Algorithmic Outing: When Your Feed Knows Before You Do #1869

1
00:00:04,080 --> 00:00:06,480
[Speaker 0]
Hello, and welcome to the Geek News Central
2
00:00:06,480 --> 00:00:10,960
[Speaker 0]
podcast episode eighteen sixty nine. I'm your host
3
00:00:10,960 --> 00:00:12,880
[Speaker 0]
tonight, Ray Cochran, and I have a great
4
00:00:12,880 --> 00:00:15,255
[Speaker 0]
episode lined up for you folks. Leading off
5
00:00:15,255 --> 00:00:18,135
[Speaker 0]
tonight, new research out of Australia says the
6
00:00:18,135 --> 00:00:21,575
[Speaker 0]
recommendation algorithms running your social feeds can figure
7
00:00:21,575 --> 00:00:24,235
[Speaker 0]
out something as private as your sexual orientation
8
00:00:24,535 --> 00:00:26,955
[Speaker 0]
before you have consciously admitted to it yourself.
9
00:00:27,860 --> 00:00:30,020
[Speaker 0]
The researchers even gave it a name. They're
10
00:00:30,020 --> 00:00:33,880
[Speaker 0]
calling it algorithmic outing. Also in tonight's headlines,
11
00:00:34,180 --> 00:00:37,480
[Speaker 0]
Meta's blueprint for building privacy into AI infrastructure,
12
00:00:38,100 --> 00:00:40,820
[Speaker 0]
the government of Alberta turning Claude loose on
13
00:00:40,820 --> 00:00:43,984
[Speaker 0]
four hundred sixty six million lines of code.
14
00:00:44,524 --> 00:00:47,265
[Speaker 0]
Ghoul DeepMind teaming up with the film studio
15
00:00:47,405 --> 00:00:51,805
[Speaker 0]
A24. Cyborg cockroaches that can scuba dive for
16
00:00:51,805 --> 00:00:56,730
[Speaker 0]
three hours. Physicists at Los Alamos making a
17
00:00:56,730 --> 00:00:59,850
[Speaker 0]
quantum system look like it's running backward in
18
00:00:59,850 --> 00:01:04,110
[Speaker 0]
time. The US Department of Energy quietly deleting
19
00:01:04,250 --> 00:01:07,930
[Speaker 0]
six thousand pages about saving energy in the
20
00:01:07,930 --> 00:01:11,204
[Speaker 0]
middle of a heat wave. And NASA scrambling
21
00:01:11,265 --> 00:01:13,585
[Speaker 0]
to catch a falling telescope before it burns
22
00:01:13,585 --> 00:01:17,204
[Speaker 0]
up in the atmosphere. Let's get into it.
23
00:01:18,145 --> 00:01:19,985
[Speaker 0]
Tonight's show is brought to you by our
24
00:01:19,985 --> 00:01:24,200
[Speaker 0]
twenty year partner, GoDaddy. Exclusive deals and pricing
25
00:01:24,259 --> 00:01:27,000
[Speaker 0]
at geek new central dot com forward slash
26
00:01:27,060 --> 00:01:31,380
[Speaker 0]
godaddy. I guess before we dive into the
27
00:01:31,380 --> 00:01:35,219
[Speaker 0]
stories, quick personal update. Hoping everyone had a
28
00:01:35,219 --> 00:01:37,795
[Speaker 0]
good holiday weekend. Spent mine working at my
29
00:01:37,795 --> 00:01:40,515
[Speaker 0]
other job, Oregon Spanish, chatting with the good
30
00:01:40,515 --> 00:01:43,634
[Speaker 0]
people of Portland. And honestly, I feel like
31
00:01:43,634 --> 00:01:45,095
[Speaker 0]
that's a big part of why I still
32
00:01:45,155 --> 00:01:47,075
[Speaker 0]
like that job. Oh, one thing I end
33
00:01:47,075 --> 00:01:48,835
[Speaker 0]
up missing out on a lot through the
34
00:01:48,835 --> 00:01:53,890
[Speaker 0]
week is, like, socializing. My bluebird work can
35
00:01:53,890 --> 00:01:56,850
[Speaker 0]
feel pretty solitary usually because outside of, like,
36
00:01:56,850 --> 00:01:59,250
[Speaker 0]
meetings, you mostly do your own thing. So
37
00:01:59,250 --> 00:02:01,810
[Speaker 0]
I kinda find myself refilling that empty social
38
00:02:01,810 --> 00:02:03,810
[Speaker 0]
meter on the weekends at the other job.
39
00:02:03,810 --> 00:02:07,774
[Speaker 0]
So it's a pretty good balance. Saturday night,
40
00:02:07,774 --> 00:02:09,375
[Speaker 0]
went out with some coworkers and we hit
41
00:02:09,375 --> 00:02:12,415
[Speaker 0]
a spot called Hungry Tiger. The bartenders there
42
00:02:12,415 --> 00:02:14,495
[Speaker 0]
are great. They've got a pretty cool selection
43
00:02:14,495 --> 00:02:16,894
[Speaker 0]
including one that Delaney keeps picking up whenever
44
00:02:16,894 --> 00:02:20,010
[Speaker 0]
we would go, the Tiger Tea. We've also
45
00:02:20,010 --> 00:02:21,610
[Speaker 0]
got a lot of vegan options for a
46
00:02:21,610 --> 00:02:23,850
[Speaker 0]
lot of the menu, corn dog, wings, that
47
00:02:23,850 --> 00:02:26,010
[Speaker 0]
kind of thing. So it was nothing too
48
00:02:26,010 --> 00:02:28,010
[Speaker 0]
crazy, just, you know, good bar food, but
49
00:02:28,010 --> 00:02:30,490
[Speaker 0]
it's always a good time. And we usually
50
00:02:30,490 --> 00:02:32,090
[Speaker 0]
end up sitting out on the patio so
51
00:02:32,090 --> 00:02:33,770
[Speaker 0]
we can, you know, hang out and smoke
52
00:02:33,770 --> 00:02:36,224
[Speaker 0]
as we please. And I gotta say, you
53
00:02:36,224 --> 00:02:38,064
[Speaker 0]
know, like a patio space has kinda become
54
00:02:38,064 --> 00:02:40,305
[Speaker 0]
a hard requirement for us, I feel like,
55
00:02:40,305 --> 00:02:42,224
[Speaker 0]
when we're, like, picking a bar to go
56
00:02:42,224 --> 00:02:44,545
[Speaker 0]
sit at. So luckily, a lot of places
57
00:02:44,545 --> 00:02:46,864
[Speaker 0]
around here keep those stuff or have kept
58
00:02:46,864 --> 00:02:51,739
[Speaker 0]
those up since, post COVID. But, yeah, enough
59
00:02:51,739 --> 00:02:58,720
[Speaker 0]
about me. Let's get into the stories. Featured
60
00:02:58,780 --> 00:03:01,980
[Speaker 0]
article tonight, algorithmic outing when your feed knows
61
00:03:01,980 --> 00:03:04,540
[Speaker 0]
before you do. Found this one on fizz
62
00:03:04,540 --> 00:03:07,475
[Speaker 0]
dot org, and it's one of those pieces
63
00:03:07,475 --> 00:03:09,075
[Speaker 0]
that kinda sits with you after you kinda
64
00:03:09,075 --> 00:03:12,195
[Speaker 0]
read it, type of research that's kinda being
65
00:03:12,195 --> 00:03:16,195
[Speaker 0]
done on, you know, people using social media
66
00:03:16,195 --> 00:03:18,035
[Speaker 0]
in general, but the the new research is
67
00:03:18,035 --> 00:03:21,575
[Speaker 0]
describing something the authors are calling algorithmic outing.
