, right? The,
953
01:01:58,532 --> 01:02:01,808
the, the point isn't the medium, but the— I wasn't,
954
01:02:01,824 --> 01:02:05,340
I wasn't necessarily saying so, you know, so forcefully, like, so why would you want
955
01:02:05,840 --> 01:02:08,857
to do it? I was just sort of saying, how does the, um, how does
956
01:02:08,953 --> 01:02:12,132
the, how does the change in, in this,
957
01:02:12,325 --> 01:02:15,520
the interface, let's call it, of accessing the
958
01:02:15,536 --> 01:02:19,534
information from the LLM? Because, uh, roughly speaking,
959
01:02:19,743 --> 01:02:23,465
you're attempting to do this specifically
960
01:02:23,722 --> 01:02:26,994
with the, um, the compressed
961
01:02:27,026 --> 01:02:30,361
information that's inside of the LLM, you're not necessarily
962
01:02:30,394 --> 01:02:33,986
hooking up to external resources because that would be somewhat redundant,
963
01:02:34,002 --> 01:02:37,097
right? Like, hey, crawl the internet,
964
01:02:37,113 --> 01:02:40,930
find another one of these web pages, and then just make me a fake one.
965
01:02:41,139 --> 01:02:44,186
Like, that would be weird, right? That would be a little strange.
966
01:02:44,731 --> 01:02:47,297
Well, not really. Anyway, I can, I can go on for very long and I'm,
967
01:02:47,329 --> 01:02:50,713
I'm kind of mindful that I don't wanna bore people, but I, I will say
968
01:02:52,100 --> 01:02:55,330
Think of a few. How far do I want to go with
969
01:02:55,830 --> 01:02:59,090
this? Okay, so I'll tell you one little bit I built there, which kind
970
01:02:59,590 --> 01:03:02,750
of changes the concept. But I will say, like, I have— so each page has
971
01:03:03,250 --> 01:03:06,670
like this little Intercom-style button at the bottom right, you know, the Intercom support
972
01:03:07,170 --> 01:03:10,270
chat that you can click? Yeah, yeah. I'm sure you do. So, but this one
973
01:03:10,770 --> 01:03:12,710
is an LLM that you can talk to and it can change the page for
974
01:03:13,210 --> 01:03:16,390
you. So if you say like, oh, put more images, like you could just like
975
01:03:16,890 --> 01:03:20,641
go to the support chat and like just ask for, you know, more images.
976
01:03:21,253 --> 01:03:25,355
And on the 10th iteration of more, more images,
977
01:03:25,716 --> 01:03:28,191
more stuff, you get what you want.
978
01:03:29,800 --> 01:03:32,987
No, but the idea being that you can sort of also mold
979
01:03:33,487 --> 01:03:36,595
the web as you want it. And presumably,
980
01:03:38,020 --> 01:03:41,258
if you're doing this on your phone, you can also install the app you just
981
01:03:41,758 --> 01:03:44,703
made as an app on your phone. Anyway, there's like I guess the whole thing
982
01:03:45,313 --> 01:03:49,084
starts from like, what can we hallucinate? Which is already kind of
983
01:03:49,584 --> 01:03:54,172
interesting and goes to like, discover the operating
984
01:03:54,204 --> 01:03:57,381
system you want for your life. But the, sorry, the thing I was
985
01:03:57,413 --> 01:04:00,591
going to say is because you can do that sort of,
986
01:04:01,008 --> 01:04:04,811
you know, just ask for a change, you could imagine this being an overlay on
987
01:04:05,311 --> 01:04:09,048
your normal web if you like, right? Mm-hmm. So when you go
988
01:04:09,548 --> 01:04:13,200
to the Gmail interface, That interface is essentially a
989
01:04:13,700 --> 01:04:16,980
negotiation between your priorities and Google's priorities,
990
01:04:17,480 --> 01:04:21,200
right? Mm-hmm. They'll put stuff there, like ads are a very easy example of this,
991
01:04:21,700 --> 01:04:25,000
right? They'll put stuff there that you definitely don't want there, but it's kind of
992
01:04:25,500 --> 01:04:28,600
like, well, if you want this website, we gotta put the ads there so that,
993
01:04:29,100 --> 01:04:31,640
you know, like you get something of what you want, we get something of what
994
01:04:32,140 --> 01:04:35,480
we want, and hopefully it works out. But like, you know, that can go in
995
01:04:35,980 --> 01:04:38,580
all sorts of directions. And usually the websites have like a ton of shit that
996
01:04:39,080 --> 01:04:41,825
you generally don't want. So if you could have a,
997
01:04:42,451 --> 01:04:45,372
like a layer in between that would just kind of give you the stuff you
998
01:04:45,420 --> 01:04:48,952
want only, there's something to that. I don't know. Or, or,
999
01:04:49,016 --> 01:04:52,467
or could scan like 3 websites and put them together into a, you know,
1000
01:04:52,547 --> 01:04:56,095
we, we, we used to talk about the mashups of the, are you that
1001
01:04:56,595 --> 01:05:00,396
old? You, you're probably that old to remember mashups and stuff, uh, about those
1002
01:05:00,493 --> 01:05:02,933
on the web. Old, old,
1003
01:05:04,265 --> 01:05:07,459
old. Okay. Okay, fine. Forget I said that. Uh, that young.
1004
01:05:08,800 --> 01:05:12,147
Anyway, so, you know, the whole idea of the programmable web and stuff, and it
1005
01:05:12,647 --> 01:05:15,170
all got stuck because we couldn't— not everybody could program.
1006
01:05:16,719 --> 01:05:19,087
And even if they could, I mean, let's face it, they probably have better things
1007
01:05:19,587 --> 01:05:21,844
to do than just sit and hack websites all day.
1008
01:05:24,217 --> 01:05:27,416
Whereas now programming is free, you know, in some
1009
01:05:28,390 --> 01:05:31,702
definitions of programming. Programming is free and
1010
01:05:32,202 --> 01:05:35,726
getting hacked is— Is even freer. Yes.
1011
01:05:36,114 --> 01:05:38,555
Yes. Anyway, as I said,
1012
01:05:39,913 --> 01:05:42,516
I can, I can talk about it for a long time. Yeah. I've, I have
1013
01:05:43,016 --> 01:05:45,605
an, I have an idea for you, Alex, and, and this is just a,
1014
01:05:45,767 --> 01:05:48,828
this is just a fun concept based off of what you're talking about. Okay.
1015
01:05:49,784 --> 01:05:54,006
When you're describing your little hallucinated website situation
1016
01:05:54,506 --> 01:05:57,532
here, I kind of get
1017
01:05:58,032 --> 01:06:02,427
an idea that this is a precursor to just having augmented
1018
01:06:02,443 --> 01:06:05,537
reality. Right.
1019
01:06:06,243 --> 01:06:09,663
Like the augmented web, so to speak. We kind of already have
1020
01:06:09,776 --> 01:06:12,987
the ability to chain together a pipeline. I just don't know if anyone's
1021
01:06:13,487 --> 01:06:17,932
done it yet, Alex, where if you, let's say, took the Apple ProVisors,
1022
01:06:18,703 --> 01:06:23,038
okay, the thing that nobody buys anymore, but technologically
1023
01:06:23,538 --> 01:06:26,780
speaking, they're actually pretty interesting if they were cheaper.
1024
01:06:28,064 --> 01:06:32,217
Yep. You could hypothetically program an
1025
01:06:32,265 --> 01:06:35,630
audio interface like a Whisper interface to then communicate
1026
01:06:36,130 --> 01:06:39,960
to a large language model, have that large language model connected
1027
01:06:40,057 --> 01:06:43,774
to something on the display that is then able to just
1028
01:06:44,274 --> 01:06:47,643
generate the correct imagery for what it is you'd like to
1029
01:06:48,143 --> 01:06:51,290
see or, and/or think about. Right, yes.
1030
01:06:52,822 --> 01:06:56,386
The most difficult and slowest part would be possibly
1031
01:06:56,451 --> 01:06:59,801
doing video. that would augment whatever you were
1032
01:07:00,106 --> 01:07:03,439
doing. But visuals would probably be fairly simple because
1033
01:07:03,471 --> 01:07:07,092
you could have— make sure that you have some Stable Diffusion-like
1034
01:07:07,621 --> 01:07:11,066
generative imagery that is based
1035
01:07:11,566 --> 01:07:14,912
off of, you know, the spherical representation. So if
1036
01:07:15,412 --> 01:07:19,431
you ever— if you've ever dealt with trying to use HDRI inside
1037
01:07:19,931 --> 01:07:23,180
of a 3D program, it's basically— you can look at it as a square image,
1038
01:07:23,244 --> 01:07:28,074
but it looks like it's It's warped, um, because it's a 360-degree
1039
01:07:28,267 --> 01:07:31,781
image that is now mapped to a rectangular, uh,
1040
01:07:32,087 --> 01:07:35,673
square. Um, so the generators can make those,
1041
01:07:36,271 --> 01:07:39,684
or you could at least train one to make one rather convincingly, and then have
1042
01:07:39,733 --> 01:07:43,272
that mapped into your Vision Pro so that now you have somewhat of an augmented
1043
01:07:43,289 --> 01:07:45,929
reality experience in real time.
1044
01:07:47,635 --> 01:07:51,193
Yeah. So, so you're saying like, for example, an experience like you're in your
1045
01:07:51,258 --> 01:07:55,202
house, right? And you go in a room. But you can project
1046
01:07:55,702 --> 01:07:59,174
over it, let's say, an escape room experience where
1047
01:07:59,674 --> 01:08:02,908
it— well, you know what escape rooms are, I assume.
1048
01:08:03,084 --> 01:08:08,695
Um, so you have a virtual reality sort
1049
01:08:09,195 --> 01:08:12,784
of experience that is also stitched to
1050
01:08:13,220 --> 01:08:16,789
actual reality so you don't bump into walls and whatnot. Right, right.
1051
01:08:17,176 --> 01:08:20,655
Yeah, yeah. I think that would be an interesting
1052
01:08:20,703 --> 01:08:24,027
experiment. Yeah. Well, I think a lot of this stuff
1053
01:08:24,332 --> 01:08:28,956
is gonna happen, uh, because the reason it doesn't, it's not happening right now is
1054
01:08:29,213 --> 01:08:32,087
ultimately compute, uh, speed.
1055
01:08:32,906 --> 01:08:36,535
And— I was actually just gonna say too many people are trapped, uh, on the
1056
01:08:37,035 --> 01:08:40,084
porn bots. Uh, too many people. That's why it hasn't happened yet. Too much.
1057
01:08:40,132 --> 01:08:44,242
Well, you know, in a few weeks that's gonna be over, so you expect
1058
01:08:44,723 --> 01:08:48,459
Even faster progress. All right, that was,
1059
01:08:48,491 --> 01:08:51,906
that was Elon's attempt at AI safety. It has not worked out,
1060
01:08:51,938 --> 01:08:55,465
so he's, he's pulling it back. All right,
1061
01:08:55,626 --> 01:08:59,073
okay. All right, moving forward, Alex,
1062
01:09:00,340 --> 01:09:03,659
um, we're gonna go to Meet the Flockers. This is basically going
1063
01:09:04,159 --> 01:09:07,218
to be a day of shitting on, on, um, the police.
1064
01:09:07,988 --> 01:09:11,868
Not necessarily by design, you know, occasionally the police do some
1065
01:09:11,980 --> 01:09:15,173
very good things. You get the nice cop that saves a
1066
01:09:15,673 --> 01:09:19,264
goat that's lost on the side of the road or something like that,
1067
01:09:19,280 --> 01:09:23,339
does something really nice and heroic and things like that. But lately,
1068
01:09:24,158 --> 01:09:27,286
when it comes to particularly this flock stuff, you're just going
1069
01:09:27,786 --> 01:09:31,843
to get a whole bunch of garbage. Alex, we have here as
1070
01:09:31,939 --> 01:09:34,924
the first story in our Meet the Flockers segment,
1071
01:09:36,047 --> 01:09:40,315
a story about the Brevard County Deputy
1072
01:09:40,524 --> 01:09:43,719
of the Year. The Brevard County
1073
01:09:43,751 --> 01:09:46,981
Sheriff recently named this man Deputy of the Year. Now,
1074
01:09:47,206 --> 01:09:51,384
I've learned he's off the job after being accused of egregious
1075
01:09:51,884 --> 01:09:55,352
misconduct. Just hours after this awards ceremony where Michael
1076
01:09:55,401 --> 01:09:59,321
Fultz was honored as Deputy of the Year, the sheriff's office says it received
1077
01:09:59,821 --> 01:10:03,274
this photo of Fultz, who appears to be performing a Nazi salute.
1078
01:10:03,660 --> 01:10:07,388
Although Fultz's face is blocked out, investigators say his mustache is shaved and waved
1079
01:10:07,888 --> 01:10:11,670
to mimic Adolf Hitler. The photo was reportedly taken shortly before Fultz
1080
01:10:11,702 --> 01:10:14,893
became a deputy. Fultz denied being racist, records show,
1081
01:10:15,086 --> 01:10:18,807
describing the photo as merely a joke. Another photo shows
1082
01:10:18,903 --> 01:10:23,281
Fultz holding a firearm to his ex-girlfriend's head and inserting
1083
01:10:23,393 --> 01:10:27,146
it into her mouth during sex. The ex-girlfriend told investigators it was
1084
01:10:27,178 --> 01:10:30,787
completely consensual. Records show Fultz admitted that he violated
1085
01:10:30,851 --> 01:10:34,307
all safety rules and firearm training. training. The Deputy of the
1086
01:10:34,807 --> 01:10:38,340
Year was also accused of logging into the Flock camera license plate
1087
01:10:38,840 --> 01:10:42,099
reader system to look up his ex-girlfriend's tag number. He denied
1088
01:10:42,148 --> 01:10:45,849
doing it to spy on her. Fultz, who just issued a public
1089
01:10:46,349 --> 01:10:49,565
apology, resigned before the Sheriff's Office could fire him.
1090
01:10:49,870 --> 01:10:51,881
Check out the full story on ClickOrlando.com.
1091
01:10:54,632 --> 01:10:58,605
Now, Alex, that raises the question— This escalated quickly.
1092
01:11:00,860 --> 01:11:02,606
This raises the question of who got passed up for the deputy.
1093
01:11:05,100 --> 01:11:07,839
The Year Award, and what are they up to?
1094
01:11:12,960 --> 01:11:17,027
What could they— I mean, maybe they were involved
1095
01:11:17,527 --> 01:11:20,905
in the WUI story. I don't know. Look, I'm not going to say— You take
1096
01:11:21,405 --> 01:11:25,024
out half your police department. I'm not going to say I've never done anything
1097
01:11:25,137 --> 01:11:28,981
wrong in my life, that I've never done anything that was ill-advised.
1098
01:11:29,094 --> 01:11:33,159
I've never— I'm not going to say that I've never had a whiff
1099
01:11:33,659 --> 01:11:36,756
of arrogance about me and done something as a cocky,
1100
01:11:36,853 --> 01:11:40,145
braggadocious asshole. So I'm gonna, I'm gonna,
1101
01:11:40,161 --> 01:11:43,421
you know, have to side with, we're all a series
1102
01:11:43,469 --> 01:11:46,617
of pluses and minuses on the good boy, bad boy charts.
1103
01:11:47,131 --> 01:11:50,455
And at the end of life, hopefully we've been more of a good
1104
01:11:50,519 --> 01:11:54,181
boy than we have been a bad boy. But I want to know— He's
1105
01:11:54,213 --> 01:11:57,377
making a list, he's checking it twice. I want to know what the good boy
1106
01:11:57,505 --> 01:12:00,771
criteria was that earned him The deputy of the year
1107
01:12:00,820 --> 01:12:04,503
to see how it compares to the bad boy side of his chart
1108
01:12:04,535 --> 01:12:08,191
here, because because
1109
01:12:08,691 --> 01:12:12,301
that's a that's some you know the funny thing is like I heard
1110
01:12:12,769 --> 01:12:16,340
the first story about the salute so to speak and I
1111
01:12:16,840 --> 01:12:19,914
was like okay that could have been a prank or like some sort of you
1112
01:12:20,414 --> 01:12:24,020
know weird Halloween costume whatever thing okay then
1113
01:12:24,707 --> 01:12:29,642
the second one you're like. Well, that is a, I mean,
1114
01:12:29,900 --> 01:12:32,902
I guess it's a private affair, I suppose.
1115
01:12:34,500 --> 01:12:37,289
And then the third one was like, oh, fuck it.
1116
01:12:37,616 --> 01:12:41,447
It just cannot be saved. Sorry. You deserve
1117
01:12:41,480 --> 01:12:44,824
what's coming to you. All right.
1118
01:12:44,936 --> 01:12:48,326
Now, Ron DeSantis actually was giving
1119
01:12:48,374 --> 01:12:52,246
somewhat of a seemingly off-the-cuff discussion
1120
01:12:52,616 --> 01:12:56,241
about flock cameras. He is apparently not
1121
01:12:56,290 --> 01:13:00,428
too enthusiastic about how they are currently being used.
1122
01:13:03,507 --> 01:13:07,005
So whether it's online with digital
1123
01:13:07,505 --> 01:13:09,535
surveillance or whether it's with cameras,
1124
01:13:10,650 --> 01:13:14,497
at what point are we just going to be free to not
1125
01:13:14,545 --> 01:13:18,006
have to deal with any of that? And so I think
1126
01:13:18,506 --> 01:13:21,992
it needs to be overhauled for sure. I know New Hampshire sure did something that
1127
01:13:22,362 --> 01:13:26,889
really kept it. It's interesting. They put something in a bill about
1128
01:13:27,034 --> 01:13:30,326
using them on state— so the
1129
01:13:30,711 --> 01:13:33,264
sheriffs and the police can use it on state right-of-ways.
1130
01:13:34,259 --> 01:13:38,193
And I asked Jared, I said, well, what's— is the state involved? State doesn't
1131
01:13:38,693 --> 01:13:41,405
get any of the information or anything. It's just like they're doing it, so we
1132
01:13:41,905 --> 01:13:44,777
don't even really know, you know, like the police department or whatever. And when I
1133
01:13:45,277 --> 01:13:49,443
see like a police officer using this to track like an ex-girlfriend.
1134
01:13:49,539 --> 01:13:52,471
I'm like, you know, what the hell's going on, you know, with this? So I
1135
01:13:52,971 --> 01:13:56,574
think this has happened very quickly. I think people are right to be concerned.
1136
01:13:57,118 --> 01:14:00,532
I wanted an AI Bill of Rights that could have addressed this and others,
1137
01:14:01,029 --> 01:14:05,051
other issues. The Senate passed a product,
1138
01:14:06,092 --> 01:14:09,682
and the House in Florida, they wouldn't even, they wouldn't even do, do anything about
1139
01:14:10,182 --> 01:14:13,884
it. Now think about it, all the stuff online about how
1140
01:14:14,591 --> 01:14:18,316
there's— I mean, I just, I don't want to be in a situation where
1141
01:14:18,765 --> 01:14:22,731
these tech companies, a handful of tech companies, know everything
1142
01:14:22,780 --> 01:14:25,798
about you and know where you're going. And I know they collect a lot as
1143
01:14:25,895 --> 01:14:28,753
it is, but I think it's going to get a lot worse unless we have
1144
01:14:28,801 --> 01:14:32,221
some protections for the people in
1145
01:14:32,237 --> 01:14:35,689
Florida. And so, it's going to need to be— but the only way it could
1146
01:14:36,189 --> 01:14:39,763
be dealt with is legislatively. This is not something that,
1147
01:14:40,196 --> 01:14:43,961
you know, it's something that's being implemented at
1148
01:14:44,461 --> 01:14:48,208
the, primarily at the local level, and you're gonna need the
1149
01:14:48,708 --> 01:14:52,102
legislature to provide protections for people. So what I would say is if it's something
1150
01:14:52,134 --> 01:14:55,659
you're concerned about, then I think it's, you should talk to the people
1151
01:14:56,159 --> 01:14:58,944
running for office, particularly for the legislature,
1152
01:14:59,329 --> 01:15:02,437
and see. And I know New Hampshire has done something.
1153
01:15:03,816 --> 01:15:07,091
And so I'm in the process of reviewing that,
1154
01:15:07,107 --> 01:15:10,767
because I think that would probably be a good place to start for us here.
1155
01:15:11,618 --> 01:15:14,203
But this has happened very quickly. I mean, this is in the last few years.
1156
01:15:14,267 --> 01:15:17,527
I mean, I think when I became governor, they had readers,
1157
01:15:17,623 --> 01:15:21,059
but it was more to toll people going through the service plazas,
1158
01:15:21,637 --> 01:15:24,864
or the toll plazas, which, you know, honestly,
1159
01:15:25,105 --> 01:15:28,139
we had a lot of toll roads when I became governor,
1160
01:15:28,155 --> 01:15:30,531
and I wish we didn't have that many toll roads. Thank you. toll roads,
1161
01:15:30,563 --> 01:15:33,629
because you do a lot. That's why we did the toll relief here in Florida.
1162
01:15:33,693 --> 01:15:37,368
We did a couple years in a row, 50% reduction. People were getting their
1163
01:15:37,384 --> 01:15:40,739
bills cut in half. That was a good, good thing, and I'm glad that we
1164
01:15:41,239 --> 01:15:43,869
were able to do that. But, but yeah, I think it's, um, I think people
1165
01:15:44,369 --> 01:15:47,833
are right to be concerned, and, and I think you got to have protections
1166
01:15:47,961 --> 01:15:51,621
for, for people. But, uh, go after criminals, great, let's all
1167
01:15:51,733 --> 01:15:55,409
do that. But surveilling just regular Floridians'
1168
01:15:55,505 --> 01:15:58,785
movement and how that data is used, you know,
1169
01:15:59,128 --> 01:16:00,958
that's a bird of a different feather.
1170
01:16:03,415 --> 01:16:06,879
Now, he mentioned talking about toll roads and using license plate
1171
01:16:07,379 --> 01:16:10,472
reader for toll roads as being what he remembers the old discussion being.
1172
01:16:10,909 --> 01:16:15,131
And I actually remember those days too, whenever the
1173
01:16:15,261 --> 01:16:18,917
license plate readers were first being put on those toll booths,
1174
01:16:19,159 --> 01:16:22,540
the people were complaining that they were saying
1175
01:16:22,572 --> 01:16:25,368
that those were gonna be used to try to spy on people to give them
1176
01:16:25,868 --> 01:16:29,154
speeding tickets. Got it. So if you went between
1177
01:16:29,250 --> 01:16:32,604
2 toll booths in a time
1178
01:16:32,637 --> 01:16:36,777
period that was logically faster than
1179
01:16:37,772 --> 01:16:41,768
the speed limit would've allowed you to, then you
1180
01:16:42,268 --> 01:16:45,010
would automatically get a ticket. And I was like, oh no, we'd never do that.
1181
01:16:45,058 --> 01:16:48,717
We'd never do that. And to my memory, I don't believe
1182
01:16:48,765 --> 01:16:52,135
that those were used for that purpose. But clearly, as soon as the
1183
01:16:52,635 --> 01:16:56,131
conversation got off of the toll booths and you got
1184
01:16:56,631 --> 01:17:00,524
you got readers everywhere else. Apparently now it's free going just to find people to
1185
01:17:01,088 --> 01:17:03,635
find crimes for or something. It's a free-for-all.
1186
01:17:07,634 --> 01:17:10,590
Now, moving forward,
1187
01:17:10,799 --> 01:17:14,511
the CEO of Flock we
1188
01:17:15,011 --> 01:17:18,675
covered last week, who had done an interview on the All
1189
01:17:18,823 --> 01:17:23,573
In podcast with the With the character
1190
01:17:24,073 --> 01:17:27,551
Jason, the host Jason. Calacanis, yes. Well,
1191
01:17:27,679 --> 01:17:31,496
it turns out, Alex, that that wasn't a one-off. He's apparently been doing
1192
01:17:31,929 --> 01:17:35,746
somewhat of a PR tour in various capacities,
1193
01:17:35,778 --> 01:17:39,178
going on different programs, trying to be a front face
1194
01:17:39,306 --> 01:17:43,396
for the company, but then also going through towns
1195
01:17:44,101 --> 01:17:47,809
and having private communications with, we'll call them
1196
01:17:47,889 --> 01:17:51,390
the clients of Flock Cameras in different jurisdictions.
1197
01:17:52,290 --> 01:17:56,450
Yeah, the subjects more like. It is reported
1198
01:17:56,787 --> 01:18:00,578
that there might be somewhat of a
1199
01:18:00,658 --> 01:18:04,400
tone difference and even in places
1200
01:18:04,432 --> 01:18:09,299
a content difference between the attempts at public positive
1201
01:18:09,331 --> 01:18:11,949
PR and what he says behind the scenes.
1202
01:18:13,569 --> 01:18:17,095
Akron's 146 flock surveillance cameras connect driver
1203
01:18:17,111 --> 01:18:20,863
information to a national network. This is from signalakron.org.
1204
01:18:21,520 --> 01:18:24,887
For drivers of the city of Akron, there's a good chance a police officer
1205
01:18:25,387 --> 01:18:28,558
can track where they went, uh, when they were there, and the route they took
1206
01:18:29,058 --> 01:18:32,598
before and after their trip. From main thoroughfares to
1207
01:18:32,806 --> 01:18:36,670
neighborhood streets, nearly 150 flock safety cameras are mounted
1208
01:18:36,734 --> 01:18:40,196
throughout the city. Logging the license plates of every vehicle that drives by
1209
01:18:40,822 --> 01:18:45,134
into a database searchable for the next 30 days by law enforcement agents nationwide.
1210
01:18:45,727 --> 01:18:49,591
Across the country, officers in more than 6,000 agencies operate
1211
01:18:49,639 --> 01:18:52,717
roughly 120,000 flock cameras that don't
1212
01:18:53,217 --> 01:18:56,628
need a warrant to discover where the driver has been. For example, a sheriff's
1213
01:18:57,128 --> 01:19:00,155
department in Texas used the technology to track a woman suspected of driving out of
1214
01:19:00,655 --> 01:19:03,265
state to have an abortion, searching cameras nationwide for her.
1215
01:19:04,068 --> 01:19:07,182
Akron City Council Member Fran Wilson, who represents Ward 1,
1216
01:19:07,294 --> 01:19:10,536
recently discovered Flock explicitly allows the technology to be used to
1217
01:19:10,584 --> 01:19:14,051
enforce anti-abortion laws in jurisdictions that don't proactively
1218
01:19:14,551 --> 01:19:18,192
set up a filter against it. The spokesperson for the Akron Police Department
1219
01:19:18,208 --> 01:19:21,610
told Signal Akron that the feature is turned off in Akron
1220
01:19:21,659 --> 01:19:25,077
for anti-abortion law enforcement, as it is for immigration enforcement.
1221
01:19:25,671 --> 01:19:28,929
A Flock spokesperson told Signal Akron Monday that the company,
1222
01:19:29,042 --> 01:19:32,171
quote, offers immigration and reproductive care filters,
1223
01:19:32,732 --> 01:19:36,114
within the system that can be activated by the agency. When the filters
1224
01:19:36,614 --> 01:19:39,879
are activated, if a user attempts to search for reasons associated with immigration
1225
01:19:40,379 --> 01:19:44,254
enforcement or reproductive care, the platform automatically prevents
1226
01:19:44,754 --> 01:19:48,084
the search from including networks that have that filter on. But for a long
1227
01:19:48,116 --> 01:19:51,530
period of time in Akron, officers infrequently logged any reason
1228
01:19:52,026 --> 01:19:55,808
for why they tracked a car. APD records from a 5-month period
1229
01:19:56,308 --> 01:20:00,856
in 2025 obtained by Signal Akron document a list of more than 37,000
1230
01:20:01,356 --> 01:20:04,941
Flock searches by officers that listed no discernible reason for the use of the camera
1231
01:20:05,441 --> 01:20:08,273
and database despite APD policy that requires it.
1232
01:20:09,090 --> 01:20:12,246
That makes it nearly impossible to know how Akron's system introduced in
1233
01:20:12,746 --> 01:20:16,396
2022 was used at that time. The city spokesperson said that
1234
01:20:16,896 --> 01:20:21,042
since October of 2025, in response to a complaint, the APD has required more thorough
1235
01:20:21,186 --> 01:20:24,534
record keeping of Flock searches nationwide. The company
1236
01:20:24,567 --> 01:20:28,059
is facing widespread accusations of facilitating stalking and other abuses by
1237
01:20:28,075 --> 01:20:31,681
law enforcement. Increasingly loud alarm bells from civil liberties advocates,
1238
01:20:32,002 --> 01:20:36,186
unprecedented vandalism of the expensive license plate tracking cameras, and growing
1239
01:20:36,686 --> 01:20:39,023
debate inside city governments about whether to continue the contracts.
1240
01:20:39,856 --> 01:20:42,966
Quote, don't trust what you read on Facebook about flock cameras.