68
00:03:21,910 --> 00:03:24,490
[Speaker 0]
And the short version is that the recommendation
69
00:03:24,710 --> 00:03:27,110
[Speaker 0]
systems running your social media feeds can pick
70
00:03:27,110 --> 00:03:30,570
[Speaker 0]
up your sexual orientation or your gender identity
71
00:03:30,710 --> 00:03:32,870
[Speaker 0]
and start serving you content around it before
72
00:03:32,870 --> 00:03:34,995
[Speaker 0]
you've worked it out, before you've come out
73
00:03:34,995 --> 00:03:37,395
[Speaker 0]
to anyone, and in some cases, before you
74
00:03:37,395 --> 00:03:40,835
[Speaker 0]
even know it yourself. So let me be
75
00:03:40,835 --> 00:03:42,195
[Speaker 0]
a little bit more clear about what the
76
00:03:42,195 --> 00:03:44,835
[Speaker 0]
study actually is because the scope is pretty
77
00:03:44,835 --> 00:03:47,939
[Speaker 0]
important. It's not a giant survey of millions
78
00:03:47,939 --> 00:03:50,579
[Speaker 0]
of accounts. This was a qualitative study, meaning
79
00:03:50,579 --> 00:03:53,379
[Speaker 0]
it's built on in-depth interviews rather than wrong
80
00:03:53,379 --> 00:03:56,500
[Speaker 0]
numbers. And it published an academic journal called
81
00:03:56,500 --> 00:03:59,974
[Speaker 0]
Gender, Place, and Culture. The researchers sat down
82
00:03:59,974 --> 00:04:03,895
[Speaker 0]
with twenty LGBTQ plus adults ranging in age
83
00:04:03,895 --> 00:04:06,455
[Speaker 0]
from eighteen to sixty, all in one region
84
00:04:06,455 --> 00:04:09,655
[Speaker 0]
of Australia, the Hunter region of New South
85
00:04:09,655 --> 00:04:11,815
[Speaker 0]
Wales. So it's a small focus group of
86
00:04:11,815 --> 00:04:14,230
[Speaker 0]
people talking about their lived experience, not a
87
00:04:14,230 --> 00:04:17,190
[Speaker 0]
sweeping statistical claim. So, yeah, that's just something
88
00:04:17,190 --> 00:04:20,150
[Speaker 0]
I want to say upfront, but what those
89
00:04:20,150 --> 00:04:22,790
[Speaker 0]
twenty people describe is, you know, a clear
90
00:04:22,790 --> 00:04:25,510
[Speaker 0]
pattern and that's real and consistent, and here's
91
00:04:25,510 --> 00:04:29,335
[Speaker 0]
how it happens. These algorithms don't really need
92
00:04:29,335 --> 00:04:31,735
[Speaker 0]
you to tell them anything. They watch what
93
00:04:31,735 --> 00:04:34,775
[Speaker 0]
researchers call engagement signals. And that's just like,
94
00:04:34,775 --> 00:04:36,775
[Speaker 0]
you know, little bread crumbs you leave behind
95
00:04:36,775 --> 00:04:39,175
[Speaker 0]
as you scroll, which videos you like, which
96
00:04:39,175 --> 00:04:41,800
[Speaker 0]
creators you follow. And one of the bigger
97
00:04:41,800 --> 00:04:43,800
[Speaker 0]
ones recently is how long you linger on
98
00:04:43,800 --> 00:04:47,240
[Speaker 0]
a given post before you swipe away. That
99
00:04:47,240 --> 00:04:49,160
[Speaker 0]
last one has a name in the industry.
100
00:04:49,160 --> 00:04:51,640
[Speaker 0]
It's called dwell time. And it's exactly what
101
00:04:51,640 --> 00:04:53,400
[Speaker 0]
it sounds like, the number of seconds your
102
00:04:53,400 --> 00:04:55,260
[Speaker 0]
thumb goes still on a piece of content.
103
00:04:55,655 --> 00:04:58,555
[Speaker 0]
You might not even consciously notice you paused,
104
00:04:58,615 --> 00:05:02,295
[Speaker 0]
but the algorithm definitely does. And from thousands
105
00:05:02,295 --> 00:05:04,135
[Speaker 0]
of those tiny signals, it starts building a
106
00:05:04,135 --> 00:05:06,055
[Speaker 0]
model of what you're drawn to, then it
107
00:05:06,055 --> 00:05:09,030
[Speaker 0]
feeds you more of it. Now the lead
108
00:05:09,030 --> 00:05:12,070
[Speaker 0]
researcher on this, doctor Justin Ellis at the
109
00:05:12,070 --> 00:05:15,590
[Speaker 0]
University of Newcastle, and the way people in
110
00:05:15,590 --> 00:05:17,510
[Speaker 0]
this study talked about it is honestly kind
111
00:05:17,510 --> 00:05:19,670
[Speaker 0]
of haunting. But let me read you the
112
00:05:19,670 --> 00:05:23,285
[Speaker 0]
line. Several participants told us the algorithm knew
113
00:05:23,285 --> 00:05:26,005
[Speaker 0]
they were queer before they did. That experience
114
00:05:26,005 --> 00:05:29,625
[Speaker 0]
was described as validating for some, but confronting
115
00:05:29,685 --> 00:05:32,565
[Speaker 0]
or even frightening for others, particularly when it
116
00:05:32,565 --> 00:05:34,824
[Speaker 0]
happened in a public or semi public settings.
117
00:05:35,979 --> 00:05:38,160
[Speaker 0]
And that's a quote from doctor Justin Ellis.
118
00:05:39,020 --> 00:05:41,580
[Speaker 0]
And that public setting piece is kind of
119
00:05:41,580 --> 00:05:44,139
[Speaker 0]
where this stops being like, abs like an
120
00:05:44,139 --> 00:05:46,780
[Speaker 0]
abstract privacy conversation and becomes a real world
121
00:05:46,780 --> 00:05:49,294
[Speaker 0]
problem. You would think about it. Your feed
122
00:05:49,294 --> 00:05:54,014
[Speaker 0]
isn't, like, a neutral private space necessarily. People
123
00:05:54,014 --> 00:05:56,014
[Speaker 0]
can now see what you like. They can
124
00:05:56,014 --> 00:05:58,975
[Speaker 0]
see what you're reposting. So oftentimes, those head
125
00:05:58,975 --> 00:06:00,895
[Speaker 0]
out into the public spaces around the people
126
00:06:00,895 --> 00:06:04,569
[Speaker 0]
that, you know, follow you. Or even, you
127
00:06:04,569 --> 00:06:06,330
[Speaker 0]
know, if you're just in public, and you
128
00:06:06,330 --> 00:06:07,690
[Speaker 0]
put your phone out like on the bus
129
00:06:07,690 --> 00:06:09,449
[Speaker 0]
or something or at the coffee shop, you
130
00:06:09,449 --> 00:06:11,130
[Speaker 0]
know, the screen is right there for anyone
131
00:06:11,130 --> 00:06:15,130
[Speaker 0]
to see. So the algorithm doesn't really take
132
00:06:15,130 --> 00:06:17,610
[Speaker 0]
that into account, just knows what you want
133
00:06:17,610 --> 00:06:19,935
[Speaker 0]
to see or what it'll, what it should
134
00:06:19,935 --> 00:06:23,535
[Speaker 0]
serve you to keep you coming back. And
135
00:06:23,535 --> 00:06:27,855
[Speaker 0]
this concept of risk or context or whether
136
00:06:27,855 --> 00:06:30,975
[Speaker 0]
you're personally ready for it, is not taken
137
00:06:30,975 --> 00:06:32,975
[Speaker 0]
into account whatsoever with those types of things
138
00:06:32,975 --> 00:06:37,670
[Speaker 0]
because they they're deterministic. So for somebody who
139
00:06:37,670 --> 00:06:41,030
[Speaker 0]
hasn't really figured that sort of concept out
140
00:06:41,030 --> 00:06:44,710
[Speaker 0]
yet, you know, the recommendation isn't something that's
141
00:06:44,710 --> 00:06:46,870
[Speaker 0]
convenient, I feel like, you know, that's that's
142
00:06:46,870 --> 00:06:51,625
[Speaker 0]
a exposure to some degree. So the study
143
00:06:51,625 --> 00:06:53,145
[Speaker 0]
found a lot of people do doing a
144
00:06:53,145 --> 00:06:54,985
[Speaker 0]
lot of work to, like, protect themselves. They
145
00:06:54,985 --> 00:06:58,025
[Speaker 0]
dig into, like, privacy settings. They're running multiple
146
00:06:58,025 --> 00:07:00,025
[Speaker 0]
accounts, one face for public and another for
147
00:07:00,025 --> 00:07:03,470
[Speaker 0]
private. Some of them just browse passively, deliberately
148
00:07:03,610 --> 00:07:06,090
[Speaker 0]
not liking or following anything because they've learned
149
00:07:06,090 --> 00:07:08,570
[Speaker 0]
that the moment they engage, the feed reshapes
150
00:07:08,570 --> 00:07:10,889
[Speaker 0]
itself around them and gives them away. And
151
00:07:10,889 --> 00:07:12,410
[Speaker 0]
that's probably why they've had to, like, dig
152
00:07:12,410 --> 00:07:14,985
[Speaker 0]
into dwell time and stuff too because people
153
00:07:14,985 --> 00:07:17,945
[Speaker 0]
have figured out that engagement leads to, you
154
00:07:17,945 --> 00:07:22,264
[Speaker 0]
know, your feed being heavily affected by an
155
00:07:22,264 --> 00:07:26,760
[Speaker 0]
actual, you know, boost of engagement. But I
156
00:07:26,760 --> 00:07:28,440
[Speaker 0]
I think it's kind of scary, you know,
157
00:07:28,440 --> 00:07:31,000
[Speaker 0]
there's a bigger idea underneath this story. And
158
00:07:31,000 --> 00:07:34,520
[Speaker 0]
then what these algorithms really do are surfacing,
159
00:07:34,520 --> 00:07:36,840
[Speaker 0]
you know, our latent structure hidden in the
160
00:07:36,840 --> 00:07:39,560
[Speaker 0]
data. And when you collect enough, like, data
161
00:07:39,560 --> 00:07:41,800
[Speaker 0]
about a person, all those likes and follows
162
00:07:41,800 --> 00:07:44,775
[Speaker 0]
and Paul pauses, the information naturally arrange itself
163
00:07:44,775 --> 00:07:47,735
[Speaker 0]
into like a shape. And mathematicians call that
164
00:07:47,735 --> 00:07:50,375
[Speaker 0]
shape a manifold, which is just a fancy
165
00:07:50,375 --> 00:07:54,215
[Speaker 0]
word for hidden geometry, your behavior traces when
166
00:07:54,215 --> 00:07:58,490
[Speaker 0]
you when you map it. So, you know,
167
00:07:58,490 --> 00:08:00,090
[Speaker 0]
the geometry holds stuff about you that you
168
00:08:00,090 --> 00:08:02,810
[Speaker 0]
never explicitly typed in anywhere. You know, your
169
00:08:02,810 --> 00:08:04,330
[Speaker 0]
orientation is one thread of it, but it's
170
00:08:04,330 --> 00:08:06,650
[Speaker 0]
almost certainly not the only one. It could
171
00:08:06,650 --> 00:08:08,345
[Speaker 0]
be figuring other stuff stuff out about you.
172
00:08:08,345 --> 00:08:10,105
[Speaker 0]
You know, the same math probably encodes hints
173
00:08:10,105 --> 00:08:12,985
[Speaker 0]
about your upbringing, your mental health, you know,
174
00:08:12,985 --> 00:08:15,705
[Speaker 0]
lifestyle choices, you know, all kinds of things
175
00:08:15,705 --> 00:08:18,525
[Speaker 0]
you don't really consciously disclose to to people.
176
00:08:19,065 --> 00:08:20,425
[Speaker 0]
You know? So it's not like reading your
177
00:08:20,425 --> 00:08:22,430
[Speaker 0]
mind, but just taking in account the stuff
178
00:08:22,430 --> 00:08:24,910
[Speaker 0]
you're interested in. So that's that's scary in
179
00:08:24,910 --> 00:08:28,210
[Speaker 0]
some ways depending on how they're unpacking that.
180
00:08:28,990 --> 00:08:30,830
[Speaker 0]
One thing else always leads to another, you
181
00:08:30,830 --> 00:08:32,270
[Speaker 0]
know. It's like there's gotta be a push
182
00:08:32,270 --> 00:08:34,029
[Speaker 0]
behind that research. You know, this is probably
183
00:08:34,029 --> 00:08:36,130
[Speaker 0]
a place where that data is being sold.
184
00:08:36,430 --> 00:08:37,735
[Speaker 0]
So So they know what kind of ads
185
00:08:37,735 --> 00:08:42,695
[Speaker 0]
to sell you. Right? So who knows? Just
186
00:08:42,695 --> 00:08:45,995
[Speaker 0]
something to think about, but, yeah, very interesting,
187
00:08:46,455 --> 00:08:49,649
[Speaker 0]
feature piece that I found earlier today. Just
188
00:08:49,649 --> 00:08:51,910
[Speaker 0]
thought it was a pretty cool dig into
189
00:08:52,129 --> 00:08:56,050
[Speaker 0]
those lake structures. But before we hit the
190
00:08:56,050 --> 00:08:58,690
[Speaker 0]
headlines, a quick word from the company that
191
00:08:58,690 --> 00:09:01,509
[Speaker 0]
has kept this show online for two decades.