1241
01:20:44,024 --> 01:20:47,342
Flock Safety CEO Garrett Langley and other executives convened
1242
01:20:47,842 --> 01:20:51,013
a conference call on Thursday, August 6th, for local Ohio
1243
01:20:51,513 --> 01:20:55,405
law enforcement and government officials. Quote, you already know what this partnership does for
1244
01:20:55,905 --> 01:20:58,964
your agency. The Wednesday morning email invitation states,
1245
01:20:59,253 --> 01:21:02,219
"Quote: This is the chance to put someone else's doubts to rest, and this is
1246
01:21:02,719 --> 01:21:06,276
not an opportunity that comes around again." In an email to Signal
1247
01:21:06,388 --> 01:21:10,333
Akron, a Flock spokesperson wrote that Thursday's discussion was a customer
1248
01:21:10,833 --> 01:21:14,679
conversation focused on communication and public safety. Wilson, a staunch critic
1249
01:21:15,179 --> 01:21:18,976
of Flock, joined the Thursday afternoon call and invited a Signal Akron reporter to sit
1250
01:21:19,476 --> 01:21:22,921
on it. The message from Flock executives to the local police
1251
01:21:23,421 --> 01:21:26,400
officials on the call, including chiefs from Medina, Strongville,
1252
01:21:26,529 --> 01:21:30,011
Euclid, South Euclid. And other area officials was clear.
1253
01:21:30,525 --> 01:21:34,442
Flock Safety executives, Langley explained, are working to counter the growing criticism
1254
01:21:34,474 --> 01:21:37,798
of the camera surveillance system. Their solution, discussed on the call,
1255
01:21:37,926 --> 01:21:41,218
is a public relations campaign to sway local and national media,
1256
01:21:41,603 --> 01:21:45,376
online critics, and the skeptical government officials whose opposition could put the
1257
01:21:45,876 --> 01:21:48,924
contracts and the technology lauded by police departments at risk.
1258
01:21:49,807 --> 01:21:53,275
As part of that, Langley said 80% of his time is now spent doing national
1259
01:21:53,775 --> 01:21:57,400
media interviews. Quote, it's the most impactful thing we're trying to do, guys.
1260
01:21:57,818 --> 01:22:01,092
I'm averaging 3 to 4 national interviews per day, he said.
1261
01:22:01,766 --> 01:22:05,009
But he encouraged the local flock clients to leave the national interviews to him.
1262
01:22:05,538 --> 01:22:09,230
Quote, let me handle the New York Times. You handle your local media,
1263
01:22:09,246 --> 01:22:12,456
Langley said. Quote, they want to hear from you. You know
1264
01:22:12,521 --> 01:22:14,896
your reporters. They'll pick up a sob story.
1265
01:22:15,972 --> 01:22:19,503
Langley also urged each agency to join a new portal used
1266
01:22:20,003 --> 01:22:23,323
by roughly 2,000 of its customers to pull flock success stories,
1267
01:22:23,741 --> 01:22:27,627
including those from across Ohio, to better counter local criticism.
1268
01:22:28,333 --> 01:22:32,315
We're seeing really positive results where now you have a place to point people saying,
1269
01:22:32,813 --> 01:22:36,586
these are the facts, Langley said. You know, don't trust what
1270
01:22:37,086 --> 01:22:40,600
you read on Facebook. Don't trust what you read on Instagram. These are the facts
1271
01:22:41,100 --> 01:22:44,887
for our town. Police officials on the call, including from the Medina Police
1272
01:22:45,387 --> 01:22:48,965
Department and Lorain County Sheriff's Office, were concerned about more frequent vandalism
1273
01:22:49,465 --> 01:22:52,798
of the cameras along with associated that might mean a loss of the tool for
1274
01:22:53,298 --> 01:22:56,660
their departments. Langley said Flock is starting to offer an insurance plan that
1275
01:22:57,160 --> 01:23:00,585
agencies can buy for every camera, later confirmed by a Flock spokesperson to
1276
01:23:01,085 --> 01:23:04,254
be $500 per year per camera, and is working to implement
1277
01:23:04,754 --> 01:23:08,355
vandalism detection in cameras, though he had concerns that a maintenance worker
1278
01:23:08,855 --> 01:23:12,970
could be labeled a vandal. A lot of this is such a contentious
1279
01:23:13,066 --> 01:23:16,430
issue that we're not going to be able to back our legislative bodies and
1280
01:23:16,446 --> 01:23:20,179
risk and ask for extra money to pay for these extra damaged cameras
1281
01:23:20,837 --> 01:23:24,368
or for any type of insurance plans, Medina Police Chief Edward Kinney told
1282
01:23:24,868 --> 01:23:28,477
Langley. We've definitely had some isolated incidents of flock cameras being damaged
1283
01:23:28,977 --> 01:23:32,586
and vandalized throughout the city, Akron Police spokesperson Lieutenant Michael Murphy told
1284
01:23:33,086 --> 01:23:36,663
Signal Akron. Wilson was one of the few other officials to ask Langley a
1285
01:23:37,163 --> 01:23:41,029
question. The council member said that they learned that a flock administrator in Akron
1286
01:23:41,529 --> 01:23:44,384
mentioned a new abortion filter and asked for more information about it.
1287
01:23:44,673 --> 01:23:48,059
Langley instead sharply criticized a widely read 404 Media
1288
01:23:48,559 --> 01:23:52,075
story last year about the Texas officer who tracked a woman accused of seeking an
1289
01:23:52,091 --> 01:23:56,043
abortion out of state, Langley said incorrectly that the reporting
1290
01:23:56,236 --> 01:23:59,352
didn't include the context that her family was worried for her safety.
1291
01:24:00,412 --> 01:24:03,930
There's a few things, right? Because one, it's not illegal to have an abortion in
1292
01:24:03,963 --> 01:24:08,043
Texas. It's illegal to administer an abortion, Langley argued. So the
1293
01:24:08,059 --> 01:24:11,320
woman could never have been under investigation for having an abortion.
1294
01:24:11,995 --> 01:24:15,256
The officer's reason for the search stated, had an abortion
1295
01:24:15,352 --> 01:24:18,547
search for female. Langley did not elaborate further
1296
01:24:18,563 --> 01:24:22,431
on Flock's current abortion-specific tracking technology, which Wilson asked
1297
01:24:22,447 --> 01:24:25,930
about, and the Ward 1 representative did not come away from the meeting feeling any
1298
01:24:25,962 --> 01:24:29,172
better than before it. Quote, generally, I walked away with a sense
1299
01:24:29,672 --> 01:24:32,912
that this technology had collected an absolute treasure trove of data and is being used
1300
01:24:32,944 --> 01:24:36,267
in ways that Flock isn't being forward enough about, Wilson said after the call.
1301
01:24:36,861 --> 01:24:40,311
Quote, I was really shocked that we have an immigration and abortion-related
1302
01:24:40,811 --> 01:24:44,693
offense filtering for a public safety technology. That fact, the fact that even exists,
1303
01:24:44,998 --> 01:24:47,805
tells us a lot about where we are in the country. It's disgusting that this
1304
01:24:47,853 --> 01:24:51,269
technology is being used for such offenses. We are in a dark time,
1305
01:24:51,798 --> 01:24:54,604
and that more— this is morally and ethically where my head is at, at the
1306
01:24:55,104 --> 01:24:59,319
end of the call. In the city's Flock surveillance logs obtained by Signal
1307
01:24:59,400 --> 01:25:03,168
Akron, APD officers and analysts searched for cars in Akron's
1308
01:25:03,668 --> 01:25:06,761
network and frequently hundreds or thousands of other networks nationwide.
1309
01:25:07,178 --> 01:25:10,465
More than 12,000 of those searches included at least 600 other networks,
1310
01:25:10,883 --> 01:25:14,560
Including 400 searches of more than 5,000 Flock networks nationwide.
1311
01:25:15,138 --> 01:25:18,542
In thousands of those instances, the searches listed indecipherable,
1312
01:25:18,654 --> 01:25:21,882
vague, or no justification at all in violation of the APD
1313
01:25:22,042 --> 01:25:25,976
rules for the technology. But the city now requires officers to select a reason from
1314
01:25:26,476 --> 01:25:29,396
a dropdown menu. They have the ability to log more justification.
1315
01:25:30,151 --> 01:25:32,591
In the 2025 records obtained by Signal Akron,
1316
01:25:33,025 --> 01:25:36,448
roughly 10,000 of those entries do not list a license plate. Instead,
1317
01:25:36,497 --> 01:25:39,816
the officers searched for a broad descriptor of the car, such as color
1318
01:25:40,316 --> 01:25:43,816
and make and model. Nearly 5,000 searches just list investigation,
1319
01:25:44,188 --> 01:25:47,244
including misspellings and abbreviations, as the reason for the search.
1320
01:25:47,667 --> 01:25:49,114
Hundreds just list crime.
1321
01:25:51,595 --> 01:25:56,130
And thousands more contain no information or seemingly random
1322
01:25:56,194 --> 01:25:59,051
letters. Bad stuff. The bad stuff.
1323
01:25:59,860 --> 01:26:03,194
During that period, a complaint showed that an APD analyst and officer tracked a
1324
01:26:03,694 --> 01:26:07,646
car Belonging to Autumn Riddle, a prominent critic of Akron police, and the November 2024
1325
01:26:08,146 --> 01:26:11,564
killing of Jasmeer Tucker. One entry listed activist
1326
01:26:12,064 --> 01:26:15,979
recording as the reason for the search, and another listed only law enforcement.
1327
01:26:16,621 --> 01:26:20,057
Internal APD investigation of the tracking said the searches
1328
01:26:20,169 --> 01:26:23,557
weren't retaliatory or harassment because of Riddle's activism,
1329
01:26:23,942 --> 01:26:27,121
but that the scant information included in them violated the department's policy.
1330
01:26:27,619 --> 01:26:31,069
Akron's independent police auditor Anthony Finnell. Backed up that finding and
1331
01:26:31,569 --> 01:26:34,950
declared that the searches failed to identify a crime, public safety concern,
1332
01:26:35,078 --> 01:26:38,526
or investigative nexus, and therefore undermined the department's
1333
01:26:38,558 --> 01:26:42,375
ability to demonstrate reasonable necessity and proportionality after
1334
01:26:42,875 --> 01:26:46,208
the fact. The records obtained by Signal Akron show that officers used
1335
01:26:46,224 --> 01:26:50,233
law enforcement tag as the sole reason for a search in 138
1336
01:26:50,733 --> 01:26:54,227
instances in that 5-month period. City of Akron spokesperson
1337
01:26:54,660 --> 01:26:58,364
Stephanie Marsh said that the 5 months of records from 2025 analyzed by
1338
01:26:58,864 --> 01:27:02,309
Signal Akron are not indicative indicative of how the department tracks the searches now
1339
01:27:02,598 --> 01:27:05,532
because more of a more thorough documentation currently required.
1340
01:27:06,270 --> 01:27:09,717
After this citizen complaint came in and we realized there was an issue with
1341
01:27:10,217 --> 01:27:13,806
how officers were documenting their searches, we changed it so that now officers have
1342
01:27:14,306 --> 01:27:17,222
to choose from a dropdown list of options for searches, Marsh wrote in an email.
1343
01:27:18,072 --> 01:27:21,487
They still do have the ability to add additional information related to a search,
1344
01:27:21,519 --> 01:27:24,886
and if it's already related to a case, they're required to include the case
1345
01:27:25,386 --> 01:27:28,430
number as well. Marsh said the cameras are essential to the city's policing efforts.
1346
01:27:28,494 --> 01:27:32,566
Quote, Flock cameras have proven to be an important tool for law enforcement,
1347
01:27:32,967 --> 01:27:36,349
helping support criminal investigations, giving officers another resource
1348
01:27:36,849 --> 01:27:40,646
to identify and apprehend those responsible for serious crimes, as well as
1349
01:27:40,982 --> 01:27:44,333
helping to recover missing persons. At the same time, we understand that there
1350
01:27:44,833 --> 01:27:47,715
are concerns for our community about the use of technology, and we look forward to
1351
01:27:48,215 --> 01:27:51,290
having thoughtful conversations with residents about those concerns in coming months.
1352
01:27:51,482 --> 01:27:52,989
Alex, your thoughts so far?
1353
01:27:55,524 --> 01:27:59,165
I don't know, man. Um, it's, it's just funny
1354
01:27:59,197 --> 01:28:01,860
that the solution is to, um, have the,
1355
01:28:03,112 --> 01:28:06,544
um, police like put, you know, have a massive
1356
01:28:06,560 --> 01:28:10,587
dropdown where, you know, like how medical coding works,
1357
01:28:11,132 --> 01:28:14,581
where, you know, the doctors have like a bunch of like, you know,
1358
01:28:14,597 --> 01:28:18,399
elaborate codes and it's all shit. And like, if you ever interact with that data,
1359
01:28:18,608 --> 01:28:22,499
it's kind of known that like people get so bored
1360
01:28:22,548 --> 01:28:26,295
that they, you know, will put whatever there.
1361
01:28:26,618 --> 01:28:30,285
So, my first thought is like I wonder what's top of that dropdown
1362
01:28:30,317 --> 01:28:33,908
because that's going to get used a lot. So, a long time
1363
01:28:33,973 --> 01:28:37,079
ago in my early days of software engineering,
1364
01:28:38,780 --> 01:28:41,120
one of the first things that I sort of, you know,
1365
01:28:42,040 --> 01:28:47,217
independently honed in as a philosophy of software
1366
01:28:47,717 --> 01:28:51,175
engineering is that anytime you have to gather data
1367
01:28:51,784 --> 01:28:55,164
from an end user, you need to
1368
01:28:55,664 --> 01:28:58,977
make gathering that data return something of value to
1369
01:28:59,009 --> 01:29:01,461
that user, or else you will get garbage.
1370
01:29:02,694 --> 01:29:05,915
Right. If the, if the data collection that you are doing serves
1371
01:29:06,415 --> 01:29:10,193
no benefit to the end user, they are not going to be helpful
1372
01:29:10,705 --> 01:29:13,798
in providing that information to you. So whatever you are doing
1373
01:29:13,830 --> 01:29:17,172
to gather that information, needs to be a byproduct of
1374
01:29:17,204 --> 01:29:20,747
their actual day-to-day work in a way that helps them. So,
1375
01:29:20,763 --> 01:29:22,808
it encourages them to be accurate.
1376
01:29:24,060 --> 01:29:27,273
So, for example, if the product was to have a specialized
1377
01:29:27,773 --> 01:29:30,584
user interface for each use case, right,
1378
01:29:32,126 --> 01:29:35,350
such that filling the dropdown would give
1379
01:29:35,850 --> 01:29:39,649
you a better experience for
1380
01:29:39,762 --> 01:29:44,665
what you were trying to do, then you might see that Yes.
1381
01:29:44,922 --> 01:29:48,209
And, and, and that just sort of adds to what
1382
01:29:48,709 --> 01:29:52,442
you were talking about of having an endless series of, of codes or, or things
1383
01:29:52,474 --> 01:29:56,130
in a dropdown that you're trying to filter through winds
1384
01:29:56,226 --> 01:29:59,192
up enticing people to be less accurate.
1385
01:29:59,593 --> 01:30:03,185
And you can counterbalance that tendency with
1386
01:30:03,201 --> 01:30:06,632
penalties for misfiled,
1387
01:30:07,450 --> 01:30:11,409
you know, things, right? So if, if But
1388
01:30:11,698 --> 01:30:15,126
I mean, look, right now everything is
1389
01:30:15,158 --> 01:30:18,731
backwards. It is incumbent— right now the situation,
1390
01:30:18,907 --> 01:30:21,807
particularly as this Akron demonstrates,
1391
01:30:22,528 --> 01:30:26,197
it is incumbent upon those who are the least
1392
01:30:26,213 --> 01:30:29,978
capable of raising an issue to find and
1393
01:30:30,478 --> 01:30:32,670
conclusively demonstrate that a problem exists,
1394
01:30:33,887 --> 01:30:37,877
prove that that problem exists to those who are least likely
1395
01:30:37,909 --> 01:30:41,643
to want to accept that the problem exists. then get them
1396
01:30:41,996 --> 01:30:45,908
to fix that problem. That is a backwards sort
1397
01:30:46,408 --> 01:30:50,365
of process in this particular way where the person was complaining,
1398
01:30:50,798 --> 01:30:54,245
oh, well, that's outdated. We've changed the policies now so that
1399
01:30:54,261 --> 01:30:57,372
issue that was raised that ran rampant for
1400
01:30:57,872 --> 01:31:01,252
many, many months is now ameliorated. Well, what about the
1401
01:31:01,752 --> 01:31:05,212
next problem? What about the next thing? The organization,
1402
01:31:05,260 --> 01:31:09,029
the organism has already demonstrated that it is not proactive enough
1403
01:31:09,480 --> 01:31:13,412
to understand how to, or even want to cut off these
1404
01:31:13,444 --> 01:31:16,989
issues ahead of time. It is only when they are
1405
01:31:17,489 --> 01:31:20,863
embarrassed and brought, you know,
1406
01:31:21,764 --> 01:31:25,096
when the light is used as the disinfectant, as they
1407
01:31:25,596 --> 01:31:26,786
say, is it addressed.
1408
01:31:30,760 --> 01:31:33,537
Yeah. No, it's, I don't know, like,
1409
01:31:34,609 --> 01:31:37,695
look, here's something that's on my mind a lot, right? And it might be a
1410
01:31:38,195 --> 01:31:41,180
little bit sideways, but let's Uh, see if we can get something useful out of
1411
01:31:41,212 --> 01:31:44,348
this. So from what I hear, a big problem
1412
01:31:44,848 --> 01:31:47,502
with these cameras is that they are used,
1413
01:31:48,357 --> 01:31:50,923
um, for, um,
1414
01:31:51,116 --> 01:31:52,455
let's say it,
1415
01:31:54,459 --> 01:31:57,754
um, like to collate data across
1416
01:31:58,204 --> 01:32:01,804
all sorts of agencies, right? So, you know, like somebody can
1417
01:32:02,012 --> 01:32:05,707
scan for a particular car across the entirety
1418
01:32:05,723 --> 01:32:09,429
of the United States or so, or, you know, stuff like that. And I'm
1419
01:32:09,929 --> 01:32:13,071
willing to bet that, you know, most of the cases that
1420
01:32:13,571 --> 01:32:16,864
are solved are solved very locally, so to speak.
1421
01:32:17,249 --> 01:32:20,865
Like, you know, it's like, okay, you're within a few blocks, like within
1422
01:32:21,365 --> 01:32:25,409
the precinct sort of area of responsibility. So I,
1423
01:32:25,961 --> 01:32:31,210
I haven't— what I haven't heard much of is sort of you
1424
01:32:31,710 --> 01:32:34,290
know, we're, we're going to use it, but we're going to have, for example,
1425
01:32:34,790 --> 01:32:37,150
local storage of the data and it's not going to get collated nationally. I think
1426
01:32:37,650 --> 01:32:40,590
that would be a big difference. Now, do you trust them?
1427
01:32:41,090 --> 01:32:44,310
Whatever. Sure. That's a different story. But like, we're assuming right
1428
01:32:44,810 --> 01:32:48,250
now, and again, that could be analyzed, that they, they do
1429
01:32:48,750 --> 01:32:51,930
what they say they do. Mm-hmm. Um, yeah.
1430
01:32:52,430 --> 01:32:56,270
So, so, you know, I'm not hearing anything about that aspect and
1431
01:32:56,770 --> 01:33:00,037
I find that concerning because it's almost like, That is kind of what
1432
01:33:00,537 --> 01:33:04,475
they'd want. Like, if they, you know, like talking about
1433
01:33:04,975 --> 01:33:08,128
like small local concerns,
1434
01:33:08,662 --> 01:33:12,307
but not remote concerns. And another element is,
1435
01:33:12,631 --> 01:33:16,225
you know, they say, well, they have limited retention,
1436
01:33:16,274 --> 01:33:20,977
right? We were talking the other week about whether it's
1437
01:33:21,477 --> 01:33:24,717
7 days or 30 days or 1 year or whatever. Okay. But what happens to
1438
01:33:25,217 --> 01:33:29,073
like query results, for example? that are extracted, are those subject to that same
1439
01:33:29,573 --> 01:33:33,113
retention? If not, what prevents something that has
1440
01:33:33,225 --> 01:33:36,383
national access to this thing from, well,
1441
01:33:36,832 --> 01:33:40,279
making some extremely broad queries that it then saves in
1442
01:33:40,779 --> 01:33:44,351
a different system? Yeah. So you're effectively—
1443
01:33:44,671 --> 01:33:47,829
just to clarify, because, you know, I don't think I'm confused about
1444
01:33:48,329 --> 01:33:51,308
what you just said, but just to clarify, you know— No, I appreciate it.
1445
01:33:51,404 --> 01:33:54,693
Please go ahead. The data retention period of, let's say,
1446
01:33:54,709 --> 01:33:58,144
7 days is only about things that haven't
1447
01:33:58,644 --> 01:34:02,316
been looked up because hypothetically someone can look up
1448
01:34:02,637 --> 01:34:06,425
a whole lot of stuff and archive their results and
1449
01:34:06,925 --> 01:34:10,004
have access to that information perpetually, in perpetuity,
1450
01:34:10,581 --> 01:34:14,738
for just the regular storage cost. Of course, the complexity being
1451
01:34:15,733 --> 01:34:19,473
getting it out of that system. So I would imagine That's easy
1452
01:34:19,585 --> 01:34:23,054
because most things these days suck and they
1453
01:34:23,102 --> 01:34:26,570
all run on Electron, which means that it's just freaking WebKit
1454
01:34:26,650 --> 01:34:30,070
under the hood and, you know, like a Chromium
1455
01:34:30,424 --> 01:34:34,117
host. As the person who probably made one of the most
1456
01:34:34,617 --> 01:34:37,649
used Electron apps at one time, I resent you bringing that up all the time.
1457
01:34:37,874 --> 01:34:41,904
But anyway, keep going. Well, let me
1458
01:34:42,404 --> 01:34:45,950
be clear in the difference there, Alex. Let's imagine
1459
01:34:46,625 --> 01:34:49,932
that this was made with, with the old Win32 API.
1460
01:34:50,109 --> 01:34:53,995
So it's actually a legitimate Windows 32 application.
1461
01:34:54,749 --> 01:34:58,651
Do you understand the complexity of scraping
1462
01:34:59,149 --> 01:35:02,537
the data out of a list box that would be returned right
1463
01:35:03,037 --> 01:35:06,679
there? Right. You would have to do a lot of Windows 32 API calls in
1464
01:35:06,695 --> 01:35:10,383
order to get that information out. In the days of LLMs, That's just
1465
01:35:10,883 --> 01:35:14,415
a matter of how many tokens you have. Point taken.
1466
01:35:15,351 --> 01:35:17,866
Even at that level, you're given some schlub,
1467
01:35:19,331 --> 01:35:22,229
that deputy of the year is not going to figure out how to use Claude
1468
01:35:22,729 --> 01:35:26,077
Card in order to get the Win32 API. What I'm concerned about,
1469
01:35:26,093 --> 01:35:30,238
again, has to do with the national syndication
1470
01:35:30,496 --> 01:35:33,965
element, because who tells me that there isn't some police department,
1471
01:35:35,089 --> 01:35:39,063
technically police department, but really part of the NSA, that technically
1472
01:35:39,563 --> 01:35:43,222
has police officers doing things, but actually is treating it as an API and not
1473
01:35:43,722 --> 01:35:46,899
siphoning off all the data. Like, you know, that's the,
1474
01:35:46,915 --> 01:35:50,062
the kind of question I'm worried about. Yeah, the, the, the deputy of the year,
1475
01:35:50,271 --> 01:35:53,354
you know, well, that is what it is. Um, now I
1476
01:35:53,854 --> 01:35:57,191
have a question. I have a question, Alex. This is just popping off the top
1477
01:35:57,691 --> 01:35:59,503
of my head here, Alex, and I don't think that you have an answer,
1478
01:35:59,519 --> 01:36:04,073
but I think you'll appreciate the question. Are You're
1479
01:36:04,573 --> 01:36:07,849
not limited to being a customer of Flock unless you're
1480
01:36:07,865 --> 01:36:11,785
law enforcement, right? Because I think they've talked about selling this to HOAs
1481
01:36:13,345 --> 01:36:16,915
and— There you go. Right? Yep.
1482
01:36:17,462 --> 01:36:21,701
Yep. Go ahead. Go ahead. Are you prohibited
1483
01:36:21,718 --> 01:36:25,675
from being part of the network, the information sharing network,
1484
01:36:25,771 --> 01:36:29,005
unless you're a— Who knows? Who knows,
1485
01:36:29,055 --> 01:36:31,885
right? That's the whole fucking thing. that this is, again,
1486
01:36:31,966 --> 01:36:35,416
like, you could get 90% of the value of the thing by having
1487
01:36:35,640 --> 01:36:39,090
a local to the police department database that is sitting on the local to the
1488
01:36:39,590 --> 01:36:43,375
police department machine and not networking the whole thing with whoever.
1489
01:36:43,920 --> 01:36:47,515
But I can tell you from the startup playbook,
1490
01:36:48,253 --> 01:36:51,205
the value of the thing is in network effects.
1491
01:36:51,606 --> 01:36:54,800
And what does that mean in this particular situation? Cross-referencing,
1492
01:36:55,377 --> 01:36:58,747
right? Right. So exactly that. It's like being able to sell You know,
1493
01:36:58,763 --> 01:37:02,258
stuff to HOAs, to this, to that, to whatever. Because I would, I would imagine
1494
01:37:02,274 --> 01:37:05,209
that the cops would want to leverage the information from the HOAs.
1495
01:37:05,963 --> 01:37:08,512
Now, would Flock allow that in reverse though,
1496
01:37:09,330 --> 01:37:12,328
where the HOAs hypothetically have access to the,
1497
01:37:13,659 --> 01:37:17,283
uh, the system? Well, you know,
1498
01:37:17,364 --> 01:37:20,907
one hand washes the other. Do you want to, do you want to start
1499
01:37:21,407 --> 01:37:24,692
a virtual HOA inside of Second Life or some other virtual game
1500
01:37:24,740 --> 01:37:27,442
and, and try to get some flock cameras,
1501
01:37:27,700 --> 01:37:31,680
Alex. That's right. Let's go into Zuckerberg's metaverse and try
1502
01:37:31,744 --> 01:37:35,098
to see what— Yeah. See if we can get some flock cameras for that and
1503
01:37:36,018 --> 01:37:39,378
see if we can tap into the Akron Police
1504
01:37:39,491 --> 01:37:42,495
Department's share there, find out what we can do.
1505
01:37:46,089 --> 01:37:47,689
Yeah. With all that said,
1506
01:37:49,322 --> 01:37:52,394
I will say that I think a lot of this is sort of beside
1507
01:37:52,894 --> 01:37:55,537
the point because this company does seem to be particularly dislikable and/or desperate,
1508
01:37:57,120 --> 01:38:00,620
Trying to figure out how to you know squirm its
1509
01:38:01,120 --> 01:38:04,607
way out of the particular situation it's in, but ultimately there are at least
1510
01:38:05,107 --> 01:38:08,399
two other companies, and frankly it is not a difficult technology to
1511
01:38:08,497 --> 01:38:11,418
build to have another temp company showing up.
1512
01:38:13,119 --> 01:38:17,520
And you know what? The next one is going to be smarter about how it
1513
01:38:18,133 --> 01:38:22,048
camouflages its cameras, or it's not it's not even going to be visible, and I
1514
01:38:22,548 --> 01:38:25,898
think. Long story short, I think this whole topic is
1515
01:38:25,914 --> 01:38:28,770
just gonna die. I don't, I don't see how. I mean, I'm,
1516
01:38:29,476 --> 01:38:32,877
I'm sort of happy, you know,
1517
01:38:32,909 --> 01:38:37,097
mostly about the fact that people are aware and are pushing back, and maybe
1518
01:38:37,597 --> 01:38:40,931
some norms get enforced or, or encoded
1519
01:38:41,431 --> 01:38:44,782
in, in laws and whatever. So maybe we get something out of it. Uh,
1520
01:38:45,119 --> 01:38:47,140
that would be the best outcome here, I think. I don't know, Gator, what do
1521
01:38:47,640 --> 01:38:49,914
you think? Yeah, I mean, I think that's the best outcome. would be some sort
1522
01:38:50,414 --> 01:38:54,258
of legal protection against some aspects of this, like the
1523
01:38:54,562 --> 01:39:00,012
collation nationally or the detention— not
1524
01:39:00,512 --> 01:39:04,035
detention, retention period. Well, but the best outcome I think would
1525
01:39:04,099 --> 01:39:07,833
be along the lines of enough occurrences
1526
01:39:08,090 --> 01:39:11,487
of these things that happen throughout the United States that something seeps up
1527
01:39:11,936 --> 01:39:15,719
and finds its way to the Supreme Court where the Fourth Amendment
1528
01:39:16,219 --> 01:39:19,794
protections are reevaluated in this context.
1529
01:39:20,613 --> 01:39:24,419
Because I said several moons ago
1530
01:39:25,173 --> 01:39:28,931
that roughly speaking, the logic behind
1531
01:39:28,963 --> 01:39:32,239
being allowed to have these, let's say, spy networks
1532
01:39:32,720 --> 01:39:36,960
is that if a cop could have seen it in public and
1533
01:39:37,056 --> 01:39:40,894
written it down, it is somewhat equivalent in
1534
01:39:40,926 --> 01:39:44,491
the eyes of the law right now to just being a
1535
01:39:44,539 --> 01:39:48,055
set of eyes that a cop would have. So you are not granted— Yeah.