192
00:09:04,995 --> 00:09:06,755
[Speaker 0]
Here's something to think about. If you've been
193
00:09:06,755 --> 00:09:09,075
[Speaker 0]
meaning to grab a domain, start a website
194
00:09:09,075 --> 00:09:11,635
[Speaker 0]
or launch a podcast of your own, GoDaddy
195
00:09:11,635 --> 00:09:14,195
[Speaker 0]
makes it stupid easy to get started. We've
196
00:09:14,195 --> 00:09:15,795
[Speaker 0]
been using them at Geek News Central for
197
00:09:15,795 --> 00:09:17,635
[Speaker 0]
over twenty years, and we still spin up
198
00:09:17,635 --> 00:09:19,555
[Speaker 0]
new projects to them because the process is
199
00:09:19,555 --> 00:09:22,410
[Speaker 0]
just painless. Hosting starts at six ninety nine
200
00:09:22,410 --> 00:09:24,810
[Speaker 0]
a month. WordPress hosting is twelve ninety nine.
201
00:09:24,810 --> 00:09:26,970
[Speaker 0]
Both come with a free domain, professional email,
202
00:09:26,970 --> 00:09:29,370
[Speaker 0]
and an SSL certificate. If you just need
203
00:09:29,370 --> 00:09:31,370
[Speaker 0]
a domain, that's eleven ninety nine and you
204
00:09:31,370 --> 00:09:33,865
[Speaker 0]
are done. Go to the geek new central
205
00:09:33,865 --> 00:09:36,904
[Speaker 0]
dot com forward slash godaddy. You get the
206
00:09:36,904 --> 00:09:39,565
[Speaker 0]
best pricing and you're directly supporting this independent
207
00:09:39,625 --> 00:09:42,745
[Speaker 0]
show, geek new central dot com forward slash
208
00:09:42,745 --> 00:09:48,900
[Speaker 0]
godaddy. Alright, first headline for the evening, met
209
00:09:48,900 --> 00:09:51,940
[Speaker 0]
his blueprint for privacy aware AI infrastructure. And
210
00:09:51,940 --> 00:09:54,580
[Speaker 0]
this one comes from engineering. Fp dot com.
211
00:09:54,580 --> 00:09:58,115
[Speaker 0]
That was a little bit interesting. So staying,
212
00:09:58,115 --> 00:10:00,115
[Speaker 0]
you know, in theme on what machines can
213
00:10:00,115 --> 00:10:02,675
[Speaker 0]
figure out about your data, the engineering team
214
00:10:02,675 --> 00:10:04,595
[Speaker 0]
over at Meta put out a really sharp
215
00:10:04,595 --> 00:10:07,175
[Speaker 0]
piece on what they call privacy aware infrastructure.
216
00:10:08,035 --> 00:10:10,035
[Speaker 0]
The whole problem they're solving is that before
217
00:10:10,035 --> 00:10:12,035
[Speaker 0]
any privacy rule can protect a piece of
218
00:10:12,035 --> 00:10:14,810
[Speaker 0]
data, the system has to actually understand what
219
00:10:14,810 --> 00:10:17,130
[Speaker 0]
that data is. And that's harder than it
220
00:10:17,130 --> 00:10:19,690
[Speaker 0]
sounds. They use the example of a field
221
00:10:19,690 --> 00:10:22,990
[Speaker 0]
just labeled age. In one place, that age
222
00:10:23,050 --> 00:10:25,610
[Speaker 0]
describes a real human being and needs strict
223
00:10:25,610 --> 00:10:28,795
[Speaker 0]
protection. In another, it's just a cache setting,
224
00:10:28,795 --> 00:10:30,795
[Speaker 0]
how many seconds to keep a file around
225
00:10:30,795 --> 00:10:33,194
[Speaker 0]
before dumping it, which is meaningless from a
226
00:10:33,194 --> 00:10:37,774
[Speaker 0]
privacy standpoint. Same word, totally different stakes. So
227
00:10:38,154 --> 00:10:40,795
[Speaker 0]
Meta lays out four jobs their system has
228
00:10:40,795 --> 00:10:43,850
[Speaker 0]
to do, which is one, understand what data
229
00:10:43,850 --> 00:10:46,730
[Speaker 0]
exists and how it's governed, two, discover which
230
00:10:46,730 --> 00:10:49,450
[Speaker 0]
data flow actually matter for a given policy
231
00:10:49,450 --> 00:10:53,130
[Speaker 0]
question, three, enforce the rules on retention and
232
00:10:53,130 --> 00:10:56,745
[Speaker 0]
access and sharing, and then four proof compliance
233
00:10:56,745 --> 00:11:00,345
[Speaker 0]
with real evidence. And this is, here's the
234
00:11:00,345 --> 00:11:01,785
[Speaker 0]
part that I kind of loved about this
235
00:11:01,785 --> 00:11:05,785
[Speaker 0]
one. You know, I've been hoping to see
236
00:11:05,785 --> 00:11:09,310
[Speaker 0]
this whole industry move kind of towards this
237
00:11:09,310 --> 00:11:12,029
[Speaker 0]
direction, but they deploy the large language model
238
00:11:12,029 --> 00:11:14,829
[Speaker 0]
like really narrowly. So this means it's only
239
00:11:14,829 --> 00:11:18,689
[Speaker 0]
on genuinely ambiguous cases. And then they distill
240
00:11:18,750 --> 00:11:21,709
[Speaker 0]
what it learns into plain deterministic rules that
241
00:11:21,709 --> 00:11:24,645
[Speaker 0]
a human reviews. Over time, the AI does
242
00:11:24,645 --> 00:11:26,565
[Speaker 0]
less and less of the live work because
243
00:11:26,565 --> 00:11:29,285
[Speaker 0]
the fixed rules have taken over. So they
244
00:11:29,285 --> 00:11:32,885
[Speaker 0]
basically set up, you know, not a how
245
00:11:32,885 --> 00:11:34,325
[Speaker 0]
do you put it? It's not just a
246
00:11:34,325 --> 00:11:36,645
[Speaker 0]
one time fix, but they actually create the
247
00:11:36,645 --> 00:11:38,810
[Speaker 0]
solution and then the solution is able to
248
00:11:38,810 --> 00:11:43,769
[Speaker 0]
handle that from there on deter deterministically rather
249
00:11:43,769 --> 00:11:46,410
[Speaker 0]
than having an AI figure out the, you
250
00:11:46,410 --> 00:11:49,050
[Speaker 0]
know, ambiguous part to try to understand what's
251
00:11:49,050 --> 00:11:51,905
[Speaker 0]
going on. So as they build more of
252
00:11:51,905 --> 00:11:54,005
[Speaker 0]
these fixed rules, the AI will be useless.