1536
01:39:48,702 --> 01:39:52,242
The same rights to privacy. And this extends to the
1537
01:39:52,259 --> 01:39:57,366
way that cars and the Fourth Amendment interplay.
1538
01:39:57,447 --> 01:40:00,306
So you go up to a car.
1539
01:40:00,919 --> 01:40:05,612
If I can see something illegal in the car,
1540
01:40:05,741 --> 01:40:10,096
that is the probable cause an officer would have to enter
1541
01:40:10,596 --> 01:40:14,695
the car and search it without any warrants needed. Whereas otherwise, if nothing were
1542
01:40:15,195 --> 01:40:18,118
visible in the car, they have to get the warrant in order to search.
1543
01:40:19,361 --> 01:40:22,670
And there's these barriers that have been built over
1544
01:40:23,170 --> 01:40:26,741
time to try to balance a situation where someone,
1545
01:40:26,951 --> 01:40:30,719
an individual, has private effects but are also
1546
01:40:30,848 --> 01:40:34,600
simultaneously out in public. It's an
1547
01:40:35,100 --> 01:40:38,872
attempt to balance. Help me out here. So what
1548
01:40:39,372 --> 01:40:43,603
if I were to tell you that we were to build another
1549
01:40:43,635 --> 01:40:47,168
app, let's call it Flockacop or maybe Flop? as a
1550
01:40:48,357 --> 01:40:51,971
shortcut, where people in public are
1551
01:40:52,019 --> 01:40:55,328
encouraged to film police whenever they see
1552
01:40:55,828 --> 01:40:59,793
them. And we have a little bit of facial recognition, a little bit of whatever,
1553
01:41:00,130 --> 01:41:03,519
you know, numbers or tags they might have, a little
1554
01:41:04,019 --> 01:41:07,277
bit of inference, you know how it goes. And then we were able to just
1555
01:41:07,777 --> 01:41:11,229
like directly map all the police in a certain area.
1556
01:41:13,220 --> 01:41:16,768
What do you think the police would do in response? I don't
1557
01:41:17,427 --> 01:41:20,638
even think we need to wonder about that.
1558
01:41:21,425 --> 01:41:25,326
They typically attempt to find a way of identifying
1559
01:41:25,439 --> 01:41:29,421
the individuals who are responsible for that idea and attempt
1560
01:41:29,453 --> 01:41:33,226
to arrest them for some crime, whether real
1561
01:41:33,726 --> 01:41:37,530
or imaginary, that involves placing the officers' lives at
1562
01:41:37,722 --> 01:41:40,932
at risk and/or safety at risk. Right, right, right, right.
1563
01:41:41,156 --> 01:41:43,949
Exactly. So there you go. That's it. That's the answer to your question.
1564
01:41:44,944 --> 01:41:48,410
Like, it's all nice and dandy until you flip the playbook
1565
01:41:48,910 --> 01:41:52,711
and then all of a sudden, you know, well, it's a different story if you're
1566
01:41:53,211 --> 01:41:56,466
placing an officer's life at risk. If doing this places an officer's
1567
01:41:56,482 --> 01:42:00,494
life at risk, presumably doing the opposite is placing people's
1568
01:42:00,575 --> 01:42:03,881
life at risk. Ergo, get fucked.
1569
01:42:03,945 --> 01:42:07,230
I don't know. Well, now the last thing I'll say before
1570
01:42:07,262 --> 01:42:11,281
we move on, Alex, because we have a parallel topic to jump
1571
01:42:11,781 --> 01:42:15,013
through, or somewhat related topic to jump through in some ways.
1572
01:42:17,807 --> 01:42:22,697
I will say that I won't come down extremely
1573
01:42:22,794 --> 01:42:26,038
hard on the Flock CEO, although I at
1574
01:42:26,538 --> 01:42:30,505
the moment generally don't like him. He rubbed me the wrong way in that Jason
1575
01:42:30,587 --> 01:42:33,835
interview, and I'm not I don't think I want to be his
1576
01:42:34,016 --> 01:42:39,117
friend. But there
1577
01:42:39,617 --> 01:42:43,856
was a— in that article, it was talking about how he was basically providing a
1578
01:42:43,937 --> 01:42:47,372
propaganda resource for the local communities to try to
1579
01:42:47,388 --> 01:42:52,745
use, saying like, look, we have a PR battle. We want
1580
01:42:53,245 --> 01:42:56,567
you to be customers still. You guys want to be customers because it's convenient to
1581
01:42:57,067 --> 01:43:00,754
have our product. You are finding value in it. So we want to
1582
01:43:00,915 --> 01:43:04,256
recruit you to be free PR
1583
01:43:04,385 --> 01:43:08,496
for us through the local outlets by giving you this resource
1584
01:43:08,996 --> 01:43:12,248
that's going to procure for you the positive stories that
1585
01:43:12,264 --> 01:43:15,454
will help sell the product in your community. Now,
1586
01:43:16,967 --> 01:43:20,619
I'm going to cut them some slack in that's somewhat the natural obvious
1587
01:43:20,796 --> 01:43:22,341
thing you're going to do as the CEO,
1588
01:43:24,315 --> 01:43:27,329
whether you're an asshole or not, or whether your company's
1589
01:43:27,829 --> 01:43:31,917
evil and I think I want it to die, that's somewhat of a natural thing
1590
01:43:31,965 --> 01:43:35,559
that you're going to do whether you're an evil person or not.
1591
01:43:35,671 --> 01:43:38,141
So I'm going to cut him some slack on there, but I just want to
1592
01:43:38,157 --> 01:43:41,719
point out that that is effectively what he's doing, is trying to
1593
01:43:41,735 --> 01:43:45,649
use a free army of his customers as
1594
01:43:45,681 --> 01:43:48,135
PR in the local communities.
1595
01:43:49,611 --> 01:43:53,557
Yeah. Well, yeah, of course.
1596
01:43:54,729 --> 01:43:58,196
Um, what he's saying is, you know, you love our
1597
01:43:58,696 --> 01:44:01,824
product, if you want to keep using it, help us out. Which is fair enough.
1598
01:44:02,755 --> 01:44:05,853
You know, I wonder if there's going to be any, uh,
1599
01:44:06,303 --> 01:44:09,513
discounts applied if you, if you have a couple
1600
01:44:10,013 --> 01:44:12,916
articles under your— The Deputy of the Year.
1601
01:44:13,285 --> 01:44:16,496
Yeah, if you're the Deputy of the Year.
1602
01:44:16,961 --> 01:44:20,220
All right, Alex, we are now going to talk a little bit about data
1603
01:44:20,268 --> 01:44:23,703
centers, and I say that's a parallel topic simply because in the,
1604
01:44:23,784 --> 01:44:28,857
in the way that there has been backlash and propaganda
1605
01:44:29,357 --> 01:44:32,647
campaigns of an anti-sentiment against the
1606
01:44:32,872 --> 01:44:36,484
idea. Now, this is less than Flock being a specific
1607
01:44:36,984 --> 01:44:40,707
company and a specific product being targeted. This is more now just a general
1608
01:44:40,804 --> 01:44:44,609
concept. There is a backlash against
1609
01:44:44,657 --> 01:44:48,366
data centers as a concept, and this is where we're
1610
01:44:48,866 --> 01:44:52,202
going to be headed now for a little bit. Company behind data center
1611
01:44:52,702 --> 01:44:55,219
near Nashville Zoo sues Metro Nashville.
1612
01:44:56,631 --> 01:45:00,113
The company behind a controversial proposed data center next to the Nashville Zoo
1613
01:45:00,226 --> 01:45:03,917
is suing the Metro Nashville government over its move to stop the tech
1614
01:45:04,417 --> 01:45:08,747
company from building the data center. The company said in its complaint, Metro's ordinance
1615
01:45:09,247 --> 01:45:12,726
defining data centers and Mayor Freddie O'Connell's executive order barring
1616
01:45:12,758 --> 01:45:16,032
data center development do not apply to their property and
1617
01:45:16,048 --> 01:45:19,193
thus cannot be enforced against them, and they seek a federal court
1618
01:45:19,241 --> 01:45:22,723
to declare Metro violated the company's due process and equal protection
1619
01:45:22,771 --> 01:45:26,157
rights granted by the Constitution. The complaint comes as part
1620
01:45:26,657 --> 01:45:30,249
of Metro's relentless efforts to prohibit DC Block's lawful invested use of
1621
01:45:30,749 --> 01:45:34,725
its property as a data center, the company said in its court rulings— court
1622
01:45:35,225 --> 01:45:39,234
filings, sorry— calling it part of a targeted attack against the company in violation
1623
01:45:39,734 --> 01:45:43,711
of the federal constitutional protections. At issue is the property directly next
1624
01:45:44,211 --> 01:45:47,820
to the Nashville Zoo parking lot, which is now legally owned by DC Blocks.
1625
01:45:48,543 --> 01:45:52,236
Initially reported as a single data center facility, the larger project
1626
01:45:52,736 --> 01:45:56,539
will actually see 2 separate data centers constructed next to the zoo, which has
1627
01:45:56,571 --> 01:45:59,526
drawn significant pushback from the zoo, nearby residents,
1628
01:45:59,847 --> 01:46:03,042
and even Metro Council officials. Alex, I don't know if this has anything to do
1629
01:46:03,154 --> 01:46:05,772
with the, the 2 missing, uh, lions,
1630
01:46:06,671 --> 01:46:10,685
uh, from the earlier, whether, whether they were— They departed in protest.
1631
01:46:12,368 --> 01:46:15,622
Metro moved quickly to define data center uses of the Metro zoning
1632
01:46:15,670 --> 01:46:19,036
code, seeking to place limits on data center sizes and
1633
01:46:19,536 --> 01:46:23,653
locations within Davidson County. The mayor also signed an executive order prohibiting
1634
01:46:23,749 --> 01:46:26,874
large-scale data center development. In their complaint,
1635
01:46:26,890 --> 01:46:30,705
DC Blox says all the moves made by Metro Council and the mayor violated
1636
01:46:30,721 --> 01:46:34,135
state law regarding vested property rights and due process for
1637
01:46:34,635 --> 01:46:38,175
the company. Further, the company says the property previously housed an office and
1638
01:46:38,675 --> 01:46:42,427
data center for Asurion with no issue, and Metro issued permits
1639
01:46:42,927 --> 01:46:46,536
for DC Blox's use of, uh, for the property with no issue before adopting the
1640
01:46:47,036 --> 01:46:50,163
new data center regulations. Most recently, the Metro Council passed
1641
01:46:50,663 --> 01:46:54,802
a bill authorizing the government to pursue eminent domain against
1642
01:46:55,302 --> 01:46:58,846
the property for future Metro use. Metro Legal Director Tyler Yarbrough
1643
01:46:59,280 --> 01:47:02,008
told News 2 her team is evaluating the claims.
1644
01:47:02,345 --> 01:47:05,604
Metro Legal is reviewing the lawsuit and preparing to defend Metro
1645
01:47:05,700 --> 01:47:09,393
against DC Blocks' claims. We are confident that Metro has acted
1646
01:47:09,893 --> 01:47:13,518
lawfully in its dealings with DC Blocks and in the enactment of the
1647
01:47:13,663 --> 01:47:17,291
moratorium on data center permitting, she said in a statement.
1648
01:47:17,355 --> 01:47:20,823
DC Blocks told News 2 in a statement, the filing speaks for itself.
1649
01:47:20,871 --> 01:47:24,403
Now, Alex, if DC Blocks said that they were
1650
01:47:24,903 --> 01:47:28,240
exclusively housing data centers for flock cameras,
1651
01:47:28,850 --> 01:47:32,960
I'm wondering how the interplay would work out in that
1652
01:47:33,460 --> 01:47:37,218
that community, whether or not the government would want
1653
01:47:37,460 --> 01:47:40,237
the flock camera data housed in the data center next to the zoo.
1654
01:47:42,022 --> 01:47:45,330
Not sure how that would play out, but— Is the fact that it is a
1655
01:47:45,830 --> 01:47:48,686
zoo somehow relevant here? The zoo?
1656
01:47:49,283 --> 01:47:52,646
Well, so yes, it is somewhat relevant
1657
01:47:52,744 --> 01:47:55,810
in that the zoo is part of the pushback. The zoo is
1658
01:47:55,924 --> 01:47:59,144
part of the pushback because Because typically speaking,
1659
01:47:59,577 --> 01:48:03,010
part of the complaints with the data centers involve extreme
1660
01:48:03,026 --> 01:48:07,662
noise pollution, and having— if
1661
01:48:08,162 --> 01:48:09,763
you take as a given that they're being— I keep hearing about this, but I
1662
01:48:10,263 --> 01:48:13,886
don't understand. Like, there's been data centers in residential
1663
01:48:13,998 --> 01:48:17,624
areas for a long time, and I— yeah,
1664
01:48:17,848 --> 01:48:21,570
I'm very confused. Like, what— I mean, I could be uninformed.
1665
01:48:22,436 --> 01:48:24,489
What is it that causes this extreme noise pollution?
1666
01:48:28,670 --> 01:48:32,236
Well, I mean, the— I'm not just
1667
01:48:32,736 --> 01:48:35,623
going to say that the fans on the servers, but, you know, you have usually
1668
01:48:35,656 --> 01:48:38,005
a large, you know,
1669
01:48:39,049 --> 01:48:42,279
power converters and things like that, large things that draw
1670
01:48:42,423 --> 01:48:46,119
energy that have a volume of noise that they emit.
1671
01:48:46,199 --> 01:48:51,943
Well, yeah. In fact, Google went so far as to advertise the fact
1672
01:48:52,443 --> 01:48:56,158
that some of their new data centers actually have a lot of enhancements in controlling
1673
01:48:56,207 --> 01:48:59,813
the noise. So there's obviously something that is specifically generating—
1674
01:49:01,152 --> 01:49:04,428
No, well, I think that's the wrong way to read it. I think a
1675
01:49:04,928 --> 01:49:08,028
lot of this stuff, um, I think these data center companies don't know what to
1676
01:49:08,528 --> 01:49:12,506
do. So one strategy is instead of fighting the,
1677
01:49:13,625 --> 01:49:16,823
uh, concern Right?
1678
01:49:16,904 --> 01:49:19,792
Like saying like, no, we don't make noise, is to say like,
1679
01:49:20,210 --> 01:49:23,323
okay, yeah, fine, we used to make 5 decibels of noise and now it's
1680
01:49:23,823 --> 01:49:27,657
gonna be zero. So let's just get it out of the way. I think that's
1681
01:49:28,157 --> 01:49:30,498
what a lot of the, you know, you also hear in new technologies about using
1682
01:49:30,546 --> 01:49:33,948
less water and whatever. I had a conversation
1683
01:49:34,448 --> 01:49:38,378
with somebody yesterday that shocked me a little bit because
1684
01:49:38,426 --> 01:49:39,132
it was like,
1685
01:49:43,262 --> 01:49:46,569
I had— this is a group of friends that, you know, we get a beer
1686
01:49:47,069 --> 01:49:49,704
every once in a while. And in the previous one, I had said, guys,
1687
01:49:49,801 --> 01:49:53,520
the whole water thing with data centers is overblown. So the guy said,
1688
01:49:53,536 --> 01:49:56,430
look, you know, I generally respect your opinion. So I looked it up and it's
1689
01:49:56,527 --> 01:50:00,052
not right because even if they don't use water themselves,
1690
01:50:00,650 --> 01:50:04,612
the energy they use is downstream
1691
01:50:04,692 --> 01:50:08,215
of water usage and therefore that counts.
1692
01:50:08,853 --> 01:50:12,912
Right? And superficially, I realize that this sounds reasonable,
1693
01:50:13,811 --> 01:50:17,019
but in practice, what does that mean? So if we want to preserve water,
1694
01:50:17,035 --> 01:50:20,292
quote unquote, if that is even a problem that the United States has,
1695
01:50:21,672 --> 01:50:25,138
we now have to what, reduce energy usage overall? Like, is that
1696
01:50:25,638 --> 01:50:29,229
where we are as a country being like, yeah, well, you know, can't have energy
1697
01:50:29,729 --> 01:50:33,208
because it wastes water? Like, in a country full of flowing waters,
1698
01:50:33,786 --> 01:50:37,363
it's like, okay, like what even, you know, the original
1699
01:50:37,797 --> 01:50:41,956
complaint had a coherence to it. It was that the cooling systems
1700
01:50:42,614 --> 01:50:44,926
evaporate a bunch of water in order to,
1701
01:50:45,344 --> 01:50:49,165
um, uh, cool the data center. And that is an additional,
1702
01:50:50,080 --> 01:50:53,067
uh, concern about data centers, that they, um,
1703
01:50:53,372 --> 01:50:57,242
waste a lot of presumably drinkable water, which also isn't
1704
01:50:57,742 --> 01:51:01,015
the case, but whatever, right? It had a, like, it was a sort
1705
01:51:01,515 --> 01:51:04,690
of thing that you could imagine being true. Um,
1706
01:51:05,190 --> 01:51:07,370
it turns out that the book that made that case had a, I don't know,
1707
01:51:07,870 --> 01:51:11,530
something like a 1,000x type of error, but fuck it, we're just gonna go with
1708
01:51:12,030 --> 01:51:15,490
it. Oh, I mean, look, within, within the, within like topics like that
1709
01:51:15,990 --> 01:51:19,990
of climate science or, or any type of climate protection science,
1710
01:51:20,490 --> 01:51:24,080
those are, those are easily just normal, uh, types of errors.
1711
01:51:24,310 --> 01:51:27,770
Small. It was a small error. Yes, it's considered a small error in the
1712
01:51:28,270 --> 01:51:31,270
field. Yeah, I understand. Anyway, but that's kind of what I'm saying. It's like one
1713
01:51:31,310 --> 01:51:34,350
I guess what Google was doing there is like, all right, fuck it. Like we're
1714
01:51:34,850 --> 01:51:38,310
not even going to get in this conversation. Let's just quote unquote invent some
1715
01:51:38,810 --> 01:51:42,550
quote unquote new technology that quote unquote, you know, reduces noise by some
1716
01:51:43,050 --> 01:51:46,310
amount and, you know, move past this, which is fair enough.
1717
01:51:46,810 --> 01:51:50,670
And I think it will happen, but you've seen my tweets recently. I will
1718
01:51:51,170 --> 01:51:54,970
say that unfortunately, I think there's a huge misunderstanding here that
1719
01:51:55,470 --> 01:51:58,590
the arguments are not the arguments. The concerns are not the concerns, and therefore the
1720
01:51:59,090 --> 01:52:02,000
solutions are not the solutions. If you see the responses to my post,
1721
01:52:02,040 --> 01:52:05,540
you'll get the vibe that people are— Are you referring to your data center in
1722
01:52:06,040 --> 01:52:09,480
space one? No, no, no, the one that I said,
1723
01:52:09,980 --> 01:52:13,220
well, if they didn't take any energy and they didn't use any water, it'd still
1724
01:52:13,720 --> 01:52:17,260
be hated just as much. And one guy literally said, well, you know, all they're
1725
01:52:17,760 --> 01:52:21,180
used for is holding spock camera data anyway, so fuck 'em. Oh,
1726
01:52:21,680 --> 01:52:25,470
okay, so my suggestion just now wasn't even a good one.
1727
01:52:25,560 --> 01:52:29,104
Okay, well. Well, yeah. So they
1728
01:52:29,329 --> 01:52:32,198
also, let's say, as an icon,
1729
01:52:32,711 --> 01:52:36,863
they represent the security state. They represent the surveillance
1730
01:52:37,363 --> 01:52:41,383
state. That's right. They are representatives
1731
01:52:41,431 --> 01:52:44,974
of what Edward Snowden would've been coming out against.
1732
01:52:45,054 --> 01:52:48,725
That's right. They are all the NSA. They're the NSA,
1733
01:52:48,805 --> 01:52:53,198
exactly. But that's what I'm saying. They're a totem
1734
01:52:53,214 --> 01:52:56,836
that people attach to them as labels, like the
1735
01:52:57,336 --> 01:53:00,707
feeling they have of loss of control, of uncertainty, about the future of
1736
01:53:00,949 --> 01:53:04,321
totalitarianism, surveillance state, you name it. And to some
1737
01:53:04,821 --> 01:53:07,785
extent, I get that. To a large extent, I get that. It's just that it
1738
01:53:08,285 --> 01:53:11,215
fucking annoys me that everybody has to fucking be talking about the fucking water all
1739
01:53:11,312 --> 01:53:15,273
the time. That's what kills me. The unfortunate
1740
01:53:15,773 --> 01:53:21,349
part is that there will be some number of people who will take the reasons
1741
01:53:21,381 --> 01:53:25,286
provided to them at face value and will strongly
1742
01:53:25,479 --> 01:53:29,031
believe that those are true. So to that extent,
1743
01:53:29,288 --> 01:53:32,885
the noise pollution is part of the complaint of, or the risk
1744
01:53:33,385 --> 01:53:36,610
of noise pollution is part of the complaint of the zoo. So hypothetically
1745
01:53:37,110 --> 01:53:40,672
speaking, if you have this data center that does contribute
1746
01:53:40,688 --> 01:53:44,301
a large amount of noise pollution, it will upset
1747
01:53:44,801 --> 01:53:48,058
the animals. It will give them a lot of stress. Maybe they will have trouble
1748
01:53:48,090 --> 01:53:51,093
sleeping. Right. You know, like I said,
1749
01:53:51,189 --> 01:53:54,909
that might be why the lions left. You know, and just went rogue.
1750
01:53:54,989 --> 01:53:58,428
Yes. So that would be the concern
1751
01:53:58,606 --> 01:54:02,094
from the zoo, and whether or not they would be
1752
01:54:02,725 --> 01:54:05,879
justified in that, maybe they truly believe that it exists.
1753
01:54:08,880 --> 01:54:13,344
Yeah. Sorry, I'm just trying to make a
1754
01:54:13,846 --> 01:54:17,324
remake of Lion King, something like, in the center of the data center,
1755
01:54:17,405 --> 01:54:19,647
the lion doesn't sleep, or something like that.
1756
01:54:22,316 --> 01:54:25,779
All right, well, we'll, we'll work on, we'll work on a music video
1757
01:54:26,279 --> 01:54:27,802
for that for next, um,
1758
01:54:29,987 --> 01:54:34,006
um, now Alex, Alex, people are getting very passionate
1759
01:54:34,184 --> 01:54:38,008
about these things. Now look, I, I am a huge Terminator—
1760
01:54:38,508 --> 01:54:40,783
The funniest one was like, uh, why don't they do something about the noise?
1761
01:54:40,897 --> 01:54:44,209
I hear they make a horrendous noise. It's like, okay, all right.
1762
01:54:47,355 --> 01:54:51,141
The problem is that you hear about it. Like, you not— you hear this,
1763
01:54:51,271 --> 01:54:54,276
the noise. You hear theoretical concerns about theoretical noise.
1764
01:54:54,356 --> 01:54:57,107
Like, um,
1765
01:54:57,496 --> 01:55:01,023
yeah. Well, have you done— have you done any dives?
1766
01:55:01,072 --> 01:55:04,848
This whole fucking thing feels like an entire— like, like a massive
1767
01:55:05,348 --> 01:55:08,567
distraction. You know what? I would be happy if I was wrong, right? I would
1768
01:55:09,067 --> 01:55:12,254
be happy if indeed a new version— well, I would— would I be happy?
1769
01:55:12,350 --> 01:55:15,883
I don't know. There will be some happiness in me if a new version
1770
01:55:16,383 --> 01:55:19,866
of data centers comes out and people are like, hey, these don't use as
1771
01:55:19,898 --> 01:55:23,688
much water and don't make as much noise or whatever, that these are actually really
1772
01:55:24,188 --> 01:55:27,301
tolerable. We're very happy to have these or something.
1773
01:55:27,366 --> 01:55:30,497
But like, my understanding is this,
1774
01:55:31,541 --> 01:55:34,833
we're not like, the whole conversation is ridiculous because this is not the reason.
1775
01:55:35,058 --> 01:55:38,816
People just feel the need to bring up some argument and they do. And if
1776
01:55:38,832 --> 01:55:42,253
all of these are solved, The next one will be they take up so much
1777
01:55:42,753 --> 01:55:45,962
space. It's like, yes, America is lacking space. That is the well,
1778
01:55:46,459 --> 01:55:49,831
well, the defining feature of the country. Well, Alex, so good to go down the
1779
01:55:50,331 --> 01:55:53,059
chain. They take up space, and if then they made them more compact, it's the
1780
01:55:53,559 --> 01:55:57,538
fact that now they they are stealing jobs from everybody because because
1781
01:55:58,038 --> 01:56:01,440
they they are making the they are making the image generators that are stealing the
1782
01:56:01,472 --> 01:56:05,117
jobs from artists. They are they are the LLMs that
1783
01:56:05,617 --> 01:56:08,859
are now now, uh, unemploying all the mathematicians, which we're gonna get into.
1784
01:56:13,376 --> 01:56:16,679
Um, I don't know why that's funny. You know, but, but, but at least that's
1785
01:56:17,179 --> 01:56:20,608
starting to get somewhere where you're getting closer to what I'm telling
1786
01:56:20,640 --> 01:56:23,244
you, like, my sense of the actual issue.
1787
01:56:23,713 --> 01:56:26,285
Like, what, what is being done in there? Right?
1788
01:56:26,770 --> 01:56:30,706
Yeah. And the fact that these things are sealed, uh, environments where,
1789
01:56:30,722 --> 01:56:34,169
you know, stuff happens and you're free to assume the worst.
1790
01:56:35,009 --> 01:56:37,905
It's kind of the issue, and, and I don't know how you solve it,
1791
01:56:37,970 --> 01:56:41,381
but hey. Well, so I was
1792
01:56:41,881 --> 01:56:45,179
saying, just right, just because we're moving on to the next story here, um,
1793
01:56:45,356 --> 01:56:48,561
I'm a, I'm a fan of the Terminator, uh, universe,
1794
01:56:48,915 --> 01:56:52,199
Alex. The, the original movie, Terminator 2 in particular,
1795
01:56:52,361 --> 01:56:55,808
and, um, the only worthy sequel to Terminator 2, which is
1796
01:56:55,840 --> 01:56:59,142
Terminator: The Sarah Connor Chronicles, which is a television show that only
1797
01:56:59,642 --> 01:57:03,322
got 2 seasons and it was taken from us. way too early,
1798
01:57:03,435 --> 01:57:07,283
but it was the only worthy follow-up to Terminator
1799
01:57:07,783 --> 01:57:11,773
2. And I have a lot of affinity. You can find
1800
01:57:12,273 --> 01:57:15,797
my timeline in the past where before the LLM explosion in the modern
1801
01:57:16,006 --> 01:57:20,399
AI era came up, I was constantly asking people if any
1802
01:57:20,576 --> 01:57:24,231
of the good ideas they had had any utility in stopping
1803
01:57:24,280 --> 01:57:27,791
the robots, because the robots will eventually be coming. Right.
1804
01:57:27,951 --> 01:57:31,191
So I have a lot of affinity to preventing the doom,
1805
01:57:31,675 --> 01:57:34,901
the Judgment Day that Skynet will eventually bring to us,
1806
01:57:35,304 --> 01:57:38,836
which I believe was just yesterday. It was either today or yesterday
1807
01:57:39,336 --> 01:57:42,453
is the— it was yesterday is the
1808
01:57:42,533 --> 01:57:44,735
anniversary of Judgment Day from Terminator 2.
1809
01:57:46,407 --> 01:57:49,553
That would have been prevented. So, you know,
1810
01:57:50,021 --> 01:57:52,429
happy Judgment Day to all those who celebrate.
1811
01:57:53,721 --> 01:57:57,678
But there are people who are
1812
01:57:58,178 --> 01:58:00,312
They are imagining themselves modern-day Sarah Connor.
1813
01:58:01,831 --> 01:58:06,341
They are the Sarah Connors who are the justified
1814
01:58:06,841 --> 01:58:08,879
terrorists in the technological fight.
1815
01:58:10,221 --> 01:58:13,443
They also remember that they are antinatalists,
1816
01:58:13,475 --> 01:58:16,543
so they can't be Sarah Connor. Oh, sorry. Well,
1817
01:58:17,619 --> 01:58:21,138
man charged after setting himself ablaze in Google's New York City
1818
01:58:21,234 --> 01:58:25,183
office building. A 63-year-old man set himself on fire Friday
1819
01:58:25,683 --> 01:58:29,000
afternoon in the lobby of a Lower Manhattan building that houses Google's New York offices
1820
01:58:29,048 --> 01:58:32,063
in an incident police say appears to have no clear motive.
1821
01:58:32,496 --> 01:58:36,184
John Butch of Brooklyn walked into the lobby of 315 Hudson Street around
1822
01:58:36,248 --> 01:58:39,599
3 PM and set himself ablaze, according to the police. He was
1823
01:58:39,615 --> 01:58:42,406
taken to a nearby hospital with a burn unit in stable condition.
1824
01:58:42,838 --> 01:58:46,831
No one else at the scene was injured. Butch was charged with criminal mischief,
1825
01:58:46,928 --> 01:58:50,038
arson, and reckless endangerment. Police have not disclosed
1826
01:58:50,538 --> 01:58:53,149
a motive, and it remains unclear whether Butch intended to target Google specifically.
1827
01:58:53,824 --> 01:58:57,099
In a statement, Google suggested it did not believe it was the intended target.
1828
01:58:57,436 --> 01:59:01,242
While the individual appeared to be seeking access to another tenant's space in the building,
1829
01:59:01,290 --> 01:59:04,068
we remain in close contact with local authorities, the statement said.
1830
01:59:04,421 --> 01:59:08,435
We are grateful to first responders. The recently renovated Hudson Square building also houses
1831
01:59:08,935 --> 01:59:12,321
a Sweetgreen and Joe and the Juice on its ground floor.
1832
01:59:12,465 --> 01:59:16,174
So I, I don't think that he was going after Joe and the Juice,
1833
01:59:16,624 --> 01:59:18,198
but maybe, who knows?
1834
01:59:20,271 --> 01:59:21,985
Depends on who he is expelling.
1835
01:59:24,460 --> 01:59:25,975
Now, this is from Axios, Alex.
1836
01:59:28,279 --> 01:59:32,160
Climate group presses Dems to seize on data center backlash.
1837
01:59:32,482 --> 01:59:36,282
A climate group that backs Democrats is giving the party's 2028 hopefuls
1838
01:59:36,653 --> 01:59:40,201
and campaign committees advice on how to capitalize on waning public support
1839
01:59:40,701 --> 01:59:44,556
for data centers, a new memo shows. Data centers have landed atop the political
1840
01:59:44,572 --> 01:59:48,264
radar, and Climate Power's memo underscores the scramble to turn the backlash
1841
01:59:48,764 --> 01:59:51,647
into a political advantage ahead of the midterms in 2028.
1842
01:59:52,513 --> 01:59:56,298
It says public opinion on data centers is cratering and argues this provides
1843
01:59:56,798 --> 02:00:00,980
a political opening around energy costs and environmental concerns. The top-line
1844
02:00:01,012 --> 02:00:05,085
message is that Trump and Republicans have given AI data centers a blank
1845
02:00:05,585 --> 02:00:08,853
check to raise our utility bills, poison our air, and drain our water,
1846
02:00:08,869 --> 02:00:12,413
Climate Powers argues. Democrats can draw a contrast with
1847
02:00:12,461 --> 02:00:16,116
President Trump's AI and data center advocacy by uniting
1848
02:00:16,164 --> 02:00:19,947
around 3 principles. It states, tech companies paying their own way,
1849
02:00:20,492 --> 02:00:24,243
powering data centers with clean energy, giving communities
1850
02:00:24,743 --> 02:00:27,673
a real say in ensuring they're informed about water and energy usage.
1851
02:00:28,250 --> 02:00:31,728
Candidates' positions on data centers don't split among partisan lines
1852
02:00:32,097 --> 02:00:34,966
as various GOP candidates and leaders offer new restrictions.
1853
02:00:35,880 --> 02:00:40,047
Demands that tech companies and developers don't offload costs onto consumers are common
1854
02:00:40,547 --> 02:00:44,214
in both parties, and the White House has the same position. And Democrats don't
1855
02:00:44,714 --> 02:00:48,348
have a united front, with high-profile incumbents and candidates taking a variety
1856
02:00:48,364 --> 02:00:51,776
of approaches. Jesse Lee, a senior adviser to Climate Power,
1857
02:00:51,904 --> 02:00:55,493
argues that Republicans are politically saddled with President Trump's overall support
1858
02:00:55,993 --> 02:00:59,579
for the sector and will bend to his goals. The memo says that while some
1859
02:00:59,675 --> 02:01:03,808
GOP candidates are now emphasizing data center accountability, Trump owns
1860
02:01:03,840 --> 02:01:07,029
the party's position. If Republicans keep control of the White
1861
02:01:07,529 --> 02:01:11,050
House and Senate, GOP lawmakers will be an army for data centers on Trump's
1862
02:01:11,550 --> 02:01:15,440
behalf, Lee said in an interview. interview. He also said that while Democrats
1863
02:01:15,940 --> 02:01:19,349
have room to debate specific policies, they're broadly united around the memo's principles.
1864
02:01:19,925 --> 02:01:24,475
The memo says that debate could include whether to pursue real enforced guardrails
1865
02:01:24,908 --> 02:01:27,359
or a temporary moratorium on data centers.
1866
02:01:28,176 --> 02:01:31,797
Data center politics are evolving quickly as the midterms loom and Democrats
1867
02:01:32,297 --> 02:01:35,754
jockey for position ahead of 2028. Climate Power said it shared the
1868
02:01:35,786 --> 02:01:39,134
memo with the Democratic National Committee, the Democratic Senatorial Campaign
1869
02:01:39,166 --> 02:01:42,787
Committee, the Democratic Congressional Campaign Committee, and the Democratic Governors
1870
02:01:43,287 --> 02:01:47,138
Association. The group declined to offer a list of potential 2828 hopefuls
1871
02:01:47,638 --> 02:01:50,609
who got it, though Lee said it's pretty much everybody that has been in the
1872
02:01:51,109 --> 02:01:53,792
conversation. Now, Alex,
1873
02:01:54,148 --> 02:01:57,971
the one issue that data
1874
02:01:58,471 --> 02:02:02,796
centers bring that I believe is strongly
1875
02:02:04,852 --> 02:02:09,288
important to review and consider does
1876
02:02:09,486 --> 02:02:13,450
involve energy consumption. But not necessarily in
1877
02:02:13,498 --> 02:02:15,781
a climate-specific way.
1878
02:02:18,090 --> 02:02:22,823
So the degree to which a data center might increase the
1879
02:02:23,131 --> 02:02:26,549
consumption on an area's energy
1880
02:02:27,838 --> 02:02:31,446
system does not mean that
1881
02:02:31,913 --> 02:02:36,006
the cost of energy is going to scale linearly to the increased consumption.
1882
02:02:36,378 --> 02:02:40,759
So what happens in certain locations, depending on
1883
02:02:40,808 --> 02:02:44,303
the grid structure itself and how much
1884
02:02:44,400 --> 02:02:46,438
energy can be maintained in such a grid.
1885
02:02:47,650 --> 02:02:51,235
Let's pretend, just as generically speaking, adding a data center
1886
02:02:51,735 --> 02:02:57,064
to a certain community doubles the wattage consumption that
1887
02:02:57,564 --> 02:03:01,244
is there, just doubles it. That doubling does not necessarily
1888
02:03:01,277 --> 02:03:05,010
come at an equal marginal expense per watt.
1889
02:03:06,278 --> 02:03:09,777
It can be that the actual average cost
1890
02:03:10,277 --> 02:03:13,840
to produce, uh, per watt under that strain goes up because
1891
02:03:13,952 --> 02:03:17,725
the, the, um, the additional capacity to generate the wattage
1892
02:03:18,127 --> 02:03:22,510
is at a much lower margin. I will say that,
1893
02:03:22,847 --> 02:03:26,251
um, if we're going to talk about something like that, uh, data centers are
1894
02:03:26,751 --> 02:03:29,671
generally baseload, right? So they take as much as they take,
1895
02:03:30,394 --> 02:03:33,397
um, because they usually try to run them at high utilization.
1896
02:03:34,006 --> 02:03:37,516
Um, which means that, you know, the, it's the power
1897
02:03:37,532 --> 02:03:41,442
usage that tends to require expensive things
1898
02:03:41,538 --> 02:03:44,230
like, um, gas, um,
1899
02:03:44,743 --> 02:03:48,124
and, and all sorts of different types of, uh,
1900
02:03:48,204 --> 02:03:51,778
peaker, peaker stations is domestic, right? Because,
1901
02:03:52,114 --> 02:03:55,031
uh, I mean, the most famous example of this is that in Britain,
1902
02:03:55,880 --> 02:03:59,021
um, it was a kind of a national, uh,
1903
02:04:01,909 --> 02:04:05,423
What do you call it? Like a pattern, like an extremely strong pattern that
1904
02:04:05,923 --> 02:04:09,001
everybody would go home after work, they would watch a particular program on the BBC,
1905
02:04:09,451 --> 02:04:12,692
and then when they were done, I shit you not, this is a real fact,
1906
02:04:12,740 --> 02:04:16,511
they would go and turn on the
1907
02:04:17,011 --> 02:04:20,875
kettle to make tea. And that spike was like humongous.
1908
02:04:20,987 --> 02:04:24,469
And like, there's like a whole documentary about like the amount of infrastructure
1909
02:04:25,095 --> 02:04:28,784
that UK power system had to
1910
02:04:29,284 --> 02:04:32,630
put up with to deal with that, like, once in a day, like,
1911
02:04:32,694 --> 02:04:36,796
extreme spike, right? Because you have to— I'm
1912
02:04:37,296 --> 02:04:40,674
not kidding you. This is a real thing. I believe you. I just
1913
02:04:40,706 --> 02:04:44,488
had never heard of that specific anecdote. So I'm curious.
1914
02:04:44,552 --> 02:04:47,308
It's very funny. But yeah. Yeah.
1915
02:04:47,597 --> 02:04:51,699
Anyway, but long story short, the thing is that your grid
1916
02:04:52,199 --> 02:04:56,207
has to be built to support the peaks. It can't be built to support your
1917
02:04:56,707 --> 02:05:00,936
average. Right. And the data center lifts the average, which is better than
1918
02:05:00,968 --> 02:05:04,753
lifting your peak, or at least, you know, in certain ways
1919
02:05:05,253 --> 02:05:08,457
just as bad, but at least pays you throughout rather than paying you for like
1920
02:05:08,957 --> 02:05:12,455
one moment. Anyway, so I don't
1921
02:05:12,955 --> 02:05:17,461
know if this helps your example, but my sense is that data
1922
02:05:17,961 --> 02:05:21,325
centers generally either bring in their own power, Like, uh, gas turbines.
1923
02:05:21,421 --> 02:05:25,659
Like, um, SpaceX now bought a turbine company because I'm assuming
1924
02:05:25,675 --> 02:05:29,239
they expect that, um, that this is just how
1925
02:05:29,303 --> 02:05:33,573
they're gonna do it. Um, or they
1926
02:05:34,665 --> 02:05:36,800
work with the power company and they get advance,
1927
02:05:37,554 --> 02:05:40,910
um, approval. Yes, we can support this additional amount.
1928
02:05:41,584 --> 02:05:44,586
But— Right. Anyway, I don't know if I have a soft floor.
1929
02:05:44,730 --> 02:05:48,292
So my specific— well, my specific, my specific thing of concern is just pointing
1930
02:05:48,388 --> 02:05:52,154
out that when you add that much extra usage,
1931
02:05:52,170 --> 02:05:56,304
it's not necessarily true that that additional consumption is
1932
02:05:57,057 --> 02:06:00,470
just linearly scaling, and therefore the
1933
02:06:00,970 --> 02:06:04,300
marginal cost of an additional watt is the same.
1934
02:06:04,365 --> 02:06:08,499
It's actually— the marginal cost for each additional
1935
02:06:08,999 --> 02:06:11,960
watt might be higher. And the problem comes— Yeah.
1936
02:06:13,307 --> 02:06:16,903
When the grid is going to take on a heavier
1937
02:06:17,561 --> 02:06:20,819
constant load like that, and the increased
1938
02:06:20,931 --> 02:06:24,575
costs to that marginal watt get pushed onto
1939
02:06:25,075 --> 02:06:28,877
the consumers that were already there. And let's just talk about the poorer
1940
02:06:29,377 --> 02:06:33,419
ones. Yeah. So, it should be, in my opinion,
1941
02:06:33,933 --> 02:06:37,496
a relatively easy argument that the data centers
1942
02:06:37,544 --> 02:06:41,075
coming in should be the ones paying that higher marginal cost.
1943
02:06:41,180 --> 02:06:45,150
Not the original consumers. So, so shall we,
1944
02:06:45,240 --> 02:06:48,640
let's put a few, um, pegs down because I think it's important
1945
02:06:49,140 --> 02:06:52,520
to, uh, put a few things out of the way. Hey, I, like all the
1946
02:06:53,020 --> 02:06:56,460
shit about eminent domain and like threatening to sue if you don't put the data
1947
02:06:56,960 --> 02:07:00,640
center here and, you know, asking for special laws and special perks
1948
02:07:01,140 --> 02:07:04,300
and all that shit. Like I'm, I'm opposed to that, like on principle,
1949
02:07:04,800 --> 02:07:08,080
uh, pretty much, right? Like I'm not talking about data centers where there
1950
02:07:08,580 --> 02:07:12,540
is, you know, um, use of lawfare In all sorts of ways,
1951
02:07:13,040 --> 02:07:17,360
or you know they attempt to sort of blackmail their ways into special
1952
02:07:17,860 --> 02:07:21,420
deals, right? I mean, there's some case to be made about you know some states
1953
02:07:21,920 --> 02:07:25,440
want to attract you know types of infrastructure with
1954
02:07:25,940 --> 02:07:29,740
you know special deals. But in general, that should be coming out roughly you
1955
02:07:30,240 --> 02:07:32,940
know as a small advantage. It should not be like a big part of the
1956
02:07:33,440 --> 02:07:36,350
conversation. Because I mean,
1957
02:07:36,540 --> 02:07:38,940
because if you know if the data center is coming out like you're making out
1958
02:07:39,440 --> 02:07:43,047
like that, it's with a with an incentive. Then why is the incentive being Um,
1959
02:07:43,127 --> 02:07:46,733
anyway, but, you know, that is one
1960
02:07:46,781 --> 02:07:50,516
side of the conversation that I think is stupid and short-sighted and hopefully is
1961
02:07:50,629 --> 02:07:53,060
going to be phased out. So I'm not talking about any of that. I'm talking
1962
02:07:53,560 --> 02:07:56,441
about data centers that are being built, you know, either with their own power,
1963
02:07:56,972 --> 02:08:00,514
uh, on-site or, um, with the approval
1964
02:08:01,014 --> 02:08:04,685
of, uh, you know, through advanced sort of approval of the power company
1965
02:08:04,718 --> 02:08:08,117
that they can support the additional load. And here's the thing,
1966
02:08:09,545 --> 02:08:13,056
Both of those cases end up improving the power availability,
1967
02:08:13,555 --> 02:08:17,211
right? A, because if you have more scale, then there's all sorts
1968
02:08:17,711 --> 02:08:21,721
of economies that come in. But B, the data
1969
02:08:22,479 --> 02:08:26,184
centers that do have their own power generation, if they do have a
1970
02:08:26,684 --> 02:08:30,816
moment where they are using less power, they tend
1971
02:08:31,316 --> 02:08:34,565
to dump that back in the grid. So then you have the opposite effect
1972
02:08:34,630 --> 02:08:38,226
of additional availability, which drops Uh,
1973
02:08:38,467 --> 02:08:41,985
prices, right? So I don't know, it's complicated,
1974
02:08:42,001 --> 02:08:45,198
I guess. But yeah, it's— I guess that's another
1975
02:08:45,698 --> 02:08:49,471
part of that that annoys me, which is that things are being presented as a
1976
02:08:49,535 --> 02:08:52,780
lot more straightforward than they are. Well,
1977
02:08:52,796 --> 02:08:56,122
my, my, my point was just to a specific item that
1978
02:08:56,218 --> 02:08:59,319
I don't, uh, that some people consider but not a lot of people consider.
1979
02:08:59,367 --> 02:09:02,853
Now, maybe it doesn't apply in like all of these instances,
1980
02:09:03,046 --> 02:09:06,852
like you're just, just, you're describing You know, ones where the power
1981
02:09:07,302 --> 02:09:10,465
generation is in-house and even pushed back
1982
02:09:10,658 --> 02:09:13,838
onto the broader network in
1983
02:09:13,950 --> 02:09:17,483
overage times, which is a completely different dynamic than the
1984
02:09:17,499 --> 02:09:20,823
one that I'm just specifically mentioning. But where the one
1985
02:09:21,016 --> 02:09:24,645
that I'm talking about applies, I think it should be, you know,
1986
02:09:24,790 --> 02:09:28,628
a reasonable discussion to have to say that the— Sure.
1987
02:09:28,804 --> 02:09:32,113
The data center company should be the one who's
1988
02:09:32,613 --> 02:09:36,285
paying the higher marginal cost for that additional wattage. consumption. That's— that was just the,
1989
02:09:36,686 --> 02:09:40,438
the only specific point I wanted to— Let me put it this
1990
02:09:40,938 --> 02:09:43,982
way. If, if there was a conversation about, um, hey,
1991
02:09:44,287 --> 02:09:47,734
these are the things I'm concerned about this new project, and if
1992
02:09:48,234 --> 02:09:51,454
they were to be, uh, redeemed, then I would be,
1993
02:09:52,224 --> 02:09:54,405
uh, remediated, uh, I would be,
1994
02:09:55,190 --> 02:09:59,391
um, you know, truly happy to support the project. That's totally
1995
02:09:59,891 --> 02:10:03,031
fine. Like, that's how politics works, right? Right.
1996
02:10:03,144 --> 02:10:07,153
Whatever. But my sense is that there is just this deep
1997
02:10:07,233 --> 02:10:10,600
visceral reaction. Yeah. And it
1998
02:10:11,100 --> 02:10:14,770
is not of that kind. And, and if you just see some of
1999
02:10:15,270 --> 02:10:18,987
the responses I got to my message, which didn't even support data centers
2000
02:10:19,019 --> 02:10:22,547
per se, um, it was— it's just
2001
02:10:23,047 --> 02:10:26,957
like, you can't argue with this, right? It's just, you know,
2002
02:10:26,973 --> 02:10:30,008
like, we, we found the object of our hate,
2003
02:10:30,652 --> 02:10:34,259
and we shall oppose it. And you know what really annoys
2004
02:10:34,275 --> 02:10:37,560
me from a dissident point of view is that this problem
2005
02:10:38,060 --> 02:10:40,359
will be solved one way or another. They'll put them in space if they have
2006
02:10:40,393 --> 02:10:40,477
to.
2007
02:10:43,526 --> 02:10:45,701
Um, and the concerns themselves are not going to be addressed.
2008
02:10:46,716 --> 02:10:49,841
Like, we're putting a ton of energy on something and making
2009
02:10:49,938 --> 02:10:53,555
arguments that are completely just, like, ridiculous in many cases.
2010
02:10:54,542 --> 02:10:57,921
Um, And with no focus on getting
2011
02:10:58,295 --> 02:11:01,752
something that actually matters, like a digital Bill of Rights. I don't
2012
02:11:02,252 --> 02:11:03,995
know. Something along those lines.
2013
02:11:07,260 --> 02:11:10,714
Well, Alex, the Bill of Rights, while written
2014
02:11:10,778 --> 02:11:14,384
originally on paper, has been digitized. So that's actually
2015
02:11:14,884 --> 02:11:18,312
already done. Oh, good news. The Bill of Rights is already digital. So now,
2016
02:11:18,506 --> 02:11:22,322
Alex, I want to preface this next article
2017
02:11:22,822 --> 02:11:26,758
here with, I didn't write the article. So if you whatever
2018
02:11:26,790 --> 02:11:30,411
your reaction is to it, do not blame me for the contents
2019
02:11:30,475 --> 02:11:34,080
or the way that it was written. Okay? Just do not blame. Do not
2020
02:11:34,096 --> 02:11:37,220
blame your lovely host for for this article. Okay, sir.
2021
02:11:38,101 --> 02:11:41,738
China is secretly fueling America's data center rage by Axios.
2022
02:11:43,132 --> 02:11:46,994
Roughly 200 accounts from a suspected Chinese bot farm have quietly tried
2023
02:11:47,494 --> 02:11:51,144
to influence Americans to oppose AI data centers on social media,
2024
02:11:51,224 --> 02:11:54,720
X said Thursday night. US and China are
2025
02:11:54,849 --> 02:11:58,348
racing to build the world's most advanced AI systems, and it appears Chinese
2026
02:11:58,444 --> 02:12:01,734
propaganda has infiltrated American circles with the hope of slowing the
2027
02:12:02,234 --> 02:12:06,405
US down. The X safety team probed suspected Chinese accounts involved
2028
02:12:06,421 --> 02:12:09,615
in influence operations and identified 200,000 accounts,
2029
02:12:10,016 --> 02:12:13,290
including 200 accounts posting in a manner that
2030
02:12:13,790 --> 02:12:17,415
could manipulate a legitimate debate about American AI and energy policy, the team said
2031
02:12:17,915 --> 02:12:21,814
in a tweet. The post included content on how AI strains
2032
02:12:21,878 --> 02:12:25,427
the electricity grid and increases utility prices and, quote,
2033
02:12:25,491 --> 02:12:28,863
cartoons that depicted data center operations enriching themselves at
2034
02:12:29,363 --> 02:12:33,086
the public's expense. It's unclear what, if any, action X will
2035
02:12:33,134 --> 02:12:35,815
take to suspend or delete accounts from the bot farm.
2036
02:12:36,490 --> 02:12:40,038
X, owned by Elon Musk, whose billions rely on data centers to boost his
2037
02:12:40,183 --> 02:12:43,314
AI enterprises, did not respond to Axios's request for
2038
02:12:43,330 --> 02:12:47,113
comment. Congressional Republicans have been digging into influence campaigns
2039
02:12:47,145 --> 02:12:51,103
from foreign adversaries designed to stall American AI development. The House
2040
02:12:51,603 --> 02:12:55,013
Energy and Commerce Committee sent a letter in June asking the FBI and White House
2041
02:12:55,513 --> 02:12:59,644
advisers to brief lawmakers on alleged propaganda campaigns by foreign adversaries—
2042
02:12:59,773 --> 02:13:03,827
specifically China, Russia, and Iran— aimed at hindering U.S. AI development and
2043
02:13:04,327 --> 02:13:08,202
data center expansion. The letter cited several reports, including one estimating
2044
02:13:08,250 --> 02:13:12,767
that opposition campaigns Have delayed or obstructed roughly $45.8
2045
02:13:13,267 --> 02:13:15,549
billion in data center projects in Virginia alone.
2046
02:13:16,244 --> 02:13:20,486
Nonetheless, Democrat and Republican officials alike have increasingly backed data center moratoriums
2047
02:13:20,904 --> 02:13:24,696
and embraced a not-in-my-backyard stance toward the massive infrastructure
2048
02:13:25,196 --> 02:13:28,571
projects. Oh no, that means that the acronym NIMBY is going to
2049
02:13:28,637 --> 02:13:32,108
be used now soon. I hate that acronym.
2050
02:13:35,208 --> 02:13:38,432
Axio, data center pushback has become a financial liability for
2051
02:13:38,932 --> 02:13:42,848
AI buildouts. Several Wall Street banks have recently cited political opposition as
2052
02:13:43,025 --> 02:13:46,472
a risk to the entire AI sector. The boom,
2053
02:13:46,649 --> 02:13:50,850
the AI boom relies on expanding data centers because they provide the computing
2054
02:13:50,963 --> 02:13:54,635
that massive AI companies need to keep growing. The all-in
2055
02:13:54,747 --> 02:13:58,403
chattering class Chinese psyop thing is worrying
2056
02:13:58,500 --> 02:14:01,643
and frustrating, Jim Prosser, a Silicon Valley
2057
02:14:02,143 --> 02:14:05,539
communications consultant and former Twitter head of corporate communications, told Axios.
2058
02:14:05,876 --> 02:14:09,308
He noted that this is the new messaging big tech wants to use to change
2059
02:14:09,324 --> 02:14:13,285
the conversation on data centers. Quote, if you get both Greg
2060
02:14:13,333 --> 02:14:17,342
Abbott and Kathy Hochul agreeing on something, it's probably not a Chinese psyop.
2061
02:14:18,096 --> 02:14:21,175
It's probably not a mass delusion people are having on these issues.
2062
02:14:22,073 --> 02:14:25,007
Quote, it's about touching grass and actually being somewhere.
2063
02:14:25,456 --> 02:14:29,899
If you're a venture capitalist in Atherton or Menlo Park opining
2064
02:14:30,171 --> 02:14:33,893
about how somebody in Ohio should feel and you've never been to Ohio,
2065
02:14:34,503 --> 02:14:38,211
that's a problem. Whatever role foreign influence campaigns may
2066
02:14:38,711 --> 02:14:41,149
play, organic opposition to data centers is very real.
2067
02:14:41,839 --> 02:14:45,579
61% of Americans oppose building new data centers in their area,
2068
02:14:45,595 --> 02:14:48,485
according to an early August University of Pennsylvania survey.
2069
02:14:49,047 --> 02:14:52,305
That's up 12 percentage points from a previous UPenn survey conducted in
2070
02:14:52,337 --> 02:14:55,692
February and March. That squares with a YouGov survey released
2071
02:14:56,192 --> 02:14:59,400
Tuesday that found only 24% of Americans say building new
2072
02:14:59,900 --> 02:15:03,505
data centers centers as very or somewhat good for the country. 47% say it's
2073
02:15:03,553 --> 02:15:06,919
bad. Flashback: OpenAI disrupted 2
2074
02:15:07,419 --> 02:15:10,622
Chinese-linked covert networks using ChatGPT to create fake content to
2075
02:15:11,122 --> 02:15:14,565
influence U.S. debates on AI energy consumption and tech tariffs, the company
2076
02:15:15,065 --> 02:15:18,717
said in a June report. The campaigns failed to catch on, but they signaled
2077
02:15:19,217 --> 02:15:22,612
an active push by foreign adversaries to exploit American economic anxieties.
2078
02:15:24,103 --> 02:15:27,245
It doesn't take much social media scrolling to see real videos of real
2079
02:15:27,277 --> 02:15:30,836
people showing up in droves to local council meetings to fight back against data centers.
2080
02:15:30,884 --> 02:15:34,447
Now, Alex, I you the preface because I didn't want you
2081
02:15:34,947 --> 02:15:38,718
to blame me as I read that 200 accounts
2082
02:15:39,232 --> 02:15:42,797
were what was being blamed on the propaganda
2083
02:15:42,861 --> 02:15:46,490
there. You know, like the, the quantity of 200,
2084
02:15:47,020 --> 02:15:50,681
uh, reads as preposterous to me to actually even care about it. I wouldn't even
2085
02:15:50,793 --> 02:15:53,972
really necessarily care about 200,000 accounts, uh,
2086
02:15:54,101 --> 02:15:57,922
because you're, you're either accepting the information with, with discernment,
2087
02:15:57,954 --> 02:16:01,518
or you're not accepting the information with no discernment, and you're thinking about it
2088
02:16:01,566 --> 02:16:05,098
and doing your due diligence. I don't really see how the number of accounts
2089
02:16:05,483 --> 02:16:09,577
really matters in this particular scenario.
2090
02:16:10,379 --> 02:16:14,168
However, Alex, you have an opponent in that article who says that you
2091
02:16:14,668 --> 02:16:18,631
need to touch grass, because you're not seeing that
2092
02:16:18,663 --> 02:16:21,922
the complaints are real, and that everybody has
2093
02:16:22,034 --> 02:16:25,468
real tangible— Yes. Well, I I'll
2094
02:16:25,968 --> 02:16:29,397
have them know that I touched grass. Okay. Just yesterday. No.
2095
02:16:29,509 --> 02:16:33,405
So fresh off my grass touching. How tall? Even while I
2096
02:16:33,905 --> 02:16:37,590
was having, I was touching grass. I was you know
2097
02:16:38,792 --> 02:16:43,170
toe high, I would say. Okay. But even while I was touching
2098
02:16:43,202 --> 02:16:46,649
said grass, I I had the same thoughts. So now what?
2099
02:16:47,451 --> 02:16:50,738
Anyway, Gator, I was saying. Sorry, my mic was
2100
02:16:51,238 --> 02:16:54,685
off. I was saying I don't blame you. I blame the Chinese. For what
2101
02:16:55,185 --> 02:16:57,865
appears to be a bad copy of the Cambridge Analytica debacle.
2102
02:16:58,909 --> 02:17:02,234
Right. Like, who the fuck is doing this thing where it's like,
2103
02:17:02,443 --> 02:17:05,592
we caught 200,000 fake accounts,
2104
02:17:06,941 --> 02:17:10,266
out of which 200 were posting stupid memes? I don't
2105
02:17:10,766 --> 02:17:12,820
know if you've seen some of them, but they were— I have not. As bad
2106
02:17:13,320 --> 02:17:16,595
as you expect. Yeah, I've not seen the memes, but— Okay.
2107
02:17:16,755 --> 02:17:20,337
Now pray tell. But it does have echoes of the 2016 election.
2108
02:17:21,477 --> 02:17:24,780
Of the blaming Russia for Trump. Right, right,
2109
02:17:25,325 --> 02:17:29,236
exactly. It's like, pray tell, what were the other 199,800
2110
02:17:29,429 --> 02:17:33,164
accounts doing? Because if they were doing
2111
02:17:33,196 --> 02:17:34,943
nothing, sorry, but like,
2112
02:17:35,985 --> 02:17:39,336
you're, you know, you, you've got the wrong query.
2113
02:17:40,154 --> 02:17:43,167
Um, and if they were doing something, then why are you talking to me about
2114
02:17:43,215 --> 02:17:48,041
200 of them? Like, I want to know that, you know, this is like
2115
02:17:48,207 --> 02:17:52,747
Like, let me put it different. If you had 200,000
2116
02:17:53,247 --> 02:17:56,623
fake accounts that were posting about all sorts of things, you could literally find
2117
02:17:57,123 --> 02:18:00,261
200 talking about any particular thing and therefore make a post like that.