253
00:11:55,025 --> 00:11:57,665
[Speaker 0]
So they're putting real numbers on it. In
254
00:11:57,665 --> 00:12:00,465
[Speaker 0]
their pipeline, the fixed rules already handle around
255
00:12:00,465 --> 00:12:02,945
[Speaker 0]
eighty five percent of the traffic and only
256
00:12:02,945 --> 00:12:04,865
[Speaker 0]
the last fifteen percent or so has to
257
00:12:04,865 --> 00:12:06,839
[Speaker 0]
fall back to the model. So that's a
258
00:12:06,839 --> 00:12:08,920
[Speaker 0]
pretty good thing because the model path cost
259
00:12:08,920 --> 00:12:12,360
[Speaker 0]
something like four hundred times more compute to
260
00:12:12,360 --> 00:12:15,260
[Speaker 0]
run. So the AI is the expensive specialist
261
00:12:15,399 --> 00:12:18,200
[Speaker 0]
you call in for the weird cases, not
262
00:12:18,200 --> 00:12:22,505
[Speaker 0]
the thing grinding through every routine decision. So
263
00:12:22,645 --> 00:12:24,645
[Speaker 0]
I really like this, you know, this concept
264
00:12:24,645 --> 00:12:27,465
[Speaker 0]
of using AI as an edge case solver
265
00:12:27,525 --> 00:12:29,625
[Speaker 0]
instead of a catch all. You know, we've
266
00:12:29,685 --> 00:12:32,325
[Speaker 0]
gotten to this theme of seeing, you know,
267
00:12:32,325 --> 00:12:36,425
[Speaker 0]
AI built in throughout the infrastructure of services.
268
00:12:37,149 --> 00:12:39,950
[Speaker 0]
And honestly, I feel like it makes them
269
00:12:39,950 --> 00:12:43,250
[Speaker 0]
worse. It genuinely makes them worse to use.
270
00:12:44,350 --> 00:12:47,310
[Speaker 0]
It makes the developers feel like they can
271
00:12:47,310 --> 00:12:50,430
[Speaker 0]
be less intuitive because there's some sort of
272
00:12:50,430 --> 00:12:52,430
[Speaker 0]
AI chatbot able to teach you how to
273
00:12:52,430 --> 00:12:55,045
[Speaker 0]
do it. One of the systems I deal
274
00:12:55,045 --> 00:13:01,865
[Speaker 0]
with at my, weekend job is just horribly
275
00:13:02,005 --> 00:13:06,890
[Speaker 0]
built horribly Laden just with AI infrastructure. And
276
00:13:06,910 --> 00:13:09,590
[Speaker 0]
it's ridiculous how much they rely on it.
277
00:13:09,590 --> 00:13:13,910
[Speaker 0]
I, I just run into new problems every
278
00:13:13,910 --> 00:13:16,550
[Speaker 0]
day that should not run, you know, that
279
00:13:16,550 --> 00:13:18,150
[Speaker 0]
should not occur in a pro in a
280
00:13:18,150 --> 00:13:22,255
[Speaker 0]
software that needs full auditability and transparency. And
281
00:13:24,155 --> 00:13:27,115
[Speaker 0]
it, it was really frustrating. So I don't
282
00:13:27,115 --> 00:13:30,155
[Speaker 0]
know. I do like this idea of taking
283
00:13:30,155 --> 00:13:34,475
[Speaker 0]
these, special cases and turning them into, you
284
00:13:34,475 --> 00:13:38,270
[Speaker 0]
know, deterministic rules that help to handle them,
285
00:13:38,330 --> 00:13:41,710
[Speaker 0]
you know, handle that specific case going forward.
286
00:13:43,290 --> 00:13:49,365
[Speaker 0]
So that's pretty cool. Up next, Alberta scans
287
00:13:49,365 --> 00:13:52,805
[Speaker 0]
four hundred sixty six million lines of code
288
00:13:52,805 --> 00:13:55,705
[Speaker 0]
with Claude. And this one came from Anthropic
289
00:13:55,765 --> 00:13:58,965
[Speaker 0]
themselves. This one ties kind of directly into
290
00:13:58,965 --> 00:14:01,385
[Speaker 0]
that same idea, but the government of Alberta
291
00:14:01,649 --> 00:14:04,529
[Speaker 0]
in Canada use Anthropic's Claude to go hunting
292
00:14:04,529 --> 00:14:07,250
[Speaker 0]
for security holes across its system, and the
293
00:14:07,250 --> 00:14:11,490
[Speaker 0]
numbers are wild. A team inside the Ministry
294
00:14:11,490 --> 00:14:14,930
[Speaker 0]
of Technology and Innovation scanned four hundred sixty
295
00:14:14,930 --> 00:14:18,315
[Speaker 0]
six million lines of code in about twenty
296
00:14:18,315 --> 00:14:22,714
[Speaker 0]
hours. Anthropic estimates the same review would have
297
00:14:22,714 --> 00:14:25,035
[Speaker 0]
taken a human team something like six and
298
00:14:25,035 --> 00:14:27,915
[Speaker 0]
a half years. And the way they pulled
299
00:14:27,915 --> 00:14:29,774
[Speaker 0]
it off is kind of the fun part.
300
00:14:29,870 --> 00:14:32,430
[Speaker 0]
They turned loose around fifty AI agents that
301
00:14:32,430 --> 00:14:35,389
[Speaker 0]
worked in parallel, basically their own automated red
302
00:14:35,389 --> 00:14:37,310
[Speaker 0]
team and blue team probing the systems at
303
00:14:37,310 --> 00:14:39,870
[Speaker 0]
the same time. Along the way, they worked
304
00:14:39,870 --> 00:14:42,745
[Speaker 0]
down years of what's called technical debt, which
305
00:14:42,745 --> 00:14:45,945
[Speaker 0]
is just a pile of outdated, messy half
306
00:14:45,945 --> 00:14:48,265
[Speaker 0]
patch code that builds up in any big
307
00:14:48,265 --> 00:14:52,845
[Speaker 0]
organization and quietly hides vulnerabilities inside it. So
308
00:14:52,985 --> 00:14:55,465
[Speaker 0]
this is a good version of AI in
309
00:14:55,465 --> 00:14:58,550
[Speaker 0]
production and notice this the same patterns, the
310
00:14:58,550 --> 00:15:00,790
[Speaker 0]
meta story AI used to clean up and
311
00:15:00,790 --> 00:15:03,750
[Speaker 0]
lock down a real system, not to run
312
00:15:03,750 --> 00:15:07,110
[Speaker 0]
the show unsupervised. So I wanna see a
313
00:15:07,110 --> 00:15:09,450
[Speaker 0]
lot more code bases get much more lean
314
00:15:09,645 --> 00:15:11,885
[Speaker 0]
And with that much more efficient, you know,
315
00:15:11,885 --> 00:15:16,125
[Speaker 0]
clearer maintainability behind every code base. And I
316
00:15:16,125 --> 00:15:17,245
[Speaker 0]
feel like that's going to be one of
317
00:15:17,245 --> 00:15:19,205
[Speaker 0]
the big things that these tools are going
318
00:15:19,205 --> 00:15:21,165
[Speaker 0]
to be able to provide us much more
319
00:15:21,165 --> 00:15:25,590
[Speaker 0]
efficient systems, especially in places that tend to
320
00:15:25,890 --> 00:15:30,130
[Speaker 0]
go underfunded such as, you know, specific government
321
00:15:30,130 --> 00:15:32,370
[Speaker 0]
utilities or whatever. You know, you don't you
322
00:15:32,370 --> 00:15:35,010
[Speaker 0]
don't know how often they're paying developers to
323
00:15:35,010 --> 00:15:38,385
[Speaker 0]
redesign systems, you know, for, you know, the
324
00:15:38,385 --> 00:15:41,585
[Speaker 0]
water ministry or whatever. I don't know, you
325
00:15:41,585 --> 00:15:44,405
[Speaker 0]
know, just to be roundabout about it. But
326
00:15:45,825 --> 00:15:47,925
[Speaker 0]
I think this is a pretty cool, application.
327
00:15:50,360 --> 00:15:54,680
[Speaker 0]
NVIDIA on reinforcement learning for AI agents. On
328
00:15:54,680 --> 00:15:57,320
[Speaker 0]
the AI building side, NVIDIA put a developer
329
00:15:57,320 --> 00:16:02,255
[Speaker 0]
piece on reinforcement learning for AI agents. Reinforcement
330
00:16:02,255 --> 00:16:04,215
[Speaker 0]
learning, if you haven't run into the term,
331
00:16:04,215 --> 00:16:06,215
[Speaker 0]
is the training approach where instead of showing
332
00:16:06,215 --> 00:16:08,615
[Speaker 0]
a model the right answer, you let it
333
00:16:08,615 --> 00:16:10,535
[Speaker 0]
try things and reward it when it does
334
00:16:10,535 --> 00:16:12,535
[Speaker 0]
well. Kind of like training a dog with
335
00:16:12,535 --> 00:16:15,790
[Speaker 0]
treats. NVIDIA's argument is that this is how
336
00:16:15,790 --> 00:16:17,950
[Speaker 0]
you take a general model and sharpen it
337
00:16:17,950 --> 00:16:21,310
[Speaker 0]
into a specialist for one specific business, where
338
00:16:21,310 --> 00:16:24,190
[Speaker 0]
the definition of a good answer is different
339
00:16:24,190 --> 00:16:27,230
[Speaker 0]
everywhere you go. So they walked through the
340
00:16:27,230 --> 00:16:30,270
[Speaker 0]
whole alphabet soup of techniques for doing it.
341
00:16:30,270 --> 00:16:34,005
[Speaker 0]
And one thing worth clarifying, because they're pretty
342
00:16:34,005 --> 00:16:36,005
[Speaker 0]
easy to mix up. There's a related tool
343
00:16:36,005 --> 00:16:40,245
[Speaker 0]
called RAG retrieval augmented generation, and NVIDIA treats
344
00:16:40,245 --> 00:16:42,245
[Speaker 0]
it as a or NVIDIA treats it as
345
00:16:42,245 --> 00:16:45,930
[Speaker 0]
a separate term from reinforcement learning. And that's
346
00:16:45,930 --> 00:16:48,090
[Speaker 0]
one where model looks up answers in your
347
00:16:48,090 --> 00:16:50,910
[Speaker 0]
database on the fly instead of being retrained.
348
00:16:50,970 --> 00:16:53,150
[Speaker 0]
So think of them as two different knobs.
349
00:16:53,290 --> 00:16:56,170
[Speaker 0]
Reinforcement learning changes how the model behaves, and
350
00:16:56,170 --> 00:16:59,210
[Speaker 0]
then rag changes what facts it can reach.
351
00:16:59,210 --> 00:17:01,395
[Speaker 0]
So if you're building agent systems, it's a
352
00:17:01,395 --> 00:17:04,375
[Speaker 0]
solid read on which knobs to turn when.
353
00:17:04,995 --> 00:17:08,695
[Speaker 0]
But pretty cool piece from nvidia dot com.
354
00:17:10,515 --> 00:17:15,070
[Speaker 0]
GitHub retires two Gemini models. Quick one, like,
355
00:17:15,070 --> 00:17:17,950
[Speaker 0]
just for a little model update. GitHub says
356
00:17:17,950 --> 00:17:20,909
[Speaker 0]
it's retiring two of Google's Gemini models, Gemini
357
00:17:20,909 --> 00:17:23,649
[Speaker 0]
two point five pro and Gemini three flash
358
00:17:24,190 --> 00:17:27,149
[Speaker 0]
across all of GitHub Copilot on July thirty
359
00:17:27,149 --> 00:17:29,775
[Speaker 0]
first. So if you're on them, the migration
360
00:17:29,775 --> 00:17:32,335
[Speaker 0]
paths are Gemini three point one pro and
361
00:17:32,335 --> 00:17:34,895
[Speaker 0]
three point five flash. The only reason I
362
00:17:34,895 --> 00:17:36,495
[Speaker 0]
brought this one up because it was, you
363
00:17:36,495 --> 00:17:39,855
[Speaker 0]
know, just say another surprise, you know, another
364
00:17:39,855 --> 00:17:42,290
[Speaker 0]
sign of the times. We're now at a
365
00:17:42,290 --> 00:17:45,030
[Speaker 0]
point where these tools that felt brand new
366
00:17:45,170 --> 00:17:48,690
[Speaker 0]
and awful, you know, and so awesome just
367
00:17:48,690 --> 00:17:50,690
[Speaker 0]
a couple of years ago are are already
368
00:17:50,690 --> 00:17:52,850
[Speaker 0]
getting sunset because, you know, the pace of
369
00:17:52,850 --> 00:17:56,475
[Speaker 0]
improvements is that fast. And it just makes
370
00:17:56,475 --> 00:17:58,475
[Speaker 0]
me wonder about all the hardware stamped with,
371
00:17:58,475 --> 00:18:02,475
[Speaker 0]
you know, an AI optimized label. You know,
372
00:18:02,475 --> 00:18:04,395
[Speaker 0]
a lot of those early chips are probably
373
00:18:04,395 --> 00:18:06,235
[Speaker 0]
gonna churn out just as quickly as the
374
00:18:06,235 --> 00:18:10,000
[Speaker 0]
models did. So, you know, I I really
375
00:18:10,000 --> 00:18:14,260
[Speaker 0]
wonder how this early AI investment will change
376
00:18:16,160 --> 00:18:18,660
[Speaker 0]
how companies are making their chips going forward
377
00:18:18,720 --> 00:18:21,360
[Speaker 0]
just because they have to continue to adapt
378
00:18:21,360 --> 00:18:25,794
[Speaker 0]
to the changing infrastructure. It's interesting just to
379
00:18:25,794 --> 00:18:28,595
[Speaker 0]
kinda like ponder as we see these two
380
00:18:28,595 --> 00:18:35,395
[Speaker 0]
models get retired from Copilot. Alright. Up next,
381
00:18:35,395 --> 00:18:39,880
[Speaker 0]
the many journeys of learning Rust. And this
382
00:18:39,880 --> 00:18:42,760
[Speaker 0]
one's for the programmers. And you guys all
383
00:18:42,760 --> 00:18:44,360
[Speaker 0]
know that I love rust. I talk about
384
00:18:44,360 --> 00:18:46,280
[Speaker 0]
it a lot on the show. Whenever I
385
00:18:46,280 --> 00:18:49,480
[Speaker 0]
host the breast blog, put out a pretty
386
00:18:49,480 --> 00:18:52,220
[Speaker 0]
cool post part of their vision doc series.