2118
02:18:00,889 --> 02:18:05,317
And it would mean nothing. Like, what are you talking about?
2119
02:18:05,703 --> 02:18:09,375
I don't know. It's just also stupid. Yes.
2120
02:18:09,568 --> 02:18:13,175
Well, it does have echoes of the 2016 election with the Russia and
2121
02:18:13,272 --> 02:18:15,966
then the Facebook. It was like it was something along the lines of they spent
2122
02:18:16,466 --> 02:18:19,927
$10,000 on Facebook ads or
2123
02:18:20,427 --> 02:18:24,310
something like that and saying, Jesus would have voted for Trump, or something
2124
02:18:24,342 --> 02:18:28,115
like that. And it's like, okay, well, I don't think that that's what moved the
2125
02:18:28,615 --> 02:18:32,531
needle here. Exactly. And by the way, this is what the defenders now of
2126
02:18:33,031 --> 02:18:37,830
the data centers are spinning up. This whole tall tale about data
2127
02:18:38,330 --> 02:18:41,683
centers are— people hate them because the Chinese want them
2128
02:18:42,183 --> 02:18:46,272
to hate them or whatever. Counter counterattack.
2129
02:18:46,320 --> 02:18:49,607
It's like, Gator, I sense
2130
02:18:50,107 --> 02:18:53,085
that you probably lived your life, a lot of your life this way. But I'm
2131
02:18:53,951 --> 02:18:57,077
for some reason this feels novel to me where I'm just like watching a debate
2132
02:18:57,381 --> 02:19:01,020
happen where both sides are just like getting angrier
2133
02:19:01,068 --> 02:19:04,258
and angrier at each other, and all the arguments are dumb,
2134
02:19:04,771 --> 02:19:07,945
like just absolutely retarded on on on both.
2135
02:19:08,394 --> 02:19:12,033
You know, that that has been generally where I have been
2136
02:19:12,225 --> 02:19:15,575
my yes my whole life. It's never been.
2137
02:19:15,752 --> 02:19:18,894
visible to me and I fucking hate it. And I'm not trying to play centrist
2138
02:19:19,394 --> 02:19:22,894
or whatever. I'm just saying, like, this,
2139
02:19:23,023 --> 02:19:24,659
you know, the issue is not the issue.
2140
02:19:26,249 --> 02:19:30,067
China's disinformation campaigns are not the issue. The water and the energy and
2141
02:19:30,567 --> 02:19:32,105
the whatevers are probably not the issue.
2142
02:19:33,935 --> 02:19:37,943
There's another conversation we're trying to have, perhaps, as a country, if that
2143
02:19:38,072 --> 02:19:41,818
sentence even makes any sense. And we seem to be dead
2144
02:19:41,850 --> 02:19:45,292
set on figuring out how not to have it. But talk about,
2145
02:19:46,114 --> 02:19:49,435
I don't know, lions not being able
2146
02:19:49,935 --> 02:19:53,197
to sleep or something. Well, in the jungle,
2147
02:19:53,633 --> 02:19:56,618
the lion sleeps tonight, but not next to the data center at the zoo.
2148
02:19:56,731 --> 02:20:00,312
Not, not next to that stupid fan there. That's right.
2149
02:20:00,990 --> 02:20:03,896
How could it? Alright. Okay. Well, uh,
2150
02:20:05,430 --> 02:20:08,127
next here, Alex, we're going to move forward here,
2151
02:20:08,482 --> 02:20:12,553
um, before we get to Well, let's see, we have traces of AI dystopia
2152
02:20:13,053 --> 02:20:16,522
coming up, but we also may take a detour
2153
02:20:17,599 --> 02:20:20,026
into square root of 2 denialism.
2154
02:20:24,210 --> 02:20:28,191
But first, Meta settled a lawsuit
2155
02:20:28,691 --> 02:20:31,920
this week. Tonight, Meta, the parent company
2156
02:20:32,420 --> 02:20:36,189
of Facebook and Instagram, agreeing to pay nearly $18 billion over claims of
2157
02:20:36,460 --> 02:20:39,875
trying to entice children to social media. And now the major change
2158
02:20:40,375 --> 02:20:43,226
is coming for teenagers on Instagram. Here's Elizabeth Schulze.
2159
02:20:44,780 --> 02:20:48,500
Tonight, social media giant Meta agreeing to pay up to $18
2160
02:20:48,613 --> 02:20:52,528
billion in a landmark deal that will make sweeping changes to how
2161
02:20:52,642 --> 02:20:56,958
young people use Instagram and Facebook. Meta reaching a settlement with 48
2162
02:20:57,458 --> 02:21:01,192
states, ending a federal case claiming the company used its technology
2163
02:21:01,305 --> 02:21:04,776
and social media platforms to entice, engage, and ultimately
2164
02:21:04,888 --> 02:21:08,497
ensnare youth and teens. This is a major moment to clean up
2165
02:21:08,997 --> 02:21:12,666
an industry that has been hurting our kids. As part of the agreement, for users
2166
02:21:13,166 --> 02:21:16,467
under age 18, Instagram and Facebook will set 2-hour
2167
02:21:16,660 --> 02:21:20,428
daily time limits, block the apps from midnight to 6 AM,
2168
02:21:20,525 --> 02:21:23,780
hide like counts, and ban some beauty filters. Finally,
2169
02:21:23,828 --> 02:21:27,324
something was done. Outside the California courtroom,
2170
02:21:27,340 --> 02:21:31,611
Victoria Hinks emotional. She says her 16-year-old daughter Alexandra
2171
02:21:32,111 --> 02:21:35,914
took her own life after getting addicted to social media. No amount
2172
02:21:36,414 --> 02:21:39,752
of money is ever going to bring my beautiful daughter back. Meta denies
2173
02:21:39,864 --> 02:21:43,300
any wrongdoing, saying in a statement tonight, we want to get this
2174
02:21:43,380 --> 02:21:47,186
right for parents and teens, and calling on TikTok and YouTube to make
2175
02:21:47,686 --> 02:21:51,232
similar changes right away. Meta still faces thousands of ongoing
2176
02:21:51,280 --> 02:21:54,739
lawsuits from school districts and families The company tonight vowing
2177
02:21:54,819 --> 02:21:58,771
to keep fighting in court. David. Elizabeth Schulze in Washington.
2178
02:21:58,900 --> 02:22:02,220
Elizabeth, thank you. So there you go,
2179
02:22:02,301 --> 02:22:05,438
Alex. They settled that lawsuit and paid lots of,
2180
02:22:05,697 --> 02:22:09,611
lots of money, and now they're gonna restrict some of the filters on
2181
02:22:10,111 --> 02:22:13,709
Instagram. It's about, it's about damn time. Those filters have been getting out of
2182
02:22:14,209 --> 02:22:15,995
control for quite a while, Alex.
2183
02:22:18,318 --> 02:22:21,865
Yep. And now I think that'll, that'll solve forever
2184
02:22:22,219 --> 02:22:25,864
the problem of, of social media and and teenagers
2185
02:22:25,897 --> 02:22:29,476
getting too caught up in their social status
2186
02:22:29,912 --> 02:22:32,604
and, uh, are you getting depressed? Gator, I have a,
2187
02:22:33,169 --> 02:22:35,889
I have a product idea for you. All right, I'm listening.
2188
02:22:37,282 --> 02:22:40,342
Um, uh, illicit beauty filters for teenagers.
2189
02:22:43,030 --> 02:22:46,193
Illicit beauty filters? Yeah. Yeah.
2190
02:22:46,822 --> 02:22:50,773
So, so secondhand gray market, gray market,
2191
02:22:50,854 --> 02:22:55,067
uh, Contraband filters. I see.
2192
02:22:55,371 --> 02:22:57,598
And, and my, and my argument is that, um,
2193
02:22:58,752 --> 02:23:02,774
some criminal gang somewhere is gonna make a lot of money doing it.
2194
02:23:02,998 --> 02:23:06,154
Why not us? Well, I'm pretty sure that there's an API
2195
02:23:06,363 --> 02:23:09,615
to connect to Instagram and publish to Instagram, so you can just
2196
02:23:09,647 --> 02:23:12,403
circumvent their filter and do your own filter.
2197
02:23:13,156 --> 02:23:17,146
And the phones do the filters now too anyway, right? The, the— any
2198
02:23:17,178 --> 02:23:20,174
phone you're gonna get is gonna have their own filters that you can put on
2199
02:23:20,222 --> 02:23:23,496
them. So you're really just— for once, Facebook can't— like,
2200
02:23:23,737 --> 02:23:27,428
literally cannot compete with you. Uh, this is
2201
02:23:27,928 --> 02:23:31,279
going to be a renaissance in beauty filters. It's going to be more, more widespread
2202
02:23:31,779 --> 02:23:36,254
than ever. And I mean, look, it's an interesting trade-off, right? We get rid of
2203
02:23:36,754 --> 02:23:39,271
the, the Instagram filters so that children can't, uh,
2204
02:23:39,496 --> 02:23:43,235
distort the way they look, but as soon as, as soon as that,
2205
02:23:43,315 --> 02:23:46,832
uh, kid uses Instagram, they are going to find the naked naked—
2206
02:23:47,027 --> 02:23:50,652
the naked ads of the, the simulated sex bots who
2207
02:23:50,684 --> 02:23:55,146
are trying to— And
2208
02:23:55,646 --> 02:23:58,992
arguably those are not using filters, so you're fine. That is, that is arguable.
2209
02:23:59,105 --> 02:24:03,288
It is arguable. It is a reality filter. It is a reality
2210
02:24:03,788 --> 02:24:07,248
filter. Um, anyway, interesting,
2211
02:24:07,281 --> 02:24:10,590
interesting development. Um, all right, it's funny because I think that,
2212
02:24:10,655 --> 02:24:13,754
uh, China had done similar things with the social media
2213
02:24:13,819 --> 02:24:17,048
a while And it was like, oh my God, you know, look at China,
2214
02:24:17,064 --> 02:24:20,302
whatever. And apparently now it's like, yeah, fine, whatever,
2215
02:24:20,691 --> 02:24:23,541
you know, we're just gonna blame China again,
2216
02:24:24,091 --> 02:24:27,728
I guess. All right, Alex, um,
2217
02:24:27,922 --> 02:24:31,268
I'll, I'll put the square root of 2 denialism on pause
2218
02:24:31,381 --> 02:24:34,453
for the moment. Um, that's a, that's a topic near and dear to my heart,
2219
02:24:34,533 --> 02:24:37,815
but I'll, I'll, I'll move forward here and we'll,
2220
02:24:37,928 --> 02:24:40,126
we'll start— Dug into infinity denialism.
2221
02:24:42,603 --> 02:24:46,078
Well, that's— I mean, listen, as your resident square root
2222
02:24:46,578 --> 02:24:49,746
of 2 denialist, you know, you could get a nice
2223
02:24:50,246 --> 02:24:53,810
rant out of me on infinities as well. Okay.
2224
02:24:54,020 --> 02:24:57,914
We'll get there. I'll bring you Patterson one time and you
2225
02:24:58,414 --> 02:24:59,336
guys can have at it.
2226
02:25:02,885 --> 02:25:06,377
So we're moving towards traces of AI dystopia. I don't know if
2227
02:25:06,877 --> 02:25:09,997
this technically fits in the category or not, but Nvidia To acquire AI
2228
02:25:10,061 --> 02:25:12,662
model repository Hugging Face for $13 billion.
2229
02:25:14,123 --> 02:25:17,752
Uh, Nvidia is reportedly moving forward to acquire Hugging Face
2230
02:25:17,768 --> 02:25:21,028
for $12.9 billion. The acquisition would
2231
02:25:21,528 --> 02:25:25,636
help the hardware giant expand and fortify its deep integration with the wider AI industry.
2232
02:25:26,503 --> 02:25:30,533
Among other things, Hugging Face is a cloud repository for AI models similar in
2233
02:25:31,033 --> 02:25:34,226
some respects to what GitHub is for conventional computer
2234
02:25:34,726 --> 02:25:37,900
software. where developers and researchers search for models that meet certain
2235
02:25:37,964 --> 02:25:42,070
criteria, download and run them, and fine-tune them into different variants
2236
02:25:42,570 --> 02:25:46,224
that then get uploaded back to Hugging Face. The deal is not finalized,
2237
02:25:46,240 --> 02:25:49,576
so it could still run into trouble ahead, but both parties
2238
02:25:49,608 --> 02:25:53,362
appear to be actively pursuing it, according to a report from The Information,
2239
02:25:53,378 --> 02:25:57,740
which cites one person with knowledge on the matter. Another source confirmed to CNBC
2240
02:25:58,174 --> 02:26:01,927
that talks are underway. Hugging Face reportedly had other companies trying
2241
02:26:01,959 --> 02:26:05,534
to buy it too, including Salesforce. Nvidia had invested in
2242
02:26:05,566 --> 02:26:09,077
Hugging Face in the past, but it wasn't the only one. Google and Microsoft had
2243
02:26:09,577 --> 02:26:12,107
also invested. The company was founded in 2016.
2244
02:26:12,940 --> 02:26:16,755
Nvidia and several other tech companies have publicly supported the role of
2245
02:26:16,867 --> 02:26:20,683
open weight models as alternatives to the proprietary closed frontier models
2246
02:26:21,019 --> 02:26:24,450
trained by companies like OpenAI and Anthropic. Nvidia has even
2247
02:26:24,482 --> 02:26:27,335
released its own family of open models called Neutron.
2248
02:26:28,409 --> 02:26:31,695
By acquiring Hugging Face, Nvidia gains significant influence
2249
02:26:32,195 --> 02:26:35,366
in that ecosystem. Additionally, Nvidia previously had plans. For its
2250
02:26:35,446 --> 02:26:38,635
own cloud AI business but had trouble getting off the ground.
2251
02:26:39,228 --> 02:26:42,993
This could help get that engine running again. Hugging Face's business
2252
02:26:43,493 --> 02:26:46,518
is reportedly not yet profitable, but the fact that it has become the de facto
2253
02:26:47,018 --> 02:26:50,604
place to store, download, or work on open weight models makes it highly
2254
02:26:51,104 --> 02:26:54,082
strategically relevant at this moment, regardless of its current bottom line.
2255
02:26:55,235 --> 02:26:58,488
The acquisition is particularly important for Nvidia, as OpenAI,
2256
02:26:58,616 --> 02:27:02,414
Anthropic, and other frontier labs have begun investing in their own specialized hard hardware
2257
02:27:02,863 --> 02:27:06,677
on which to train and run those models, seeking vertical integration and freedom from
2258
02:27:06,693 --> 02:27:09,963
Nvidia's leverage. With Hugging Face in its portfolio,
2259
02:27:10,075 --> 02:27:13,265
Nvidia may take measures to encourage that side of the market to continue
2260
02:27:13,765 --> 02:27:17,657
relying on its hardware. Hugging Face is best known for hosting large language models,
2261
02:27:17,769 --> 02:27:21,616
but it's increasingly invested in the types of models used in robotics
2262
02:27:22,116 --> 02:27:25,414
and other physical AI applications. Nvidia is already one of the biggest
2263
02:27:25,462 --> 02:27:28,684
players in that space, so the acquisition may open opportunities there
2264
02:27:29,184 --> 02:27:32,520
in the future. Alex, your thoughts first. on, on the potential acquisition?
2265
02:27:34,980 --> 02:27:38,468
Um, I think that I'm actually kind of positive about it.
2266
02:27:38,727 --> 02:27:42,247
Um, so I think that NVIDIA has
2267
02:27:42,360 --> 02:27:45,901
realized something as the other, you know,
2268
02:27:46,062 --> 02:27:48,424
so, so basically this is in, in the context where,
2269
02:27:49,661 --> 02:27:53,099
um, OpenAI just revealed their own
2270
02:27:53,260 --> 02:27:56,494
chip with some impressive claims about its performance,
2271
02:27:57,319 --> 02:28:02,560
uh, while at the same time Elon has abandoned essentially some
2272
02:28:03,060 --> 02:28:06,270
plans to compete with NVIDIA and has gone with being forever with NVIDIA. There are
2273
02:28:06,770 --> 02:28:10,070
rumblings that Anthropic might want to design its own chips. Amazon is already doing it,
2274
02:28:10,570 --> 02:28:14,310
et cetera, et cetera. And I think NVIDIA, regardless of
2275
02:28:14,810 --> 02:28:17,970
whether the individual chips from the other companies are going to work out or they're
2276
02:28:18,470 --> 02:28:21,770
always going to be NVIDIA customers, I think has a very
2277
02:28:22,270 --> 02:28:25,930
interesting strategy of training its own models, as you read
2278
02:28:26,430 --> 02:28:30,308
in this story. which essentially
2279
02:28:30,487 --> 02:28:33,362
create a floor, right? The thing that NVIDIA makes,
2280
02:28:33,427 --> 02:28:36,846
NVIDIA gives for free to everybody. And then everybody,
2281
02:28:36,862 --> 02:28:39,676
of course, has to buy more NVIDIA compute in order to run those things.
2282
02:28:39,789 --> 02:28:43,120
So they're not doing it for free. They have a very good justification for it.
2283
02:28:43,589 --> 02:28:47,424
But it also creates an impetus for the labs to create better-than-that
2284
02:28:47,473 --> 02:28:50,858
models in order to differentiate themselves, for which they
2285
02:28:51,358 --> 02:28:55,573
need a lot of NVIDIA compute. compute. I don't know if you're seeing the picture
2286
02:28:56,073 --> 02:28:59,693
here, but it's like a win-win for NVIDIA. So in that regard,
2287
02:28:59,725 --> 02:29:03,507
they are very well incentivized to support the open source ecosystem.
2288
02:29:04,918 --> 02:29:08,332
And I guess by taking over Hugging Face officially, they are
2289
02:29:08,364 --> 02:29:11,554
becoming the GitHub essentially of AI models.
2290
02:29:12,548 --> 02:29:15,817
Which again, open source helps
2291
02:29:16,317 --> 02:29:20,530
their strategy both as itself and by encouraging
2292
02:29:20,547 --> 02:29:23,589
getting, uh, labs to keep accelerating their pace,
2293
02:29:24,431 --> 02:29:27,845
for which they need to pay Nvidia a bunch of money. So it's,
2294
02:29:27,926 --> 02:29:31,113
it's all, it's all a win for, uh, for Jensen Huang.
2295
02:29:32,520 --> 02:29:36,462
Does that, does that make sense? Okay, what do you think?
2296
02:29:36,849 --> 02:29:40,523
What do you— how do— how does the Hugging Face purchase
2297
02:29:40,990 --> 02:29:44,697
specifically potentially help Nvidia accomplish?
2298
02:29:46,252 --> 02:29:49,667
that goal? Because they could still
2299
02:29:50,179 --> 02:29:54,267
do their own models and push them to Hugging Face just
2300
02:29:54,767 --> 02:29:57,810
the same without the acquisition. So what does the acquisition buy them?
2301
02:29:58,291 --> 02:30:02,490
Enables them to, uh, control that node,
2302
02:30:02,522 --> 02:30:06,370
right? So consider what are the alternatives here. It could be that Anthropic
2303
02:30:06,870 --> 02:30:09,303
buys it, or OpenAI buys it, or Microsoft buys it,
2304
02:30:09,960 --> 02:30:14,116
or somebody else. And then that somebody else could could start doing things
2305
02:30:14,133 --> 02:30:17,526
like requiring special, you know, making it
2306
02:30:18,026 --> 02:30:21,604
difficult basically to publish open
2307
02:30:22,104 --> 02:30:24,438
source models by saying like, well, you have to be approved or you have to
2308
02:30:24,938 --> 02:30:27,920
have something like this or you can't download the open models. So, so,
2309
02:30:28,082 --> 02:30:32,084
so Nvidia could hypothetically have an Apple
2310
02:30:32,116 --> 02:30:35,867
Store-like policy.
2311
02:30:35,964 --> 02:30:41,190
Exactly. So taking it away from its open source Um,
2312
02:30:41,690 --> 02:30:43,970
sort of get all the way— Not Nvidia, but some other acquirer could do something
2313
02:30:44,470 --> 02:30:47,510
like that. Yes. Okay. So Nvidia's not going to do that. They're not going to
2314
02:30:48,010 --> 02:30:51,450
be like Apple and, and have a review process. No, I think Nvidia benefits
2315
02:30:51,950 --> 02:30:55,170
from wide, uh, availability and use of open models,
2316
02:30:55,670 --> 02:30:58,830
right? So by purchasing the repository for such models,
2317
02:30:59,330 --> 02:31:03,090
like they can defensively prevent, uh, that from being,
2318
02:31:03,590 --> 02:31:06,970
you know, made more frictionful by somebody who has more to gain by doing that.
2319
02:31:07,470 --> 02:31:09,570
Like, you know, Anthropic, when I say could Anthropic buy it, I mean, the purchases
2320
02:31:10,070 --> 02:31:14,127
happen happening for $13 billion, right? Anthropic is
2321
02:31:14,143 --> 02:31:18,056
rumored to be going public, uh, at a valuation of, uh, $2 trillion.
2322
02:31:18,954 --> 02:31:22,097
So it will be a rounding error for them to just sweep in and,
2323
02:31:22,161 --> 02:31:25,833
and purchase it. So that's why I'm saying like NVIDIA buying
2324
02:31:25,929 --> 02:31:28,992
it means that, you know, it, it,
2325
02:31:29,169 --> 02:31:32,825
it is a— it could be a defensive move to prevent
2326
02:31:33,325 --> 02:31:36,208
something like that. There's, there's times where companies like this are purchased and then we
2327
02:31:36,708 --> 02:31:39,886
don't really see much happen. happen with them that seems
2328
02:31:40,386 --> 02:31:44,098
to explain the value that the company got
2329
02:31:44,307 --> 02:31:48,330
out of that asset, right? Other than owning the
2330
02:31:48,830 --> 02:31:52,114
intellectual property of the site, right? Or the— like, they don't get
2331
02:31:52,275 --> 02:31:56,688
anything obviously from it. What do you think specifically
2332
02:31:57,768 --> 02:32:02,004
they would get out of it? I could see them possibly
2333
02:32:02,101 --> 02:32:05,481
integrating Hugging Face with their own software
2334
02:32:05,631 --> 02:32:09,475
ecosystem. system so that they provide a nice
2335
02:32:09,507 --> 02:32:13,297
pipeline for end consumers to— well, end consumers
2336
02:32:13,797 --> 02:32:16,845
slash actual people who use these in the— in a— in an environment to,
2337
02:32:17,166 --> 02:32:21,084
to pipeline, um, the usage of NVIDIA
2338
02:32:21,584 --> 02:32:26,078
product pipelines and have a nice, nicer plug-and-play
2339
02:32:26,608 --> 02:32:30,365
ecosystem to, to load the models instead of somewhat of
2340
02:32:30,413 --> 02:32:33,630
the, um, by-hand stuff
2341
02:32:34,130 --> 02:32:37,159
that you have to do when you're loading these things with Python or something like
2342
02:32:37,659 --> 02:32:40,543
that. You already have some platforms.
2343
02:32:42,728 --> 02:32:46,173
Yeah, yeah. Exactly. Exactly. I could see that happening.
2344
02:32:46,431 --> 02:32:50,681
Because the— there are platforms,
2345
02:32:51,036 --> 02:32:54,272
like, I don't know if you are in this environment at all,
2346
02:32:54,288 --> 02:32:57,726
but there's different platforms. Like one of them is called
2347
02:32:57,759 --> 02:33:01,146
ComfyUI. ComfyUI. And you basically, like, if you
2348
02:33:01,450 --> 02:33:05,088
get the ComfyUI desktop app, you basically can get a
2349
02:33:05,168 --> 02:33:09,157
system where you can visualize a workflow
2350
02:33:09,657 --> 02:33:14,044
pipeline that has Python in the background, operates as a sandboxed instance.
2351
02:33:14,749 --> 02:33:18,723
And you can download the models from Hugging Face
2352
02:33:18,996 --> 02:33:22,280
and plug and play them in a visualized workflow where
2353
02:33:22,360 --> 02:33:25,565
all of the input and outputs are able to be chained
2354
02:33:26,065 --> 02:33:29,932
visually. So that you can understand the models without having to deal
2355
02:33:29,980 --> 02:33:33,657
with reading Python documents and stuff like that,
2356
02:33:33,689 --> 02:33:37,157
and manuals, and download them all manually, install them all manually.
2357
02:33:37,221 --> 02:33:40,272
There's an ecosystem that helps people pipeline that. Now,
2358
02:33:40,288 --> 02:33:43,756
if Nvidia owned that type of process
2359
02:33:44,270 --> 02:33:48,621
and had already access to the full library to
2360
02:33:48,637 --> 02:33:52,970
provide customers of Nvidia products, to allow
2361
02:33:53,050 --> 02:33:56,367
them to get the maximum utility out of their product line,
2362
02:33:57,104 --> 02:33:59,812
I could see that benefiting NVIDIA directly.
2363
02:34:00,469 --> 02:34:04,187
Outside of that, the best I have as an idea
2364
02:34:04,315 --> 02:34:08,866
is maybe promoting their models or
2365
02:34:08,978 --> 02:34:12,328
the— or like— We'll see.
2366
02:34:12,584 --> 02:34:16,109
But my sense is that again, um, for NVIDIA, open source
2367
02:34:16,609 --> 02:34:19,413
in general is valuable. And they don't want to leave,
2368
02:34:19,798 --> 02:34:23,937
you know, a critical component in that ecosystem to chance
2369
02:34:24,771 --> 02:34:28,365
to somebody else to purchase. I will see, but yeah.
2370
02:34:28,926 --> 02:34:32,776
They will still promote the open source nature then, is your bet that
2371
02:34:33,276 --> 02:34:35,680
they are not going to close it off? I think so, yes. Okay. Look,
2372
02:34:35,825 --> 02:34:39,980
I mean, again, right? Nvidia. So so a key element of strategy
2373
02:34:40,557 --> 02:34:44,505
in technology at least is called commoditize your complement. I don't
2374
02:34:45,005 --> 02:34:48,682
know if you've heard the— I have not. Yeah, give me that. Um, so the
2375
02:34:49,182 --> 02:34:52,699
idea is like, for example, you're Google, right? What is, what is
2376
02:34:53,004 --> 02:34:56,426
a complement of the Google search engine is the browser, right? So what
2377
02:34:56,926 --> 02:34:59,575
does Google want to do? Wants to make the browser free, available to everybody.
2378
02:34:59,993 --> 02:35:03,382
Once you've got the browser, what's next? The mobile operating
2379
02:35:03,882 --> 02:35:07,977
system, right? You don't want anything that's between you and the people's
2380
02:35:08,477 --> 02:35:11,959
ability to use you, right? If Apple comes in and is putting conditions on
2381
02:35:11,975 --> 02:35:15,919
people being able to install Google's browser or using
2382
02:35:16,419 --> 02:35:19,030
the browser to get to Google, that's a problem for Google, right? Mm-hmm. That's a
2383
02:35:19,530 --> 02:35:23,199
bottleneck that they don't even want to have to compete for
2384
02:35:23,699 --> 02:35:27,448
the default position because now they're paying a ton of money to Apple to be
2385
02:35:27,948 --> 02:35:31,617
the default in their search box. So in that same sort of
2386
02:35:32,117 --> 02:35:33,845
logic, you know,
2387
02:35:34,262 --> 02:35:37,245
NVIDIA's product is GPUs.
2388
02:35:37,774 --> 02:35:40,871
GPUs run models. Right? So the models
2389
02:35:41,064 --> 02:35:45,239
are complement— a complement of the GPU. So what NVIDIA
2390
02:35:45,739 --> 02:35:49,221
wants is people to be putting all their money in buying GPUs. Sure. And therefore
2391
02:35:49,349 --> 02:35:53,155
it wants the models to be free, right? They don't— they want the models
2392
02:35:53,655 --> 02:35:55,756
to be something that is just— there's a lot of them, they're out there,
2393
02:35:55,836 --> 02:35:59,385
they're easy to get, there's no friction in getting that. It's a commodity,
2394
02:35:59,931 --> 02:36:03,543
right? So then you focus all your spending and
2395
02:36:04,043 --> 02:36:07,540
your focus Um, on, uh,
2396
02:36:08,170 --> 02:36:11,724
you know, purchasing more and more. So that's sort of, that's sort of where
2397
02:36:12,224 --> 02:36:16,488
I'm, I'm headed. But my, my question comes in like, there, there are other
2398
02:36:16,988 --> 02:36:20,137
GPUs in the world. So AMD,
2399
02:36:20,782 --> 02:36:24,424
while it doesn't even come close to having the, the competitive
2400
02:36:24,924 --> 02:36:28,078
footprint or a, or a competitive, strong competitive footprint,
2401
02:36:28,549 --> 02:36:32,525
is technically a competitor. And yes,
2402
02:36:32,542 --> 02:36:36,524
technically a competitor. And the one thing that Nvidia has
2403
02:36:36,877 --> 02:36:39,928
and has had above, um,
2404
02:36:40,313 --> 02:36:44,391
uh, AMD and other competitors is
2405
02:36:44,889 --> 02:36:48,807
the ecosystem. The CUDA platform, the CUDA programming
2406
02:36:48,903 --> 02:36:53,383
platform is what gave them a, a, a hold on—
2407
02:36:53,817 --> 02:36:57,882
Yep. The ecosystem. So you're
2408
02:36:57,996 --> 02:37:01,375
sort of indicating to me a belief, not explicit at
2409
02:37:01,875 --> 02:37:05,965
the moment, so that's why I'm articulating it, that Hugging Face would remain
2410
02:37:06,077 --> 02:37:09,540
neutral since it
2411
02:37:10,040 --> 02:37:14,107
would be open source and not necessarily restricted, so that you could run non-CUDA-based
2412
02:37:14,610 --> 02:37:18,225
things on any GPU, hypothetically, provided that
2413
02:37:18,419 --> 02:37:21,583
it was developed in a way that it could be run on it. Yeah,
2414
02:37:21,599 --> 02:37:24,542
I don't, I don't think they're gonna put their— I mean, I would be surprised
2415
02:37:25,042 --> 02:37:27,630
if they put their thumb on the scale that way. I think their, uh,
2416
02:37:27,905 --> 02:37:31,075
attitude is that they kind of own the GPU game for whatever reason.