387
00:18:52,684 --> 00:18:55,165
[Speaker 0]
And it's all about how people actually learn
388
00:18:55,165 --> 00:18:58,525
[Speaker 0]
Rust. So for anybody who doesn't write code,
389
00:18:58,525 --> 00:19:01,325
[Speaker 0]
Rust is a programming language built around memory
390
00:19:01,325 --> 00:19:04,525
[Speaker 0]
safety. Its most famous feature is something called
391
00:19:04,525 --> 00:19:07,270
[Speaker 0]
a bar checker. And it's a strict part
392
00:19:07,270 --> 00:19:09,850
[Speaker 0]
of the compiler that polices how your data
393
00:19:09,910 --> 00:19:13,590
[Speaker 0]
gets used. The post is full of honest
394
00:19:13,590 --> 00:19:16,230
[Speaker 0]
stories, people wrestling with the bar checker until
395
00:19:16,230 --> 00:19:18,950
[Speaker 0]
it finally clicks, the idea they call clone
396
00:19:18,950 --> 00:19:21,925
[Speaker 0]
guilt, where beginners think good Rust means never
397
00:19:21,925 --> 00:19:24,405
[Speaker 0]
making a copy of anything, and they tie
398
00:19:24,405 --> 00:19:27,045
[Speaker 0]
themselves in knots over it. When seasoned Rust
399
00:19:27,045 --> 00:19:29,765
[Speaker 0]
folks say the opposite, make your copies without
400
00:19:29,765 --> 00:19:32,245
[Speaker 0]
stressing early on and tighten things up once
401
00:19:32,245 --> 00:19:34,980
[Speaker 0]
you actually understand the problem. In a really
402
00:19:34,980 --> 00:19:38,100
[Speaker 0]
interesting thread on silent attrition, the folks who
403
00:19:38,100 --> 00:19:40,580
[Speaker 0]
quietly bounce off the language and walk away
404
00:19:40,580 --> 00:19:43,140
[Speaker 0]
who are hard to study precisely because they
405
00:19:43,140 --> 00:19:47,300
[Speaker 0]
leave without saying anything. There's a whole section
406
00:19:47,300 --> 00:19:50,315
[Speaker 0]
on people using AI models as a learning
407
00:19:50,315 --> 00:19:53,035
[Speaker 0]
aid. One company saying it's training high school
408
00:19:53,035 --> 00:19:55,935
[Speaker 0]
grads into working Rust developers with that help.
409
00:19:56,555 --> 00:19:58,495
[Speaker 0]
I always tell you folks, getting your brain
410
00:19:58,555 --> 00:20:00,795
[Speaker 0]
onto a memory safe language rewires how you
411
00:20:00,795 --> 00:20:03,539
[Speaker 0]
think with, you know, about a problem. Rust
412
00:20:03,539 --> 00:20:05,539
[Speaker 0]
isn't right for every job. It really depends
413
00:20:05,539 --> 00:20:08,259
[Speaker 0]
on your use case, but it's absolutely paradigm
414
00:20:08,259 --> 00:20:10,419
[Speaker 0]
worth exploring, you know, just to understand how
415
00:20:10,419 --> 00:20:13,380
[Speaker 0]
it works. But a cool one from the
416
00:20:13,380 --> 00:20:17,865
[Speaker 0]
Rust blog, check it out, folks. Up next,
417
00:20:17,924 --> 00:20:22,005
[Speaker 0]
Google DeepMind partners with a twenty four. This
418
00:20:22,005 --> 00:20:25,625
[Speaker 0]
one comes from deep mind dot Google. But
419
00:20:25,684 --> 00:20:27,765
[Speaker 0]
it is a cool interest inclusion of world.
420
00:20:27,765 --> 00:20:29,625
[Speaker 0]
Maybe it was kind of bound to happen.
421
00:20:29,779 --> 00:20:32,659
[Speaker 0]
But Google DeepMind, the AI lab is teaming
422
00:20:32,659 --> 00:20:34,580
[Speaker 0]
up with a twenty four. And if you're
423
00:20:34,580 --> 00:20:37,299
[Speaker 0]
not a field person, a twenty four is
424
00:20:37,299 --> 00:20:41,220
[Speaker 0]
the studio behind movies like Hereditary and Everything
425
00:20:41,220 --> 00:20:43,865
[Speaker 0]
Everywhere All at Once. They've got a real
426
00:20:43,925 --> 00:20:47,445
[Speaker 0]
reputation for backing filmmakers. The two of them
427
00:20:47,445 --> 00:20:49,365
[Speaker 0]
are calling it a first of its kind
428
00:20:49,365 --> 00:20:53,125
[Speaker 0]
research partnership where DeepMind's researchers and a twenty
429
00:20:53,125 --> 00:20:55,285
[Speaker 0]
four work side by side to build and
430
00:20:55,285 --> 00:20:58,200
[Speaker 0]
test new creative tools shaped by the artists
431
00:20:58,200 --> 00:21:01,640
[Speaker 0]
who'd actually use them. And one detail kind
432
00:21:01,640 --> 00:21:04,520
[Speaker 0]
of worth adding, Google also made investment in
433
00:21:04,520 --> 00:21:06,520
[Speaker 0]
a twenty four. So this is money on
434
00:21:06,520 --> 00:21:10,525
[Speaker 0]
the table, not just a research handshake, but
435
00:21:10,525 --> 00:21:15,005
[Speaker 0]
they're being pretty deliberately vague. No named films,
436
00:21:15,005 --> 00:21:17,905
[Speaker 0]
no specific products, no release dates or nothing.
437
00:21:18,605 --> 00:21:22,145
[Speaker 0]
So I think, this is a pretty cool
438
00:21:22,890 --> 00:21:25,150
[Speaker 0]
look at, you know, the types of investments
439
00:21:25,290 --> 00:21:28,330
[Speaker 0]
that are gonna be very worthwhile in the
440
00:21:28,330 --> 00:21:31,950
[Speaker 0]
long run for Google, especially with successful filmmakers
441
00:21:32,090 --> 00:21:34,170
[Speaker 0]
or film studios like a twenty fifth, a
442
00:21:34,170 --> 00:21:36,670
[Speaker 0]
twenty four. Can I speak? A twenty four.