2417
02:37:31,139 --> 02:37:33,663
We can get into why they have that, uh,
2418
02:37:34,456 --> 02:37:37,433
uh, belief and whether it's true or not. And I mean,
2419
02:37:37,772 --> 02:37:41,017
substantially true, right? I, I mostly say it's CUDA,
2420
02:37:41,097 --> 02:37:45,115
but yeah. Yeah, no, it is CUDA, but you know, what does CUDA like
2421
02:37:45,180 --> 02:37:48,555
mean? You know what, what do you mean? If it's a programming interface, why hasn't
2422
02:37:49,055 --> 02:37:52,053
there like, uh,
2423
02:37:53,321 --> 02:37:57,079
compatibility layer showing up, right? Yeah. Like, that's the typical answer
2424
02:37:57,496 --> 02:38:00,788
anyway. Yeah. Um, but for whatever reason, I think
2425
02:38:00,868 --> 02:38:04,497
NVIDIA is confident that they will continue to dominate the GPU
2426
02:38:04,997 --> 02:38:08,399
game. Yeah. Um, it is kind of surprising
2427
02:38:08,431 --> 02:38:11,192
that this is the situation we're in right now, but this is what the situation
2428
02:38:11,225 --> 02:38:15,333
we're in right now. Um, and, and From
2429
02:38:15,349 --> 02:38:18,192
that point of view, they are asking, you know,
2430
02:38:18,338 --> 02:38:21,763
what threatens— or
2431
02:38:21,957 --> 02:38:23,896
you don't have to see it as a threat, you can see it as an
2432
02:38:24,396 --> 02:38:27,818
opportunity. Like, how do we ensure that open models continue to
2433
02:38:28,173 --> 02:38:31,660
improve and spread and grow and be more used and whatever?
2434
02:38:32,400 --> 02:38:36,589
It's like, well, we should shepherd the company
2435
02:38:37,089 --> 02:38:39,756
that is the, you know, the community hub.
2436
02:38:40,230 --> 02:38:44,260
for that development and definitively
2437
02:38:44,760 --> 02:38:48,580
make sure that it doesn't fall into the wrong hands, right? Because that's
2438
02:38:49,080 --> 02:38:50,870
kind of, you know, GitHub is similar.
2439
02:38:52,860 --> 02:38:56,020
If you follow the GitHub acquisition by Microsoft, Microsoft is not going to stop
2440
02:38:56,520 --> 02:38:59,540
you from putting Linux software on GitHub, right?
2441
02:39:00,040 --> 02:39:02,430
Fair. To do that would be to destroy their GitHub acquisition,
2442
02:39:02,840 --> 02:39:07,360
right? Fair. You know, yeah. So it's
2443
02:39:07,860 --> 02:39:10,080
more like that. Twitter is actually another great example. And in that case,
2444
02:39:11,370 --> 02:39:15,038
You could say that maybe some fingers have been put on certain scales.
2445
02:39:19,570 --> 02:39:23,684
But overall, it's that the hubs have a difficulty
2446
02:39:24,184 --> 02:39:26,288
monetizing themselves to a point where,
2447
02:39:27,300 --> 02:39:31,952
you know, they are just gloriously, you know, strong businesses,
2448
02:39:32,451 --> 02:39:35,735
but their value lies in being who they are, which is
2449
02:39:36,331 --> 02:39:38,891
the place everybody goes to, even if you rename the thing X.
2450
02:39:40,161 --> 02:39:44,423
And fired 80% of the employees. If I were to argue that
2451
02:39:45,050 --> 02:39:48,588
Microsoft did not actually do much with GitHub as a property and
2452
02:39:49,088 --> 02:39:52,430
kind of left it for the most part alone, you would
2453
02:39:52,446 --> 02:39:56,387
say that the benefit is just in that ecosystem
2454
02:39:57,146 --> 02:40:01,573
existing and being maintained, having originally
2455
02:40:02,073 --> 02:40:05,190
an auxiliary value for the host company that has now acquired it. it.
2456
02:40:05,287 --> 02:40:08,997
And so therefore, just simply even just maintaining its existence
2457
02:40:09,497 --> 02:40:12,868
is the benefit to the company. Well, for GitHub, they have built an entire,
2458
02:40:13,238 --> 02:40:16,803
uh, developer ecosystem on top of it that runs on Azure, right? Like all this
2459
02:40:17,303 --> 02:40:20,161
GitHub CI stuff is an entire use case for Azure. Uh,
2460
02:40:20,466 --> 02:40:24,032
and then they built their agents. I will confess that I've never— I've,
2461
02:40:24,160 --> 02:40:28,143
I've never used GitHub for, for Azure anything.
2462
02:40:28,690 --> 02:40:32,352
So— Well, but you, you wouldn't know. That's what I'm saying. If you put
2463
02:40:32,852 --> 02:40:35,225
a GitHub CI JSON file, whatever it's called,
2464
02:40:35,979 --> 02:40:39,283
that that does builds, you know, and releases on GitHub.
2465
02:40:39,813 --> 02:40:43,262
You are using Azure like it or not. It's not called
2466
02:40:43,762 --> 02:40:47,609
Azure, right? Azure is a it's a monstrosity on the one end, but the GitHub
2467
02:40:48,109 --> 02:40:50,256
is the happy smiley face on top of the same compute.
2468
02:40:51,395 --> 02:40:55,437
Mm-hmm. So, you know, in a way they've been careful not to
2469
02:40:55,951 --> 02:40:59,721
blend the brands too much. They haven't called it Windows Server 2000 GitHub
2470
02:41:00,221 --> 02:41:03,490
Express. Pro. Uh, thank God. Um,
2471
02:41:03,990 --> 02:41:07,850
Copilot. But yeah, well, they have a Copilot on GitHub. Well, they, yes.
2472
02:41:08,350 --> 02:41:10,830
No, no, I meant, I was, I was just adding Copilot to your naming.
2473
02:41:10,930 --> 02:41:13,950
I was just adding Copilot to your naming because it was missing. It was missing
2474
02:41:14,450 --> 02:41:18,070
from the brand. Yes. Sorry. Uh, I'm, I'm, I'm not, I'm, I'm out
2475
02:41:18,570 --> 02:41:22,510
of date. Um, anyway, but you know, um, they, GitHub is
2476
02:41:23,010 --> 02:41:26,770
arguably one of the best acquisitions of all time to me. Like it is,
2477
02:41:27,270 --> 02:41:31,380
you know, Microsoft got a foothold. in a developer ecosystem that had
2478
02:41:31,880 --> 02:41:35,040
largely rejected them. Yeah. And they completely flipped it,
2479
02:41:35,540 --> 02:41:37,700
which is why then they went and bought npm, which is even crazier,
2480
02:41:38,140 --> 02:41:41,390
right? As a, as an acquisition. Why would Microsoft bought npm?
2481
02:41:41,400 --> 02:41:45,540
Oh, okay. Oh, a long time. A long time.
2482
02:41:45,880 --> 02:41:49,500
Okay. Okay. Yeah. Anyway, that's how much I hate
2483
02:41:50,000 --> 02:41:53,440
Node. So, you know, I, I, I definitely didn't know that. I just hate it.
2484
02:41:53,940 --> 02:41:57,400
Yep. Um, but it's another case of the community hub,
2485
02:41:57,900 --> 02:42:01,683
right? The thing where all the activity is cannot monetize itself to become a valuable
2486
02:42:01,715 --> 02:42:05,165
business that's worth running on its own. And you end up in, again, the same
2487
02:42:05,665 --> 02:42:09,224
situation, right? Where you have all the traffic and none of the money effectively.
2488
02:42:09,657 --> 02:42:13,588
Right. So you need a sugar daddy, for lack of a better term, to like,
2489
02:42:13,813 --> 02:42:17,246
you know, make sure that, you know, there's something valuable happening
2490
02:42:17,262 --> 02:42:19,765
here clearly. Right. Like this is not a company that should like die.
2491
02:42:20,648 --> 02:42:23,969
Right. But the way to monetize it
2492
02:42:24,081 --> 02:42:27,948
is not like a straight line where you like, you launch this like
2493
02:42:28,448 --> 02:42:31,785
companion product and boom, Boom. It's, it's, it's as being part
2494
02:42:32,285 --> 02:42:36,008
of a larger ecosystem, right? Okay. And Nvidia is— yeah, anyway,
2495
02:42:36,056 --> 02:42:39,395
sorry. All right, I'll buy it. I'll buy it. Boring, but no, I'll buy,
2496
02:42:39,412 --> 02:42:42,205
I'll buy that argument. I just, I was just trying to see if there was
2497
02:42:42,254 --> 02:42:46,011
a path to, um, expanding the value for Nvidia
2498
02:42:46,511 --> 02:42:50,314
past just ownership and maintenance of it. Right, right, right. Um, now as
2499
02:42:50,362 --> 02:42:54,376
a side question, Alex, you mentioned, um, the OpenAI chip
2500
02:42:57,165 --> 02:43:00,636
Jalapeño. Okay. And I want to state, a jalapeño is not a chip.
2501
02:43:02,070 --> 02:43:05,478
Jalapeño— You can make jalapeño chips,
2502
02:43:05,510 --> 02:43:06,825
maybe? Is a pepper.
2503
02:43:09,027 --> 02:43:12,963
Okay. What do you know about this chip and what is it— why
2504
02:43:12,979 --> 02:43:16,996
are they doing this? What is this benefit that they're supposedly trying
2505
02:43:17,012 --> 02:43:20,231
on this product line? What is this supposed to do for them? Okay. So,
2506
02:43:20,264 --> 02:43:23,151
I want to introduce you to my model of the situation,
2507
02:43:23,264 --> 02:43:26,316
quote unquote, like the whole AI AI game. Um,
2508
02:43:27,232 --> 02:43:29,721
because that, that's what I, you know,
2509
02:43:30,348 --> 02:43:33,833
I, I, I, I, I built that like, I don't know, 2 years ago or
2510
02:43:34,333 --> 02:43:37,189
something, and I keep coming back to it, right? Which is that you have,
2511
02:43:37,816 --> 02:43:40,578
um, money, and money buys, um,
2512
02:43:40,948 --> 02:43:43,662
energy. Energy goes into data centers, data centers,
2513
02:43:44,321 --> 02:43:48,159
um, train models, and models underpin applications
2514
02:43:48,256 --> 02:43:52,286
which make money, which feeds back into loop. Okay, if you
2515
02:43:52,786 --> 02:43:55,535
sort of draw that out as a self-reinforcing loop,
2516
02:43:56,347 --> 02:43:58,929
and, you know, of course, the money goes into the model, then the money goes
2517
02:43:58,961 --> 02:44:02,205
into, you know, there's like, it's not like a fully complete
2518
02:44:02,705 --> 02:44:04,606
picture, but it just gives you the general sense of what's happening.
2519
02:44:06,961 --> 02:44:10,554
And then you have companies in different parts of the game, right? So the data
2520
02:44:11,054 --> 02:44:13,921
centers and the chips inside them, NVIDIA is an enormous player,
2521
02:44:14,715 --> 02:44:18,522
right? And there's like certain NeoClouds and whatever, which some people say
2522
02:44:18,538 --> 02:44:21,792
are basically like 2 NVIDIAs in the trench. coat,
2523
02:44:21,824 --> 02:44:24,631
but there's a lot of activities in the data center section.
2524
02:44:25,438 --> 02:44:29,487
And then companies that are, quote unquote, the labs,
2525
02:44:29,955 --> 02:44:32,580
which are now becoming the biggest companies in the world in the so-called labs,
2526
02:44:34,329 --> 02:44:37,837
train models using compute that they
2527
02:44:38,227 --> 02:44:40,290
acquire through these data centers.
2528
02:44:42,342 --> 02:44:45,501
So that means that each one of these companies needs to pay a bunch
2529
02:44:46,001 --> 02:44:49,740
of money to NVIDIA, to Elon sometimes, anybody who can give them enough
2530
02:44:50,030 --> 02:44:53,607
GPUs to train
2531
02:44:53,703 --> 02:44:58,365
and serve their models. And then we— and
2532
02:44:58,865 --> 02:45:02,720
by we, it could be anybody from large companies to random people
2533
02:45:02,963 --> 02:45:06,507
like you and me— pay a bunch of money to these companies
2534
02:45:06,847 --> 02:45:09,339
to either use the models for ourselves or to build applications.
2535
02:45:11,638 --> 02:45:14,129
And that conversion then starts the cycle all over again.
2536
02:45:14,868 --> 02:45:18,724
But I guess what I'm saying is that if you see
2537
02:45:19,224 --> 02:45:22,899
that loop, though, as a strategic control sort of type of game.
2538
02:45:23,508 --> 02:45:27,049
Uh, if you're OpenAI and you're living in the, in the model sort of quadrant,
2539
02:45:27,995 --> 02:45:31,424
um, you are subject to the approval of
2540
02:45:31,937 --> 02:45:35,606
NVIDIA, say, right? That NVIDIA will, um, make available
2541
02:45:35,622 --> 02:45:38,987
for you enough chips fast enough. They won't prioritize anybody else.
2542
02:45:39,051 --> 02:45:42,160
They won't give anybody else special treatment, whatever, whatever. Um,
2543
02:45:42,977 --> 02:45:45,381
and, and if you're getting very, very, very big,
2544
02:45:46,102 --> 02:45:48,762
uh, big enough to be able to do things like make chip,
2545
02:45:49,787 --> 02:45:52,800
um, that can, that can act as a backup for your,
2546
02:45:53,793 --> 02:45:56,613
um, uh, strategy, let's say.
2547
02:45:56,998 --> 02:46:00,619
Even, even just to scare Nvidia, right? It doesn't even have to work.
2548
02:46:00,923 --> 02:46:04,465
And by work, I mean you produce it at like large enough volumes that you
2549
02:46:04,965 --> 02:46:08,070
can install it in data centers and just abandon Nvidia altogether. It could be that
2550
02:46:08,570 --> 02:46:12,364
you're using your own chips and Nvidia chips like as
2551
02:46:12,508 --> 02:46:14,848
fast as you can make them and as fast as they can make them.
2552
02:46:14,944 --> 02:46:16,370
You're like, we'll, we'll just buy everything.
2553
02:46:17,767 --> 02:46:21,700
But NVIDIA has to know then that you have
2554
02:46:22,200 --> 02:46:25,890
an option, right? So Google has the TPUs. Tesla tried
2555
02:46:26,390 --> 02:46:30,095
to do the Dojo computer, and that didn't work out very
2556
02:46:30,128 --> 02:46:33,707
well. But Amazon,
2557
02:46:33,723 --> 02:46:37,432
I believe, has Trainium. So the concept
2558
02:46:37,576 --> 02:46:41,830
of companies that are technically sort of model providers
2559
02:46:42,232 --> 02:46:45,586
or data center providers or whatever, whatever, making their own chips to cut Nvidia out
2560
02:46:46,086 --> 02:46:49,924
of the picture is not new. What's new is that this chip appears
2561
02:46:49,956 --> 02:46:52,643
to be particularly good and has been made extremely quickly.
2562
02:46:54,580 --> 02:46:58,559
So apparently the whole process from start
2563
02:46:59,059 --> 02:47:02,588
to now has been done in 9 months. And the chip itself,
2564
02:47:03,640 --> 02:47:06,911
according to numbers released by OpenAI, right? So it could, you know, this could be
2565
02:47:07,056 --> 02:47:11,570
wrong, does appear to be impressively good
2566
02:47:12,070 --> 02:47:15,495
at putting out lots and lots of tokens really, really quickly out of,
2567
02:47:15,752 --> 02:47:19,629
you know, with the chip itself and its energy usage being low
2568
02:47:20,129 --> 02:47:22,174
enough to give NVIDIA a real scare.
2569
02:47:24,458 --> 02:47:27,766
So if they're able to scale this, et cetera, et cetera, and not hit any
2570
02:47:27,799 --> 02:47:31,383
weird walls, right? And by walls, I mean supply chains
2571
02:47:31,883 --> 02:47:35,850
and everything else. This could be a serious threat for NVIDIA.
2572
02:47:36,550 --> 02:47:39,931
And to close off with one more point that I think you'll find very interesting,
2573
02:47:40,591 --> 02:47:43,972
what they claimed is that they used their latest model that has not
2574
02:47:44,472 --> 02:47:48,257
yet been released, the Astra model, to help this out, in order to port— to
2575
02:47:48,757 --> 02:47:52,255
port— no, no, to port models to it. So in effect, to sidestep
2576
02:47:52,755 --> 02:47:56,569
the CUDA advantage that NVIDIA has. So, you know,
2577
02:47:57,704 --> 02:48:01,902
that's definitely something to think about. Now, my thoughts
2578
02:48:02,402 --> 02:48:06,021
would be I followed
2579
02:48:06,521 --> 02:48:10,616
everything you said. My thoughts would be that to the point where they
2580
02:48:10,696 --> 02:48:13,878
actually gave Nvidia some concern,
2581
02:48:14,456 --> 02:48:18,151
one of the things that— yes, Nvidia is benefiting a lot
2582
02:48:18,489 --> 02:48:22,023
from this expansion in the need for compute,
2583
02:48:22,103 --> 02:48:25,204
both to train and to run these models.
2584
02:48:26,184 --> 02:48:29,061
But one of the things that Nvidia benefited from from,
2585
02:48:29,642 --> 02:48:33,486
and still benefits from, in my opinion, is the
2586
02:48:34,762 --> 02:48:38,995
multifaceted number of uses that their products actually
2587
02:48:39,173 --> 02:48:42,373
have. They are not limited to crypto,
2588
02:48:42,567 --> 02:48:45,136
AI, gaming, et cetera. Exactly.
2589
02:48:46,546 --> 02:48:51,236
Yes. Their demand is not specifically just
2590
02:48:51,918 --> 02:48:55,158
running an LLM. It is
2591
02:48:55,578 --> 02:49:00,805
everything that you could possibly imagine. In fact, the training LLMs
2592
02:49:01,305 --> 02:49:05,550
LLMs and running LLMs on GPUs was the novel usage.
2593
02:49:05,889 --> 02:49:10,206
Like, that was an invented novel use case of their systems.
2594
02:49:10,254 --> 02:49:13,810
It was not the intended usage. Like, doing anything on these machines
2595
02:49:14,232 --> 02:49:16,246
for those types of purposes was novel.
2596
02:49:18,200 --> 02:49:21,823
And designing something
2597
02:49:22,500 --> 02:49:26,623
that is exceptionally good at running your current specific
2598
02:49:26,832 --> 02:49:30,681
architecture is a potential limitation
2599
02:49:31,427 --> 02:49:36,402
on the versatility of your product.
2600
02:49:37,665 --> 02:49:40,756
So it might be great at running the models that
2601
02:49:41,256 --> 02:49:45,785
they currently have and could be available hypothetically
2602
02:49:46,285 --> 02:49:49,449
if they could produce them at scale to be useful as
2603
02:49:49,707 --> 02:49:52,935
a replacement for Nvidia product to run LLMs.
2604
02:49:54,414 --> 02:49:57,829
But that is not the only thing people might ever want to run on
2605
02:49:57,861 --> 02:50:01,916
them. And what happens if there is a breakthrough in the LLM technology that
2606
02:50:02,670 --> 02:50:06,148
makes that current set of chips now not
2607
02:50:06,196 --> 02:50:09,595
particularly useful because it's not— it's not the next
2608
02:50:10,095 --> 02:50:13,618
target of attack? You mean the jalapeño-style chips? Yeah, yeah, yeah. So, so that is
2609
02:50:13,650 --> 02:50:17,321
a concern. And, and actually, the part of the, the cheat code
2610
02:50:17,821 --> 02:50:21,249
that they use is that they custom tailored the chips to work,
2611
02:50:21,826 --> 02:50:24,839
you know, uh, close to the current technology of LLMs, right?
2612
02:50:25,016 --> 02:50:28,541
Which opens the question, what happens if, as you say,
2613
02:50:28,621 --> 02:50:33,461
let's say we crack continuous learning. So let's
2614
02:50:34,951 --> 02:50:39,181
put this down a little bit better because I think there's a really valuable thing
2615
02:50:39,454 --> 02:50:42,659
here to discuss, which is that basically you have a
2616
02:50:42,899 --> 02:50:45,383
trade-off between generality and specificity.
2617
02:50:46,088 --> 02:50:48,572
So when something is more general, it could be more expensive to make, but it
2618
02:50:49,072 --> 02:50:52,389
can do more things. And then somebody like OpenAI could say,
2619
02:50:52,437 --> 02:50:55,487
well, we only want these chips to do one thing, right? So we can kind
2620
02:50:55,552 --> 02:50:58,795
of cheat and do something that might not be a full-blown NVIDIA GPU,
2621
02:50:59,742 --> 02:51:02,713
but we'll make it specifically for our use case. And we know our use case
2622
02:51:03,213 --> 02:51:06,583
better than anybody else. So we know exactly what we need. And because, because GPUs,
2623
02:51:06,807 --> 02:51:09,938
because the GPUs are so universal, some of the
2624
02:51:10,438 --> 02:51:13,198
pipelines between how data gets transferred internally,
2625
02:51:13,326 --> 02:51:18,400
so, so data moves from A to B and there might be bottlenecks
2626
02:51:18,900 --> 02:51:22,681
that are unpleasant in use cases like running LLMs in them. And the bottlenecks can
2627
02:51:22,697 --> 02:51:26,242
just be circumvented if you're creating your own chip to specifically run that application.
2628
02:51:26,290 --> 02:51:30,525
That's right. That's right. But the problem is that all of that turns
2629
02:51:30,557 --> 02:51:34,695
over when you're talking about a fast-moving field,
2630
02:51:35,112 --> 02:51:38,865
for example, AI, right? And then the generality
2631
02:51:38,898 --> 02:51:41,721
becomes future-proofing, right? So you,
2632
02:51:42,346 --> 02:51:46,324
your chip design pipeline literally cannot
2633
02:51:46,824 --> 02:51:50,613
not, um, keep up with the changes in the architecture itself,
2634
02:51:51,094 --> 02:51:54,546
such that having built a general chip would have been a better idea to begin
2635
02:51:55,046 --> 02:51:58,865
with, if that makes sense. Mm-hmm. So, and then sort
2636
02:51:59,365 --> 02:52:03,296
of to the 4th layer of all of this is that OpenAI
2637
02:52:03,796 --> 02:52:06,074
comes in and says, you know what, yes, we hadn't built this chip to run
2638
02:52:06,106 --> 02:52:09,606
those models, but we use our general purpose coding
2639
02:52:09,638 --> 02:52:13,062
model to help bridge the gap with some models that we
2640
02:52:13,562 --> 02:52:16,992
never intended to run on this thing. So it's a shit show. Like, look,
2641
02:52:17,509 --> 02:52:21,990
from a sort of entrepreneurship, like nerdiness point
2642
02:52:22,490 --> 02:52:26,064
of view, this is the Super Bowl, right? Like you have
2643
02:52:26,564 --> 02:52:30,051
incredibly competent operators. Say what you want about Altman,
2644
02:52:30,347 --> 02:52:32,535
it doesn't matter. Like he is very, very good at what he does.
2645
02:52:34,689 --> 02:52:38,113
And you have Elon, you have Dario, you have Jensen Huang, you have— just
2646
02:52:38,613 --> 02:52:41,967
don't let them do your security. Uh, uh, run your
2647
02:52:42,016 --> 02:52:45,063
security in your company or do your networking security.
2648
02:52:46,185 --> 02:52:50,034
Nobody's perfect. Okay. Um, anyway, but long story short,
2649
02:52:50,178 --> 02:52:53,722
um, they just turn the Wi-Fi off on the, in the bottom taskbar and that's
2650
02:52:54,222 --> 02:52:57,972
your sandbox. Um, and if you don't
2651
02:52:58,052 --> 02:53:01,901
like it, you know what, use Anthropic. Um, oh wait,
2652
02:53:02,943 --> 02:53:06,023
they have the same problem too. Anyway, long story short,
2653
02:53:07,088 --> 02:53:10,483
Yeah, any sort of high technology company operator
2654
02:53:10,983 --> 02:53:13,830
that's worth their salt right now, so that would be not me,
2655
02:53:15,540 --> 02:53:18,832
is cutting their
2656
02:53:18,945 --> 02:53:22,737
teeth at finding a base somewhere along the elements that I was talking about,
2657
02:53:22,753 --> 02:53:25,894
which is the applications, the models, the data centers,
2658
02:53:25,943 --> 02:53:29,486
the energy game, or then when you
2659
02:53:29,986 --> 02:53:33,367
have a base on one or more of these, you try to expand your
2660
02:53:33,579 --> 02:53:37,621
foothold to start controlling more of these, right? So ultimately what
2661
02:53:37,702 --> 02:53:40,577
I suspect we're looking at is these companies,
2662
02:53:41,316 --> 02:53:45,202
um, becoming competing castles that have all 4,
2663
02:53:45,572 --> 02:53:48,977
uh, baked in potentially. So that could be fun.
2664
02:53:50,182 --> 02:53:53,587
Okay. You know, again, with each one of them having its own
2665
02:53:53,619 --> 02:53:57,281
advantages and not necessarily stopping from selling
2666
02:53:57,393 --> 02:54:01,312
to each other. But if you, like, for example, look at Elon's, uh,
2667
02:54:01,328 --> 02:54:04,799
consortium right now, right? They've got the energy down. They've got the data center down.
2668
02:54:04,864 --> 02:54:08,127
They're lacking the chips, which they are building their own chips,
2669
02:54:08,239 --> 02:54:10,731
but only for inference and only the cars. So that's a different story.
2670
02:54:11,560 --> 02:54:15,377
But they have the models and they have the applications, right? So Elon,
2671
02:54:15,603 --> 02:54:18,856
even though technically he's behind in that he doesn't have
2672
02:54:19,356 --> 02:54:23,090
a model that is competing with Fable, if you see
2673
02:54:23,590 --> 02:54:26,549
it in terms of like how broad his base is, he's probably the only one
2674
02:54:27,049 --> 02:54:29,399
that has credible entries in all 4,
2675
02:54:30,402 --> 02:54:32,926
uh, places. Mm-hmm. Okay.
2676
02:54:34,383 --> 02:54:37,958
All right. Well, let's, let's move forward here, Alex,
2677
02:54:38,168 --> 02:54:41,853
uh, to some, uh, traces of AI dystopia here.
2678
02:54:42,144 --> 02:54:45,404
Some real traces of AI dystopia. Yes.
2679
02:54:46,970 --> 02:54:50,778
Autobots, transform and roll out!
2680
02:54:51,988 --> 02:54:56,469
I had to throw in, uh, several Transformers-related
2681
02:54:56,969 --> 02:55:00,047
Easter eggs in this show, Alex, just because they have— they— we've lost my boy
2682
02:55:00,176 --> 02:55:02,678
Optimus Prime, and I'm not going to get over that anytime soon.
2683
02:55:03,448 --> 02:55:07,427
All right. Elon Musk's xAI
2684
02:55:07,700 --> 02:55:10,170
used child porn to train Grok models,
2685
02:55:10,282 --> 02:55:12,448
lawsuit says. This is from Ars Technica.
2686
02:55:12,994 --> 02:55:16,587
xAI has now been accused of training Grok on child sexual
2687
02:55:16,683 --> 02:55:20,213
abuse material as regulators and courts continue to probe
2688
02:55:20,713 --> 02:55:23,612
how far the problem goes. And some Grok users have been arrested.
2689
02:55:24,477 --> 02:55:27,971
In a complaint filed on Wednesday, a plaintiff known as Jane Doe explained
2690
02:55:28,308 --> 02:55:32,395
that she was preschool age in the early 2000s when adult men repeatedly raped
2691
02:55:32,895 --> 02:55:36,787
her to create CSAM to sell to pedophiles online.
2692
02:55:37,155 --> 02:55:40,730
Since then, Doe's images have been hashed by groups like the National Center
2693
02:55:41,230 --> 02:55:44,929
for Missing and Exploited Children and the Canadian Center for Child Protection. For her
2694
02:55:45,429 --> 02:55:48,920
safety, Doe has opted to receive alerts from the United States Department
2695
02:55:49,420 --> 02:55:53,167
of Justice Victim Notification system anytime she may be a victim in
2696
02:55:53,667 --> 02:55:57,269
a new criminal investigation. Although she has received countless alerts,
2697
02:55:57,333 --> 02:56:00,874
she was shocked when the CCCP notified her that
2698
02:56:01,374 --> 02:56:04,735
it had identified AI generate an AI generated CSAM
2699
02:56:05,168 --> 02:56:09,014
on XAI that depicted her. This retraumatized
2700
02:56:09,350 --> 02:56:13,484
Doe, whose complaint alleged that messages were found online on online
2701
02:56:13,516 --> 02:56:17,041
forums between offenders chatting about creating AI generated
2702
02:56:17,541 --> 02:56:21,175
CSAM. Of plaintiff and other similarly situated known legacy victims
2703
02:56:21,207 --> 02:56:24,525
of CSAM. Now, Doe fears that XAI has not
2704
02:56:24,573 --> 02:56:28,066
only made it easier to make more volatile images of the most distressing
2705
02:56:28,098 --> 02:56:31,223
time in her life, but also that XAI allegedly has stored
2706
02:56:31,723 --> 02:56:35,549
the images that Grok generates and uses those outputs to further train Grok.
2707
02:56:36,190 --> 02:56:39,492
Because of this, she believes that Grok has been trained on both the initial set
2708
02:56:39,992 --> 02:56:43,049
of images that have haunted her for more than 20 years and the more recent
2709
02:56:43,145 --> 02:56:47,576
AI. Generated ones. This is the first case to accuse xAI
2710
02:56:48,076 --> 02:56:51,815
of training on CSAM, and the complaint does not go into great
2711
02:56:51,831 --> 02:56:55,573
detail on that claim. Previously, Ars reported on a controversial
2712
02:56:55,605 --> 02:56:59,506
dataset that was later scrubbed after researchers found CSAM in the training data,
2713
02:56:59,779 --> 02:57:03,553
but there's no indication xAI trained on that data. In a
2714
02:57:04,053 --> 02:57:07,824
press release from lawyers representing Doe, it explained that Doe's images were included in
2715
02:57:08,324 --> 02:57:12,159
a CSAM hash list maintained by NCMEC, and that
2716
02:57:12,659 --> 02:57:16,559
same material allegedly was part of the dataset xAI used to build Grok's image and
2717
02:57:16,623 --> 02:57:20,460
video generation capabilities. The complaint similarly only
2718
02:57:20,492 --> 02:57:23,672
alleged that CSAM depicting plaintiff with its longstanding
2719
02:57:23,704 --> 02:57:27,589
well-known hash values has been used as part of the dataset used by xAI.
2720
02:57:28,521 --> 02:57:32,149
More detail shared on the claims alleging that Grok trains
2721
02:57:32,198 --> 02:57:36,043
on AI CSAM, however, Because Grok's terms
2722
02:57:36,075 --> 02:57:39,754
treat public ex posts and Grok's own outputs as
2723
02:57:39,851 --> 02:57:43,433
training data by default, publicly posting an image does
2724
02:57:43,933 --> 02:57:47,587
not just expose it to viewers, but also feeds it directly into the pipeline
2725
02:57:47,619 --> 02:57:51,403
xAI uses to train and improve its model, thereby generate further
2726
02:57:51,484 --> 02:57:54,761
images, Doe's complaint says. I'm really curious about that aspect, not,
2727
02:57:54,874 --> 02:57:59,028
not, you know, outside of the sadness
2728
02:57:59,302 --> 02:58:01,945
of the, the subject nature,
2729
02:58:02,943 --> 02:58:06,563
but don't you wind up with degradation if
2730
02:58:07,063 --> 02:58:11,107
you wind up using the outputs back as training inputs? I'd have to imagine
2731
02:58:11,527 --> 02:58:15,486
that you wind up with some— Not really, but I
2732
02:58:15,986 --> 02:58:19,591
don't— I'm really confused as to what exactly happened.
2733
02:58:20,254 --> 02:58:23,733
Do you understand what they're
2734
02:58:23,830 --> 02:58:26,888
talking about? So the loss— I'll continue and see if it— we'll
2735
02:58:27,388 --> 02:58:30,767
see if it answers any So the lawsuit further notes that while xAI filters
2736
02:58:31,267 --> 02:58:34,925
out violent content in Grok outputs to exclude from training data, AI's terms
2737
02:58:34,941 --> 02:58:38,859
notably do not specify if CSAM, non-consensual intimate
2738
02:58:38,955 --> 02:58:42,793
imagery, or non-safe work material are excluded categories. It seems
2739
02:58:42,809 --> 02:58:46,582
to follow then that because full removal of the training examples' influence
2740
02:58:47,082 --> 02:58:50,356
from an already trained model is technically difficult and not something that xAI has publicly
2741
02:58:50,856 --> 02:58:54,752
claimed to have done, any CSAM ingested in the training before takedown likely
2742
02:58:54,784 --> 02:58:58,277
continued to shape the model's outputs even after the original images
2743
02:58:58,293 --> 02:59:01,354
were removed from public view. In her proposed class action— Let
2744
02:59:01,854 --> 02:59:05,520
me just— Go ahead. So, sorry. So let me just understand
2745
02:59:06,020 --> 02:59:08,661
the claim and maybe you can correct me, or maybe we'll see if they clarify
2746
02:59:08,725 --> 02:59:11,994
further. I think what she's saying is somebody posted the photo of
2747
02:59:12,494 --> 02:59:15,952
her on Twitter at some point, and since Grok
2748
02:59:16,112 --> 02:59:19,670
uses Twitter data in general to train
2749
02:59:20,441 --> 02:59:26,642
And since they don't clarify whether they remove
2750
02:59:26,756 --> 02:59:30,463
material that is, you know, severely objectionable,
2751
02:59:30,708 --> 02:59:34,542
like the one she's talking about, therefore we assume that this
2752
02:59:35,042 --> 02:59:38,752
was used to train Grok. I think that's
2753
02:59:39,365 --> 02:59:42,707
the— So rereading the paragraph here that, that contains
2754
02:59:43,207 --> 02:59:46,468
somewhat of the identification, this is the first case to accuse xAI of
2755
02:59:46,520 --> 02:59:50,165
training on CC. CSAM, and the complaint does not go into
2756
02:59:50,181 --> 02:59:53,628
great detail on the claim. Previously, ours reported
2757
02:59:54,128 --> 02:59:57,635
on a controversial dataset that was later scrubbed after researchers found CSAM in
2758
02:59:58,135 --> 03:00:01,418
the training data, but there's no indication xAI trained on that data.
2759
03:00:02,044 --> 03:00:05,282
In a press release from lawyers representing Doe, it was
2760
03:00:05,330 --> 03:00:08,808
explained that Doe's images were included in a CSAM hash
2761
03:00:08,856 --> 03:00:12,992
list and that the same material on the hash list
2762
03:00:13,265 --> 03:00:16,855
was part of the dataset xAI used to build Grok's image and video
2763
03:00:17,192 --> 03:00:19,857
generating capabilities. So I don't know how they identified that,
2764
03:00:20,743 --> 03:00:24,398
if they have some— if someone has— if that— that the,
2765
03:00:25,010 --> 03:00:28,409
the dataset it uses is publicly known or known to specific individuals
2766
03:00:28,909 --> 03:00:31,954
or not. But they're claiming that the hash list was in that same dataset.
2767
03:00:36,380 --> 03:00:40,189
The hash list itself is not— is not itself
2768
03:00:41,400 --> 03:00:45,128
CSAM, right? The— well, the hash list
2769
03:00:45,628 --> 03:00:48,706
would not be the, the material itself. No, it would be some sort of signature
2770
03:00:48,723 --> 03:00:52,654
of the material. Right, so that
2771
03:00:53,154 --> 03:00:57,117
doesn't make any sense. But I think from other paragraphs, what I'm gathering
2772
03:00:57,149 --> 03:01:00,954
is that the hash list was used to notify her of her
2773
03:01:02,069 --> 03:01:06,047
pictures appearing on X. And I think
2774
03:01:06,095 --> 03:01:09,717
what she's sort of— what the article is
2775
03:01:09,749 --> 03:01:13,662
sort of saying is that because it appeared on X
2776
03:01:13,856 --> 03:01:17,837
and Grok uses X as training data, Therefore— well,
2777
03:01:17,950 --> 03:01:21,946
the original thing that tipped her off was a generated image that
2778
03:01:22,156 --> 03:01:25,391
was very similar to her likeness that
2779
03:01:25,891 --> 03:01:29,362
got flagged as being her. Right. Okay.
2780
03:01:29,427 --> 03:01:32,847
But that cannot be a hash list, right? Unless the hash is very liberal,
2781
03:01:33,057 --> 03:01:36,090
as in it's kind of fingerprint in a very
2782
03:01:36,590 --> 03:01:39,922
sort of open-ended kind of way. No, no.
2783
03:01:40,019 --> 03:01:43,884
They're saying the hash, according to this, and I don't know how they specifically
2784
03:01:43,949 --> 03:01:48,151
confirmed it, they're saying that the the original source
2785
03:01:48,200 --> 03:01:51,479
images that are in the hash
2786
03:01:51,609 --> 03:01:55,491
list were identified to have been in
2787
03:01:55,507 --> 03:01:58,467
the training set. I don't know how they made that determination,
2788
03:01:59,368 --> 03:02:03,792
but she was originally tipped off because of an AI-generated
2789
03:02:03,873 --> 03:02:07,609
image that was tagged by a group that
2790
03:02:08,109 --> 03:02:11,829
does monitoring as having been
2791
03:02:11,926 --> 03:02:14,911
her or was similar enough to being her. Right. No, I,
2792
03:02:15,218 --> 03:02:18,004
I, I get it. It's just that then we start to wonder whether,
2793
03:02:18,863 --> 03:02:22,604
um, you know, how good the algorithms of that
2794
03:02:23,104 --> 03:02:26,662
group are. Um, well, the— that, that does—
2795
03:02:26,855 --> 03:02:30,559
that, that will, I guess, be somewhat fought out in the
2796
03:02:31,059 --> 03:02:34,021
court, uh, whether— unless they settle, in which case it won't be fought, it'll just
2797
03:02:34,521 --> 03:02:37,576
be settled. Um, yeah, but, you know,
2798
03:02:37,772 --> 03:02:38,947
there's, there's, uh,
2799
03:02:41,647 --> 03:02:44,915
Certainly lots of difficulties even
2800
03:02:44,963 --> 03:02:50,997
in, like, okay, let's agree for a moment that no
2801
03:02:51,302 --> 03:02:55,399
material originating as child abuse material should
2802
03:02:55,899 --> 03:02:59,110
ever be used in a training set for generative images. Let's agree
2803
03:02:59,610 --> 03:03:03,242
to that as a principle, whether or not that is necessarily the principle
2804
03:03:03,742 --> 03:03:06,031
to agree on. Let's agree to that for a moment. Well, they're not even Not
2805
03:03:06,531 --> 03:03:10,956
even just that, but also I don't know why they would want
2806
03:03:11,456 --> 03:03:14,965
that. But in any case, now, like,
2807
03:03:15,271 --> 03:03:19,104
if you at all allow versions of
2808
03:03:19,491 --> 03:03:22,748
nudity, the vectors
2809
03:03:22,861 --> 03:03:26,338
that are involved in identifying the shapes
2810
03:03:26,838 --> 03:03:30,928
that produce that imagery could probably be combined to produce
2811
03:03:30,979 --> 03:03:33,377
something like that no matter what,
2812
03:03:34,742 --> 03:03:38,082
right? Like the ability to produce imagery. If you're going
2813
03:03:38,098 --> 03:03:40,330
to have naked women and young children,
2814
03:03:41,229 --> 03:03:43,990
right? Right. It's, yeah.
2815
03:03:44,279 --> 03:03:47,908
Even if you don't have the combination, the way
2816
03:03:48,408 --> 03:03:51,810
these models work, they could produce the combination. Correct. I mean, listen, you can have
2817
03:03:53,190 --> 03:03:56,771
panda bears wearing suits playing tennis, right?
2818
03:03:57,285 --> 03:04:00,762
Those are just concepts, right? And so the degree
2819
03:04:01,262 --> 03:04:05,236
to which these image generators
2820
03:04:05,414 --> 03:04:08,975
are trained on concepts, the concepts can be combined. And so that means that
2821
03:04:09,475 --> 03:04:12,892
you won't necessarily be able to stop it from generating that type
2822
03:04:13,392 --> 03:04:16,717
of imagery at the concept level. You have to have it
2823
03:04:16,895 --> 03:04:20,912
on the pipeline adhering
2824
03:04:21,025 --> 03:04:24,177
to requests level, because you can't stop the image
2825
03:04:24,275 --> 03:04:28,435
generator from combining vectors describing image
2826
03:04:28,935 --> 03:04:34,050
shapes and somehow making
2827
03:04:34,082 --> 03:04:38,012
it not generate those specific ones. You have to have a layer that stops
2828
03:04:38,512 --> 03:04:42,206
that in advance, like something that takes the request as text and goes,
2829
03:04:42,303 --> 03:04:45,721
no, this violates my policy. Because I don't— I just don't see how you can
2830
03:04:46,221 --> 03:04:47,308
do that at the image generation level.
2831
03:04:50,111 --> 03:04:52,857
That's— yeah, I don't, I don't see— you have to,
2832
03:04:53,402 --> 03:04:56,598
because the— these general diffusion image generators,
2833
03:04:56,710 --> 03:05:00,371
which there's several other things in the pipeline now, but the general image diffusion
2834
03:05:00,483 --> 03:05:04,224
layer is take this, this random noise
2835
03:05:05,204 --> 03:05:08,961
and un— use an unnoising algorithm, a denoising
2836
03:05:09,138 --> 03:05:11,835
algorithm on it, and start to, uh,
2837
03:05:12,445 --> 03:05:15,801
classify that image. And based off of the prompt,
2838
03:05:15,978 --> 03:05:19,413
pick the one that is closest to what the user asked
2839
03:05:19,913 --> 03:05:23,522
for. Then you add a little bit more noise and then denoise
2840
03:05:24,022 --> 03:05:28,048
it again and continue that iteration. And then do a couple
2841
03:05:28,096 --> 03:05:31,675
extra things, but basically keep denoising it, removing noise
2842
03:05:31,691 --> 03:05:35,222
from it, until you get an image that matches what the user asked
2843
03:05:35,722 --> 03:05:39,218
for as a classification. And there's some extra little tricks that they
2844
03:05:39,315 --> 03:05:42,798
do, but that process is what winds up generating that
2845
03:05:43,298 --> 03:05:46,390
image. So that process doesn't know what
2846
03:05:46,890 --> 03:05:50,046
a naked child is. It has no concept of that.
2847
03:05:50,222 --> 03:05:53,622
It's getting text that then turns it into
2848
03:05:54,122 --> 03:05:57,791
vectors of representation. Those vectors of representation get turned
2849
03:05:58,291 --> 03:06:00,934
into a denoising algorithm to turn into matching and say, hey,
2850
03:06:01,929 --> 03:06:05,280
enhance this vector, don't enhance this vector, decrease the value on
2851
03:06:05,296 --> 03:06:08,952
this vector. And then you get an image that matches the classification
2852
03:06:08,984 --> 03:06:12,404
that you asked for to the best of the generator's
2853
03:06:12,904 --> 03:06:16,037
ability. I don't, I don't know how you block that, that content
2854
03:06:16,409 --> 03:06:19,575
outside of, uh, the, the pre-filter.
2855
03:06:20,787 --> 03:06:23,179
Look, we're requesting it. That, all that aside,
2856
03:06:24,306 --> 03:06:27,840
um, and without sort of getting into what this one has gone through, which is
2857
03:06:28,033 --> 03:06:30,776
definitely seems, uh, pretty, uh,
2858
03:06:31,470 --> 03:06:36,037
awful, um, I'm, I'm I'm
2859
03:06:36,537 --> 03:06:39,936
getting very weird vibes from this person saying, I, I was traumatized by seeing the
2860
03:06:39,984 --> 03:06:43,126
images that I have specifically registered myself
2861
03:06:43,626 --> 03:06:46,881
to be on a notification list to see. Uh, well, the new, the new
2862
03:06:46,995 --> 03:06:50,657
generation is, is what she said she was re-traumatized
2863
03:06:51,157 --> 03:06:53,672
by. Not— sure, not, not an, uh,
2864
03:06:54,588 --> 03:06:57,850
not a same old same old. The generation is now
2865
03:06:59,344 --> 03:07:03,858
people taking— hypothetically, people taking her likeness And
2866
03:07:04,358 --> 03:07:06,751
putting her in new scenarios. Oh,
2867
03:07:07,152 --> 03:07:10,607
yes, to the degree that we agree that it is indeed her likeness
2868
03:07:11,107 --> 03:07:14,222
and whatever. But all I'm saying is that, like,
2869
03:07:14,785 --> 03:07:18,480
there does seem to be, you know, in this particular lawsuit,
2870
03:07:18,721 --> 03:07:22,031
like, some attempt to
2871
03:07:23,526 --> 03:07:26,739
make a case that will, you know, become a class action and
2872
03:07:27,239 --> 03:07:30,880
make somebody rich. Um, because I,
2873
03:07:31,394 --> 03:07:34,859
again, the whole, like, there's no motive here. Like, why would you
2874
03:07:35,359 --> 03:07:38,853
want to train on these images anyway and, and expose yourself to a risk like,
2875
03:07:39,029 --> 03:07:42,510
well, this lawsuit? Um, let's say it's a,
2876
03:07:42,719 --> 03:07:46,168
let's say that the, the issue is not necessarily that you wanted to,
2877
03:07:46,232 --> 03:07:49,937
but that you, um, let's, let's take the Hanlon's
2878
03:07:50,437 --> 03:07:53,322
razor on this one. Let's, let's take the Hanlon's razor on this one. No,
2879
03:07:53,386 --> 03:07:56,436
no, it could be that the, the data was somewhere, it got got hoovered up
2880
03:07:56,821 --> 03:08:00,067
and ended up in the training set. Yeah. Right?
2881
03:08:00,838 --> 03:08:04,935
Yeah. Well, and in that case, like, I get it, but like,
2882
03:08:04,951 --> 03:08:06,140
at the same time, you know,
2883
03:08:09,273 --> 03:08:13,659
yeah, I don't know. I tend to be skeptical that
2884
03:08:13,771 --> 03:08:16,771
the story being presented, I guess. And if not, you know, whatever,
2885
03:08:16,916 --> 03:08:20,794
they should pay something and scrub their training set and move
2886
03:08:21,294 --> 03:08:24,736
on. All right, next story on
2887
03:08:24,784 --> 03:08:27,236
the block here, Alex. Next story on the block.
2888
03:08:28,614 --> 03:08:32,749
This evening, that ICE is set to deploy a new tool in its deportation
2889
03:08:32,781 --> 03:08:36,676
campaign, a fleet of robot dogs to be used in operations
2890
03:08:36,948 --> 03:08:40,538
that might be unsafe for human agents. Here's ABC's Chief Justice
2891
03:08:41,038 --> 03:08:44,561
Correspondent Pierre Thomas. Tonight, as the Trump administration
2892
03:08:45,061 --> 03:08:48,888
accelerates its massive deportation campaign, ICE is now putting in
2893
03:08:49,388 --> 03:08:53,824
a $2 million order for robot Dogs to track and apprehend undocumented
2894
03:08:53,872 --> 03:08:57,750
migrants. The robots can open doors and are equipped
2895
03:08:57,782 --> 03:09:00,875
with cameras that can see 360 degrees.
2896
03:09:01,067 --> 03:09:04,801
They're increasingly being deployed by law enforcement domestically and abroad.
2897
03:09:05,298 --> 03:09:09,464
A Secret Service robot dog seen here patrolling the grounds at Mar-a-Lago,
2898
03:09:10,121 --> 03:09:13,326
and the Secret Service recently using them at a NATO conference.
2899
03:09:13,550 --> 03:09:17,060
Work with— They are symbolic of ICE's rapid expansion. The agency
2900
03:09:17,092 --> 03:09:20,062
Recently approving the purchase of so-called shock gloves,
2901
03:09:20,303 --> 03:09:24,044
which can briefly incapacitate a person delivering up to 380
2902
03:09:24,157 --> 03:09:27,657
volts. Give me your hands, give me your hands.
2903
03:09:27,770 --> 03:09:31,864
The president's deportation efforts ramping up. Nearly 47,000
2904
03:09:32,364 --> 03:09:36,601
arrests by ICE in July, the most since October 2022. The majority
2905
03:09:37,101 --> 03:09:40,615
of those arrested without criminal records. And with one other piece
2906
03:09:41,115 --> 03:09:44,313
of news to report tonight, DHS tells us that controversial
2907
03:09:44,813 --> 03:09:48,355
right-wing provocateur Milo Yiannopoulos, a British citizen, is now
2908
03:09:48,855 --> 03:09:52,594
in ICE detention after overstaying his visa. The feds releasing this
2909
03:09:52,772 --> 03:09:55,732
picture of Yiannopoulos after his arrest in New Orleans.
2910
03:09:56,249 --> 03:10:00,179
A former high-profile Trump supporter, he is now awaiting deportation.
2911
03:10:00,890 --> 03:10:03,221
Whit. Pierre Thomas with those developments. Thank you.
2912
03:10:05,487 --> 03:10:09,104
Now, just to be clear, Milo Yiannopoulos is not part
2913
03:10:09,604 --> 03:10:14,418
of the AI dystopia. He was just part of the normal dystopia that we all
2914
03:10:14,918 --> 03:10:18,198
lived in previous— But he is in the training data. He is unfortunately in the
2915
03:10:18,698 --> 03:10:21,487
training data. Grox, most specifically.
2916
03:10:22,065 --> 03:10:23,962
Um, now, uh,
2917
03:10:26,325 --> 03:10:30,011
uh, Alex, the, the, the dogs, the robot dogs, I believe,
2918
03:10:30,269 --> 03:10:34,027
um, or, or the, uh, the K-1001
2919
03:10:34,140 --> 03:10:36,707
units, uh, as I, as I put them,
2920
03:10:37,452 --> 03:10:41,564
um, those— I don't like the robot dogs because when,
2921
03:10:41,661 --> 03:10:45,230
when Boston Robotics was first coming out and, and showing us how how
2922
03:10:45,439 --> 03:10:49,061
awesome their things were. They were kicking their robot dogs, and I had
2923
03:10:49,142 --> 03:10:52,748
a visual— I had a visceral reaction to them kicking the
2924
03:10:52,845 --> 03:10:56,502
robot dog. I felt bad for the robot dog, and that's not good.
2925
03:10:56,631 --> 03:11:00,517
I need to be able to kill these things without remorse if
2926
03:11:01,017 --> 03:11:04,201
necessary, and I can't— I can't have that visceral reaction.
2927
03:11:04,298 --> 03:11:08,621
I can't have that connection. So I don't like the robot dogs. I don't.
2928
03:11:09,396 --> 03:11:11,219
Yeah, yeah. And all the kicking is also in the training.
2929
03:11:12,900 --> 03:11:17,140
So, uh, they know,
2930
03:11:17,626 --> 03:11:20,749
they know. You know, last summer. And speaking in
2931
03:11:20,782 --> 03:11:24,652
between of robot dogs and Boston
2932
03:11:24,685 --> 03:11:25,792
Robotics and running,
2933
03:11:27,900 --> 03:11:32,048
world-first humanoid robot games— world-first
2934
03:11:32,548 --> 03:11:35,486
humanoid robot games show runners breaking records bursting into flames.
2935
03:11:36,340 --> 03:11:39,914
Viral videos of world humanoid robot game World Humanoid
2936
03:11:39,930 --> 03:11:43,623
Robot Games shows sprinting robots beating the human 100-meter record
2937
03:11:44,123 --> 03:11:48,022
held by Usain Bolt, but also features running robots crashing into barriers and falling
2938
03:11:48,054 --> 03:11:51,571
down because they cannot readily stop running. A few fallen robots
2939
03:11:51,587 --> 03:11:54,959
even break the waist at the waist in a shower of sparks or catch on
2940
03:11:55,459 --> 03:11:59,278
fire. That demonstration of both robotic prowess and limitations came from the second
2941
03:11:59,778 --> 03:12:03,645
edition of the World Humanoid Robot Games hosted in Beijing from August 22
2942
03:12:03,790 --> 03:12:07,045
through 26. Like the inaugural event held in 2025,
2943
03:12:07,077 --> 03:12:10,716
this year's event challenged humanoid robots to run fast, perform kung fu
2944
03:12:10,844 --> 03:12:14,259
or dance moves, fight one another in kickboxing, and play soccer
2945
03:12:14,759 --> 03:12:18,507
or table tennis. This year's games also tested robots in more humdrum practical
2946
03:12:18,523 --> 03:12:21,473
scenarios like washing and hanging laundry,
2947
03:12:22,066 --> 03:12:25,529
where they appeared to lag behind typical human working speeds.
2948
03:12:26,218 --> 03:12:29,809
That suggests that the event organizers, including Chinese state media and the Beijing city
2949
03:12:29,890 --> 03:12:33,703
government are keen to showcase the commercial potential of humanoid robots at
2950
03:12:34,203 --> 03:12:37,580
a time when China's robotics industry is rapidly developing and testing such
2951
03:12:37,612 --> 03:12:41,666
robots. The fact that humanoid robots went from lagging behind human record
2952
03:12:41,682 --> 03:12:45,623
holders last year to surpassing human world records in the 100-meter
2953
03:12:46,123 --> 03:12:50,093
sprint and standing high jump events this year suggests significant progress in robotic
2954
03:12:50,414 --> 03:12:52,993
whole-body control for accomplishing specific tasks.
2955
03:12:53,714 --> 03:12:57,223
Deepam Patel, a PhD candidate in computer science at
2956
03:12:57,723 --> 03:13:01,536
Purdue University and a research assistant at the U.S. Army DEVCOM Army Research Lab,
2957
03:13:01,552 --> 03:13:04,824
told Ars Technica, quote, all of, all of that does
2958
03:13:04,936 --> 03:13:09,443
display whole body control because you need to move all the joints— hands,
2959
03:13:09,491 --> 03:13:13,356
legs, arm, body, everything— to accomplish the task, said Patel, who is also
2960
03:13:13,856 --> 03:13:17,366
a graduate student member of IEEE. So there's definitely been a huge push in
2961
03:13:17,866 --> 03:13:20,686
engineering from where they were this time to last time.
2962
03:13:21,696 --> 03:13:25,096
However, the robots on display in the World Humanoid Robot Games are still very far
2963
03:13:25,224 --> 03:13:28,673
from general-purpose autonomous robots capable of handling practically any task.
2964
03:13:28,993 --> 03:13:32,202
Instead, these robots were engineered to accomplish just one task,
2965
03:13:32,506 --> 03:13:35,779
as shown by the fact that the running robots sprinted in a straight line and
2966
03:13:35,795 --> 03:13:39,131
could not even slow down by, uh, by continuing around the track after
2967
03:13:39,163 --> 03:13:42,468
the finish line. If you cross the finish line, that's the goal. It doesn't
2968
03:13:42,968 --> 03:13:45,163
matter if you stop, and it doesn't matter if you fall into different pieces,
2969
03:13:45,515 --> 03:13:49,253
Patel told Ars. Some companies had different types of robots for different games
2970
03:13:49,285 --> 03:13:53,087
or different tasks because they were engineered for that task to be best in
2971
03:13:53,135 --> 03:13:56,872
that. The robotic races and other flashy events provided for an entertaining
2972
03:13:56,904 --> 03:14:00,370
spectacle that can lead to easy news headlines and viral Video clips.
2973
03:14:00,916 --> 03:14:03,950
They also gave robotics companies an opportunity to attract
2974
03:14:04,450 --> 03:14:08,093
more publicity and investor money by potentially standing out from the competition,
2975
03:14:08,141 --> 03:14:11,449
Patel said. One type of humanoid robot developed by the
2976
03:14:11,949 --> 03:14:15,575
Beijing-based company X Humanoid gained viral online fame for winning the
2977
03:14:15,591 --> 03:14:19,268
small group 400-meter race with an unusual shy person hiding
2978
03:14:19,300 --> 03:14:23,121
behind their hands posture. The robot's developers told Global Times
2979
03:14:23,378 --> 03:14:26,395
that the robot learned through reinforcement training. training, uh,
2980
03:14:26,508 --> 03:14:30,034
reinforcement learning trials and simulation to shift from a human-style runner's
2981
03:14:30,050 --> 03:14:34,026
arms to swing holding its arms up near its face and swinging
2982
03:14:34,074 --> 03:14:37,937
them back and forth through hip rotations. Measures of robotic progress
2983
03:14:38,450 --> 03:14:41,592
at the games could prove even more meaningful if, for example, they challenged
2984
03:14:42,092 --> 03:14:45,584
robots to run the same race 10 times, Patel suggested. That would require the
2985
03:14:46,084 --> 03:14:49,512
robots to demonstrate they have the software and hardware to perform reliably over multiple
2986
03:14:49,528 --> 03:14:53,029
trials, and that maybe even learn how to slow down safely without
2987
03:14:53,529 --> 03:14:54,873
endangering themselves or bystanders.
2988
03:14:57,440 --> 03:15:01,648
So Alex, did you see any videos of these fun games? Oh yeah.
2989
03:15:02,051 --> 03:15:05,754
On your feed? Any, any thoughts on, on what you saw there?
2990
03:15:07,289 --> 03:15:11,299
Well, you know, at least they're making, making them entertaining
2991
03:15:11,719 --> 03:15:15,472
as opposed to, you know, like, I mean, it's, it is said,
2992
03:15:15,782 --> 03:15:17,527
and please don't accuse me of Chinese propaganda,
2993
03:15:19,390 --> 03:15:20,876
But, um,
2994
03:15:21,992 --> 03:15:25,579
the Chinese in general are much more positive
2995
03:15:26,079 --> 03:15:30,492
about AI than Americans are. Um, so it
2996
03:15:30,992 --> 03:15:34,579
kind of is of a piece with that vibe. You see
2997
03:15:35,079 --> 03:15:37,187
what I mean? Okay. All right.
2998
03:15:38,536 --> 03:15:41,878
Well, are— were the entrants,
2999
03:15:42,071 --> 03:15:45,348
uh, only from Japanese— or sorry, uh, Chinese companies,
3000
03:15:45,380 --> 03:15:49,356
or did they have more global admissions if you meet certain
3001
03:15:49,856 --> 03:15:51,258
criteria. I'm not aware.
3002
03:15:54,130 --> 03:15:56,380
Now, I would like to say something.
3003
03:15:58,566 --> 03:16:03,267
You know, the algorithms for training these robots are actually not that
3004
03:16:03,658 --> 03:16:03,756
new.
3005
03:16:07,766 --> 03:16:11,813
Yeah, they're not that new. The difficulty from training these robots really
3006
03:16:12,313 --> 03:16:15,991
comes down to having an accurate simulator, because you need to actually have,
3007
03:16:16,167 --> 03:16:20,486
you know, extreme throughput to be able to get enough repetitions,
3008
03:16:21,225 --> 03:16:24,228
um, to be able to train these models, uh, because you can't,
3009
03:16:24,372 --> 03:16:28,194
you can't do it in host really. You can
3010
03:16:28,306 --> 03:16:31,662
get some biofeedback to help, uh, to help you once
3011
03:16:32,162 --> 03:16:36,013
you put it in a, a body to then test out. And you— I
3012
03:16:36,077 --> 03:16:39,835
call biofeedback sensors on the, on the, the chassis
3013
03:16:40,349 --> 03:16:44,389
to, to try to get real-world feedback
3014
03:16:44,501 --> 03:16:48,589
from the algorithm you run when you run it, but the simulations
3015
03:16:49,089 --> 03:16:53,237
have to actually be prime. And the simulations then are going to be sensitive
3016
03:16:53,253 --> 03:16:56,747
to the chassis that you actually have ultimately put it in. So you're gonna have,
3017
03:16:57,885 --> 03:17:01,796
it's gonna be difficult to train a universal model unless
3018
03:17:01,892 --> 03:17:06,252
the simulations are extremely good. If the simulations
3019
03:17:06,752 --> 03:17:09,570
are not extremely good and you don't have enough variation in in,
3020
03:17:10,022 --> 03:17:13,266
in those simulations to represent various models, you're kind of
3021
03:17:13,766 --> 03:17:17,074
going to have a one-to-one correspondence between the model
3022
03:17:17,107 --> 03:17:19,513
itself and its ability to perform the task like that.
3023
03:17:20,742 --> 03:17:24,235
Uh, the Gator, all of this brings me to the post I just, uh,
3024
03:17:25,109 --> 03:17:28,644
shared to the space, which— Okay, I saw something
3025
03:17:28,660 --> 03:17:32,111
happen. Have you seen Microsoft?
3026
03:17:32,176 --> 03:17:35,319
I've, I have received, I received this and I've seen it
3027
03:17:35,819 --> 03:17:39,261
on my feed. I am, I am a huge fan of ducks,
3028
03:17:40,095 --> 03:17:42,949
but I also don't want something that is technically a spy,
3029
03:17:43,542 --> 03:17:47,648
uh, a spy apparatus that I am destined
3030
03:17:48,148 --> 03:17:52,010
to love because of its representation, uh, inside of my house.
3031
03:17:53,004 --> 03:17:56,131
Well, I mean, apparently this is completely open source and you
3032
03:17:56,631 --> 03:17:59,835
write the software for it, so, um, you can rest assured, um, okay,
3033
03:17:59,900 --> 03:18:03,699
all right, camera fleet, unless you want it to Which is
3034
03:18:03,715 --> 03:18:04,622
a whole other story.
3035
03:18:06,945 --> 03:18:09,989
I'm just really worried about foul play in this scenario.
3036
03:18:11,462 --> 03:18:11,608
And,
3037
03:18:12,368 --> 03:18:16,248
y'know,
3038
03:18:16,506 --> 03:18:19,529
I— Anyway, yes, if you want to buy one, they're only $399.
3039
03:18:20,542 --> 03:18:23,446
So, uh, while supplies last. Alright. So,
3040
03:18:23,688 --> 03:18:26,867
for potentially our last Traces
3041
03:18:27,060 --> 03:18:30,811
of AI Dystopia story, Alex, this is related
3042
03:18:31,311 --> 03:18:35,146
to something that you, uh, sent me, uh, which I had also seen,
3043
03:18:35,259 --> 03:18:37,652
but you also sent me in, in DM.
3044
03:18:39,205 --> 03:18:42,514
Uh, Paxton promises to ban Chinese technology and chatbots that
3045
03:18:43,014 --> 03:18:45,172
endanger children in data center platform.
3046
03:18:46,074 --> 03:18:49,504
Attorney General Ken Paxton on Monday pledged to ban Chinese
3047
03:18:49,536 --> 03:18:53,117
technology from U.S. data centers. I don't know what that
3048
03:18:53,150 --> 03:18:56,523
means. And hold companies criminally liable if their data centers
3049
03:18:56,555 --> 03:19:00,093
power AI chatbots that endanger children if elected to
3050
03:19:00,593 --> 03:19:04,222
Senate. His data center platform stakes out his position on what has become
3051
03:19:04,286 --> 03:19:07,484
a flashpoint in Texas' U.S. Senate race after months of
3052
03:19:07,984 --> 03:19:11,340
pressure from Democratic opponent James Talarico, who rolled out his plan
3053
03:19:11,840 --> 03:19:15,422
more than a month ago. He also backed the DATA Act, a bill by Senator
3054
03:19:15,922 --> 03:19:19,439
Tom Cotton that would let data centers build off-the-grid power systems
3055
03:19:19,939 --> 03:19:23,311
exempt from federal regulation, and a champion Governor
3056
03:19:23,811 --> 03:19:27,630
Greg Abbott's state-level regulatory push, including a repeal of the industry's sales
3057
03:19:28,130 --> 03:19:31,147
tax exemptions, among other outdated incentives,
3058
03:19:31,484 --> 03:19:34,711
per the plan. He also said he would back a ban on AI data
3059
03:19:35,211 --> 03:19:38,500
centers in rural Texas, which Abbott says has also floated
3060
03:19:38,549 --> 03:19:42,097
on the campaign trail. The 2 Senate candidates are backing similar
3061
03:19:42,597 --> 03:19:46,384
policy regulations requiring water systems that don't guzzle nearly as much
3062
03:19:46,512 --> 03:19:50,125
of the resource, getting the industry to pay its own infrastructure costs,
3063
03:19:50,446 --> 03:19:54,437
and terminating the tax breaks. However, as a freshman state senator in 2013,
3064
03:19:54,469 --> 03:19:58,094
Paxton voted for the data center sales tax exemption that helped spark
3065
03:19:58,594 --> 03:20:02,337
the industry's boom in Texas, a break the state comptroller— a word I
3066
03:20:02,450 --> 03:20:05,636
hate so much— now projects will cost
3067
03:20:05,765 --> 03:20:08,641
$3.2 billion in revenue over the next 2 years.
3068
03:20:10,360 --> 03:20:14,297
Paxton's announcement caps a months-long stretch of silence on the front
3069
03:20:14,797 --> 03:20:18,807
that exploded into the center of the campaign. Data centers, large warehouses of computer servers
3070
03:20:19,307 --> 03:20:22,822
that power artificial intelligence and cloud computing have proliferated across
3071
03:20:22,870 --> 03:20:26,012
Texas faster than almost anywhere in the country, drawing fierce
3072
03:20:26,076 --> 03:20:29,459
backlash from communities concerned about noise, water use,
3073
03:20:29,523 --> 03:20:33,579
and rising electricity bills. A Texas Tribune analyst found the state
3074
03:20:33,659 --> 03:20:37,458
has 335 existing data centers with more than 248
3075
03:20:37,958 --> 03:20:41,225
in the works. And June poll— and a June poll from the Texas
3076
03:20:41,725 --> 03:20:45,772
Politics Project at the University of Texas at Austin found a majority of Texans oppose
3077
03:20:45,804 --> 03:20:48,273
construction of facilities in their communities.
3078
03:20:50,220 --> 03:20:53,876
Talarico, an Austin state representative, has used the GOP's early support for data
3079
03:20:54,376 --> 03:20:57,921
center investments as a persistent line of attack, unveiling his Hold Data
3080
03:20:58,421 --> 03:21:02,174
Centers Accountable plan in July. His plan goes one step further than
3081
03:21:02,271 --> 03:21:05,873
his GOP opponent, calling for giving local officials approval
3082
03:21:05,905 --> 03:21:07,663
power over data center projects.
3083
03:21:09,366 --> 03:21:13,277
Talarico pressed Paxton over a request from Hood County, a rural community
3084
03:21:13,777 --> 03:21:17,021
southwest of Fort Worth, where developers have proposed 8 data centers spanning
3085
03:21:17,117 --> 03:21:20,547
more than 7,600 acres. County officials
3086
03:21:20,580 --> 03:21:24,395
were asked— they asked the attorney general's office in February
3087
03:21:24,427 --> 03:21:28,049
whether the county could impose a moratorium on data center development.
3088
03:21:28,498 --> 03:21:30,951
6 months later, the opinion is still without a decision.
3089
03:21:31,544 --> 03:21:35,054
And Notice reported this month that Paxton's political operation
3090
03:21:35,391 --> 03:21:39,270
has taken in nearly $450,000 from data center-affiliated donors.
3091
03:21:39,655 --> 03:21:43,470
Quote, Ken Paxton has taken nearly half a million dollars from the data center industry
3092
03:21:43,486 --> 03:21:47,034
voted to flood Texas with data centers by giving them billions in tax
3093
03:21:47,534 --> 03:21:50,709
handouts and has refused to use his power as attorney general to respond
3094
03:21:51,209 --> 03:21:53,647
to Texans' requests to block data center development in their communities,
3095
03:21:54,016 --> 03:21:57,917
Tallarico said in a statement Monday. Now he's scared his ties
3096
03:21:57,965 --> 03:22:01,159
to big data center companies will lose him the election, so he's
3097
03:22:01,659 --> 03:22:05,172
making a sad attempt to reverse course by releasing a plan he literally
3098
03:22:05,252 --> 03:22:08,430
used AI to make. Paxton used similar language
3099
03:22:08,479 --> 03:22:11,672
in his statement on the plan's release to jab Tallarico Rico and his
3100
03:22:11,736 --> 03:22:16,337
campaign infrastructure that accepted millions of dollars from data center-affiliated
3101
03:22:16,385 --> 03:22:20,024
donors. Reid Hoffman, LinkedIn's co-founder and an outspoken advocate
3102
03:22:20,524 --> 03:22:23,727
of expanding data centers to edge out China, has given $10 million to Lone Star
3103
03:22:23,775 --> 03:22:25,698
Rising, a super PAC supporting Tallarico.
3104
03:22:26,420 --> 03:22:29,914
Paxton's plan targeting China's interest dovetails with a broader Republican
3105
03:22:29,946 --> 03:22:33,633
effort to reframe the race around national
3106
03:22:34,133 --> 03:22:38,041
security. All right, Alex. So specifically the idea
3107
03:22:38,541 --> 03:22:42,029
that you were pointing to was you pointed
3108
03:22:42,061 --> 03:22:45,840
to a post from Adam Theurer, who I believe works
3109
03:22:46,340 --> 03:22:49,888
for FIRE, FIRE.org. And he points out—
3110
03:22:50,388 --> 03:22:54,188
The second prong of Ken Paxton's data center regulatory
3111
03:22:54,688 --> 03:22:58,596
platform would enshrine into law just about the most open-ended and destructive theory
3112
03:22:58,628 --> 03:23:01,986
of vicarious liability in history, stating that data centers
3113
03:23:02,002 --> 03:23:04,604
will be held criminally liable for downstream uses of AI.
3114
03:23:04,812 --> 03:23:08,425
Chatbots that undermine child safety. However, that ends up
3115
03:23:08,925 --> 03:23:12,326
being defined. This is like saying an electric electricity company
3116
03:23:12,342 --> 03:23:15,875
will be held criminally liable for problems that develop with any appliance or
3117
03:23:15,891 --> 03:23:19,022
product plugged into the grid, or if applied to
3118
03:23:19,522 --> 03:23:23,325
gun manufacturers, it would mean a firearm ammunition maker will be criminally
3119
03:23:23,357 --> 03:23:27,066
liable for any act of violence committed by a person using their gun or ammo.
3120
03:23:27,676 --> 03:23:30,405
And that's I'll just cut that off there so you can have the discussion.
3121
03:23:30,726 --> 03:23:33,890
As this is probably the last segment here. Yeah,
3122
03:23:34,018 --> 03:23:37,358
I don't know, I'll tell you, man. I mean, you hear—
3123
03:23:37,470 --> 03:23:40,553
I don't know, this whole thing is so stupid, I'm about to
3124
03:23:41,163 --> 03:23:45,033
harm myself. I don't know if that would— Alex
3125
03:23:45,533 --> 03:23:48,678
is kidding about threatening self-harm or
3126
03:23:49,031 --> 03:23:51,728
any of the— He's sarcastic. Yes, yes.
3127
03:23:52,901 --> 03:23:55,919
I don't know who's gonna be held responsible though, were I not kidding.
3128
03:23:56,970 --> 03:24:00,528
Because this is not me. Not me. I'm not sure. I'm not sure. Watcher.
3129
03:24:01,984 --> 03:24:05,559
What do you think, Gator? I mean, what do you think about all this?
3130
03:24:06,433 --> 03:24:09,847
Well, of course, that type of proposal is the,
3131
03:24:10,252 --> 03:24:13,504
you know, the phenomenon we've talked about many times,
3132
03:24:14,320 --> 03:24:18,125
which is the desire to keep pushing things down the chain of
3133
03:24:18,221 --> 03:24:21,253
causality. Right. So, so,
3134
03:24:21,994 --> 03:24:25,997
you know, So the, the, the device
3135
03:24:26,125 --> 03:24:29,786
that is harming children has a company that's hosting
3136
03:24:29,963 --> 03:24:32,950
it. It's not the company and the host, or maybe it's also the company and
3137
03:24:33,450 --> 03:24:37,092
the host, but then now it's just the, the hardware that was running
3138
03:24:38,232 --> 03:24:41,106
and jamming the giant matrices together.
3139
03:24:41,460 --> 03:24:44,719
I don't know. Are they having a hard time figuring out who
3140
03:24:44,735 --> 03:24:47,128
was hosting the application that was damaging?
3141
03:24:48,125 --> 03:24:51,071
The children— like, this is different because it's not even that,
3142
03:24:52,890 --> 03:24:56,127
like, they can't stop X, so they're stopping Y. It's like,
3143
03:24:56,207 --> 03:24:59,267
no, no, we can totally fine the company and, and sue them. We're just also
3144
03:24:59,283 --> 03:25:02,622
going to somehow sue you for providing them infrastructure,
3145
03:25:03,875 --> 03:25:07,279
just, just because we can. Right. And, and,
3146
03:25:07,538 --> 03:25:10,850
and that— so where they're like, anything bad
3147
03:25:10,882 --> 03:25:14,275
that comes downstream of your thing, we'll just attach
3148
03:25:15,586 --> 03:25:20,148
So let's get
3149
03:25:20,648 --> 03:25:23,281
off conceptual issues with it as far as like,
3150
03:25:23,859 --> 03:25:26,944
that's a bad practice to have to continue to push things down
3151
03:25:26,992 --> 03:25:31,282
the causal chain and create criminal
3152
03:25:31,571 --> 03:25:35,156
statutes and make behavior crimes that are not otherwise
3153
03:25:35,253 --> 03:25:40,650
criminal. Meaning having a data center that computes
3154
03:25:40,730 --> 03:25:44,389
giant matrices and tensor multiplications together
3155
03:25:44,889 --> 03:25:48,344
is not something that you're gonna charge a crime for because, you know,
3156
03:25:48,360 --> 03:25:52,037
multiplying tensors is not— and matrices is not a crime.
3157
03:25:52,617 --> 03:25:55,228
Okay? It's a perfectly legal thing to do.
3158
03:25:57,048 --> 03:26:00,255
We need this in the— This has a practical problem. This has a practical problem.
3159
03:26:01,659 --> 03:26:03,979
It has a practical problem because not only,
3160
03:26:05,123 --> 03:26:07,846
not only can the same,
3161
03:26:08,442 --> 03:26:11,761
quote, AI chatbot
3162
03:26:12,261 --> 03:26:16,307
bot be used 2 billion times for innocuous purposes and harm a child
3163
03:26:17,045 --> 03:26:21,152
once, right? It is also a
3164
03:26:21,216 --> 03:26:25,612
problem where you cannot anticipate in advance necessarily
3165
03:26:26,190 --> 03:26:29,880
what is going to cause harm to a child. If you
3166
03:26:30,380 --> 03:26:34,051
have all of— if the company that is using this data center to
3167
03:26:34,131 --> 03:26:37,386
host the chatbot has a myriad of
3168
03:26:37,418 --> 03:26:41,023
safeguards that it tries to actually enforce, not just
3169
03:26:42,466 --> 03:26:45,879
laziness, but an actual myriad of them, and those safeguards
3170
03:26:46,379 --> 03:26:50,862
still get circumvented, the data center has absolutely
3171
03:26:50,910 --> 03:26:54,916
no tie to controlling that outcome.
3172
03:26:56,086 --> 03:26:59,643
It has no ability. And in fact, anything they do to control
3173
03:27:00,143 --> 03:27:04,005
that outcome is things that we consider bad. monitoring exactly
3174
03:27:04,037 --> 03:27:07,473
what the apps are doing, intervening in how the apps are operating. Having to give
3175
03:27:07,973 --> 03:27:10,990
approval for every activity that is being done.
3176
03:27:11,825 --> 03:27:15,052
Yes. Anyway, but, but this gets very stupid very fast,
3177
03:27:15,100 --> 03:27:19,018
right? Like, what if like a gang of like ne'er-do-wells
3178
03:27:20,399 --> 03:27:23,931
use Google Calendar to meet up and go and beat up a child?
3179
03:27:24,269 --> 03:27:27,833
Not like— now what are— okay, well,
3180
03:27:28,026 --> 03:27:31,977
yes, the app was used to, you know, cause harm to a child or
3181
03:27:32,041 --> 03:27:34,724
led to harm to a child or whatever the lingo is.
3182
03:27:35,833 --> 03:27:39,705
So we're going to punish Google or no,
3183
03:27:39,769 --> 03:27:42,885
no, no, no, we're not going to punish Google. We're going to punish whoever was
3184
03:27:42,998 --> 03:27:46,520
hosting the chips that Google used
3185
03:27:47,021 --> 03:27:51,159
to serve Google Calendar. Now, now this
3186
03:27:51,659 --> 03:27:54,389
is also, this is also a crapshoot. Okay.
3187
03:27:55,000 --> 03:27:58,389
This is, this is a, this is a spinning wheel.
3188
03:27:59,333 --> 03:28:02,750
of blame. All right, if you are a large company,
3189
03:28:03,729 --> 03:28:07,531
a large company, you are, you're typically not
3190
03:28:07,595 --> 03:28:11,413
going to have all of your eggs in one basket. You are going
3191
03:28:11,509 --> 03:28:14,750
to have redundancy upon redundancy. Now maybe,
3192
03:28:15,295 --> 03:28:20,268
maybe you have a contract with one company's
3193
03:28:20,768 --> 03:28:24,263
data centers that, that you are signing an
3194
03:28:24,763 --> 03:28:28,549
agreement with. For example, For example, I might have an app that
3195
03:28:28,597 --> 03:28:32,595
has redundancy in the standpoint of it propping
3196
03:28:32,691 --> 03:28:36,625
up different, spinning up different things in the background to
3197
03:28:37,125 --> 03:28:40,752
make sure it can handle its throughput. But those could be geographically spaced out to
3198
03:28:41,252 --> 03:28:45,569
the point where there isn't an actual geography that is consistent across the entire application.
3199
03:28:45,617 --> 03:28:48,812
It is very variable. And Microsoft, let's say I'm
3200
03:28:48,844 --> 03:28:52,985
doing this all in a Microsoft-hosted environment. Microsoft themselves
3201
03:28:53,098 --> 03:28:57,939
might have cut deals with a different data center to
3202
03:28:58,439 --> 03:29:01,134
handle some of its consumption.
3203
03:29:01,489 --> 03:29:04,684
Yeah. It is not necessarily the prime owner
3204
03:29:04,717 --> 03:29:07,895
of all of the real estate/data centers themselves.
3205
03:29:07,944 --> 03:29:11,489
They might contract out to another data center to run their
3206
03:29:11,795 --> 03:29:15,097
infrastructure on. Okay? Which means that
3207
03:29:15,178 --> 03:29:19,736
I am not guaranteed that the thing that I've developed
3208
03:29:20,236 --> 03:29:23,716
developed is going to execute on any given specific device, any specific
3209
03:29:23,957 --> 03:29:27,136
given server in any one of these data centers. When I'm saying it's in,
3210
03:29:27,409 --> 03:29:31,439
you know, East US 1 or East US 2 or something like that,
3211
03:29:32,065 --> 03:29:35,790
it is now— Look, we are now sitting here getting nerd-sniped
3212
03:29:36,290 --> 03:29:40,269
about the idiocies that are clearly the hallucinations of a boomer
3213
03:29:40,769 --> 03:29:43,609
mind who says, if you hurt children, we're going to punish you. And if you
3214
03:29:43,625 --> 03:29:47,014
use any Chinese software, I'm going to punish you. It doesn't have to make any
3215
03:29:47,514 --> 03:29:50,146
sense, and it doesn't. And we're struggling, our brains are struggling to try to make
3216
03:29:50,646 --> 03:29:53,905
sense of what the fuck he's talking about. He's not talking about anything per se,
3217
03:29:54,403 --> 03:29:57,921
okay? He's putting nice words in order that sound good because he
3218
03:29:58,421 --> 03:30:01,680
wants to get elected. And we're here now stuck trying to figure out
3219
03:30:01,696 --> 03:30:04,925
what harm to children means when, you know, you've got 6
3220
03:30:05,425 --> 03:30:08,539
degrees of infrastructure in between. I would say that that is what we signed up
3221
03:30:09,039 --> 03:30:11,639
to do when we named this show This Dumb Week. Yeah, fair enough.
3222
03:30:12,153 --> 03:30:16,424
Fair enough. But we, we need, we need some policies here.
3223
03:30:17,356 --> 03:30:21,033
But legitimately, legitimately, what, what, what you've developed,
3224
03:30:21,049 --> 03:30:24,308
this thing that could some, you know, somewhere endanger children,
3225
03:30:24,357 --> 03:30:28,178
is now possibly running on a random, random spinning
3226
03:30:28,291 --> 03:30:31,871
wheel of different places. And it is just a crapshoot
3227
03:30:31,904 --> 03:30:35,356
which one of those companies is going to be the data center that gets shit
3228
03:30:35,420 --> 03:30:39,547
on because the code executed on that specific— But you don't understand,
3229
03:30:39,675 --> 03:30:41,230
like Ultimately, what happens is,
3230
03:30:43,500 --> 03:30:48,120
were anything remotely like this to pass
3231
03:30:48,620 --> 03:30:52,232
in Texas, which it wouldn't, and if it did, it would fall in some
3232
03:30:52,732 --> 03:30:56,221
Supreme Court or other, but let's say somehow it came into actual effect,
3233
03:30:57,732 --> 03:31:02,003
the data centers would simply get up and leave Texas. There's no— Yeah.
3234
03:31:02,067 --> 03:31:05,306
You don't put yourself in the position of having a
3235
03:31:05,806 --> 03:31:09,068
Russian roulette to your head and sitting there happily hoping that— Mind you,
3236
03:31:09,084 --> 03:31:12,791
this is a Senate campaign. This is a Senate campaign. This is a Senate
3237
03:31:13,291 --> 03:31:16,722
campaign. So this is not just Texas proposal. So you
3238
03:31:17,222 --> 03:31:21,118
leave the country. I mean, it's just not, not feasible to operate in
3239
03:31:21,618 --> 03:31:24,520
an environment where you literally have a gun to your head. Well, I mean,
3240
03:31:25,049 --> 03:31:29,574
what, what, uh, data centers does, uh, do, uh, does Roblox
3241
03:31:29,638 --> 03:31:33,008
run on? Because, uh, let's talk about some harm to children.
3242
03:31:33,072 --> 03:31:36,440
Roblox is— yes, exactly. Whoever hosts a Roblox server
3243
03:31:36,940 --> 03:31:39,992
is doomed under that proposed legislation. Okay. If it's,
3244
03:31:40,121 --> 03:31:43,174
if it's about a Senate campaign,
3245
03:31:43,191 --> 03:31:47,393
then fuck it. Like, there's zero chance that any of this gets
3246
03:31:48,729 --> 03:31:52,143
applied because like, this is like some rando from Texas
3247
03:31:52,643 --> 03:31:55,247
is gonna come in and say, vote for my bill. And everybody is gonna be
3248
03:31:55,747 --> 03:31:57,916
like, get on, get in the line. And no.
3249
03:31:59,620 --> 03:32:02,821
So, so it, it's all signaling. And, but,
3250
03:32:02,837 --> 03:32:06,032
but, but I, I, you know, I continue to think
3251
03:32:06,532 --> 03:32:10,689
that, you know, trying to make technical sense of these things is hopeless.
3252
03:32:11,010 --> 03:32:14,334
And it— the situation makes a lot more sense
3253
03:32:14,834 --> 03:32:18,139
if you just think about data centers as bad thing that does bad things.
3254
03:32:18,717 --> 03:32:21,832
And the laws as like, we don't, you know, we're tough on bad things.
3255
03:32:22,346 --> 03:32:24,835
You know, it used to be tough on crime. Now it's tough on data centers,
3256
03:32:24,851 --> 03:32:26,424
apparently. Good job.
3257
03:32:29,217 --> 03:32:32,419
Yay! Those big bad data centers are gonna get what's coming to them.
3258
03:32:33,095 --> 03:32:36,169
We're gonna— You know, I was wondering, why do they— why— how,
3259
03:32:36,555 --> 03:32:39,948
how do they come to be called data centers? Because yeah, okay, they have a
3260
03:32:39,996 --> 03:32:43,130
lot of data in them, but they also have a lot of like chips and
3261
03:32:43,630 --> 03:32:46,891
computers and— Well, yeah, they're actually primarily not even doing much.
3262
03:32:47,165 --> 03:32:50,326
They're, they're data intermediaries, you know.
3263
03:32:50,471 --> 03:32:54,100
They're throughput systems. They're not even how, you know, often the
3264
03:32:54,600 --> 03:32:57,977
housing Yeah, exactly. So, you know,
3265
03:32:57,993 --> 03:33:01,540
it is a weird name for them, but I, I think it's definitely not helping.
3266
03:33:02,160 --> 03:33:05,657
Maybe, um, you know, the next generation should
3267
03:33:06,157 --> 03:33:09,928
be called something else to throw people off. Compute centers, I don't
3268
03:33:10,428 --> 03:33:13,801
know, um, they'll be more accurate. Well,
3269
03:33:13,978 --> 03:33:17,188
Alex, uh, we'll, we'll put that as the tail end here.
3270
03:33:17,349 --> 03:33:21,059
We're sounding incoherent, but it's just the subject matter that is doing
3271
03:33:21,170 --> 03:33:24,919
doing this, people. Take my word for it. Well, uh,
3272
03:33:25,112 --> 03:33:28,621
we're approaching— we're approaching a time period where maybe my coherency
3273
03:33:28,733 --> 03:33:31,986
seems to decrease, uh, as every minute, uh, it goes
3274
03:33:32,083 --> 03:33:35,688
on. So we will try to cut this off and round this episode
3275
03:33:36,188 --> 03:33:38,396
out. It's been a good one. I think, uh, that puts a tail end on
3276
03:33:38,429 --> 03:33:42,002
Traces of AI Dystopia. Unfortunately, um, you know, we lost
3277
03:33:42,499 --> 03:33:46,137
Optimus Prime and we're not going to get him back. So, um, the world
3278
03:33:46,637 --> 03:33:48,646
is a little bit weaker out there than it was before,
3279
03:33:49,741 --> 03:33:53,171
but we will try to carry on in his legacy. Alex, uh, I hope you
3280
03:33:53,204 --> 03:33:57,004
have a wonderful week, and I will see you next time.
3281
03:33:57,810 --> 03:33:58,889
Same to you, Gator.