443
00:21:39,835 --> 00:21:42,955
[Speaker 0]
Up next, another one from Google, Google's million
444
00:21:42,955 --> 00:21:46,155
[Speaker 0]
dollar Africa indie game fund. So this one
445
00:21:46,155 --> 00:21:49,755
[Speaker 0]
came from, blog dot Google, and it's a
446
00:21:49,755 --> 00:21:52,554
[Speaker 0]
good news for game developers. If you're based
447
00:21:52,554 --> 00:21:55,730
[Speaker 0]
out of Africa, Google's launching an indie games
448
00:21:55,730 --> 00:21:59,030
[Speaker 0]
fund aimed at Sub Saharan Africa, a region
449
00:21:59,090 --> 00:22:01,490
[Speaker 0]
whose gaming scene is expanding about as fast
450
00:22:01,490 --> 00:22:04,450
[Speaker 0]
as anywhere on the planet. They're putting up
451
00:22:04,450 --> 00:22:07,590
[Speaker 0]
one million dollars split across ten local studios,
452
00:22:08,075 --> 00:22:11,035
[Speaker 0]
with each studio getting somewhere between fifty thousand
453
00:22:11,035 --> 00:22:13,915
[Speaker 0]
and two hundred thousand dollars plus mentorship and
454
00:22:13,915 --> 00:22:17,035
[Speaker 0]
hands on technical support. I think the applications
455
00:22:17,035 --> 00:22:19,515
[Speaker 0]
are now open. So if that's you or
456
00:22:19,515 --> 00:22:22,715
[Speaker 0]
somebody you know, is, you know, working at
457
00:22:22,715 --> 00:22:25,530
[Speaker 0]
a indie game studio in Africa, you know,
458
00:22:25,530 --> 00:22:28,510
[Speaker 0]
tell them that the applications closed at noon
459
00:22:28,650 --> 00:22:32,809
[Speaker 0]
on July thirty first. But that's cool, to
460
00:22:32,809 --> 00:22:34,650
[Speaker 0]
see that type of funding going out, you
461
00:22:34,650 --> 00:22:37,934
[Speaker 0]
know, across the world. But looking forward to
462
00:22:37,934 --> 00:22:40,255
[Speaker 0]
seeing what kind of games come from this
463
00:22:40,255 --> 00:22:48,495
[Speaker 0]
investment. WhatsApp app usernames are now here to
464
00:22:48,495 --> 00:22:52,240
[Speaker 0]
reserve. So WhatsApp is finally moving off of
465
00:22:52,240 --> 00:22:55,440
[Speaker 0]
phone numbers as your identity. They're rolling out
466
00:22:55,440 --> 00:22:58,000
[Speaker 0]
usernames, which means somebody can find you and
467
00:22:58,000 --> 00:23:00,879
[Speaker 0]
start a conversation without ever seeing your phone
468
00:23:00,879 --> 00:23:03,735
[Speaker 0]
number. Starting this week, you can reserve the
469
00:23:03,735 --> 00:23:05,735
[Speaker 0]
name you want ahead of the full launch
470
00:23:05,735 --> 00:23:08,855
[Speaker 0]
later this year. The privacy win here is
471
00:23:08,855 --> 00:23:11,255
[Speaker 0]
real. There's no directory to browse and no
472
00:23:11,255 --> 00:23:14,054
[Speaker 0]
suggestions. Somebody has to know your exact username
473
00:23:14,054 --> 00:23:15,975
[Speaker 0]
to reach you in, you know, the first
474
00:23:15,975 --> 00:23:18,240
[Speaker 0]
time. So if you want to grab yours
475
00:23:18,240 --> 00:23:21,680
[Speaker 0]
before it's taken, head into the settings, then
476
00:23:21,680 --> 00:23:25,860
[Speaker 0]
account, then username to reserve your WhatsApp username.
477
00:23:28,400 --> 00:23:31,220
[Speaker 0]
Up next is a little earnings announcement. I
478
00:23:31,360 --> 00:23:34,015
[Speaker 0]
just thought this was something to look forward
479
00:23:34,015 --> 00:23:36,915
[Speaker 0]
to that people might be interested in. Intel
480
00:23:37,135 --> 00:23:40,175
[Speaker 0]
is, or has announced that their Q2 earnings
481
00:23:40,175 --> 00:23:42,815
[Speaker 0]
date. And I'm just interested in to kind
482
00:23:42,815 --> 00:23:46,095
[Speaker 0]
of see how they're performing right now with
483
00:23:46,095 --> 00:23:49,480
[Speaker 0]
the chip infrastructure changing and with all the
484
00:23:49,480 --> 00:23:52,200
[Speaker 0]
investments being all over the place. You know,
485
00:23:52,200 --> 00:23:55,500
[Speaker 0]
we've seen a lot of us government investment
486
00:23:55,560 --> 00:23:59,000
[Speaker 0]
into Intel recently. So I am interested in
487
00:23:59,000 --> 00:24:03,165
[Speaker 0]
seeing, you know, their, their earnings, but they'll
488
00:24:03,165 --> 00:24:05,645
[Speaker 0]
report their second quarter results on July twenty
489
00:24:05,645 --> 00:24:08,605
[Speaker 0]
third, right after market close with a earnings
490
00:24:08,605 --> 00:24:12,205
[Speaker 0]
call at two pm Pacific. So keep an
491
00:24:12,205 --> 00:24:14,125
[Speaker 0]
ear to the ground on that one. It'll
492
00:24:14,125 --> 00:24:15,565
[Speaker 0]
tell us a lot about how they're holding
493
00:24:15,565 --> 00:24:19,040
[Speaker 0]
up in the middle of all this AI
494
00:24:19,100 --> 00:24:27,660
[Speaker 0]
infrastructure spending and their own restructuring. Up next,
495
00:24:27,660 --> 00:24:30,540
[Speaker 0]
a little health shift. Your smartwatch might spot
496
00:24:30,540 --> 00:24:33,755
[Speaker 0]
illness before you do Engadget has put out
497
00:24:33,755 --> 00:24:37,355
[Speaker 0]
a little piece on the wearables x AI
498
00:24:37,355 --> 00:24:39,695
[Speaker 0]
wave that's, you know, starting to actually see
499
00:24:39,995 --> 00:24:42,955
[Speaker 0]
see little results. The the core insight is
500
00:24:42,955 --> 00:24:47,100
[Speaker 0]
pretty simple. These devices are catching the moment
501
00:24:47,100 --> 00:24:49,760
[Speaker 0]
your body drifts off of your general baseline.
502
00:24:50,620 --> 00:24:52,860
[Speaker 0]
And it's they're saying that this type of
503
00:24:52,860 --> 00:24:55,100
[Speaker 0]
blip is often the first nudge to go
504
00:24:55,100 --> 00:24:57,580
[Speaker 0]
get yourself checked out. And there's a twenty
505
00:24:57,580 --> 00:24:59,820
[Speaker 0]
twenty five study out of Texas A and
506
00:24:59,820 --> 00:25:04,265
[Speaker 0]
M and Stanford suggesting smartwatches can pick up
507
00:25:04,265 --> 00:25:07,325
[Speaker 0]
early signs of COVID or the flu within
508
00:25:07,465 --> 00:25:11,945
[Speaker 0]
hours of infection. And Apple watches irregular rhythm
509
00:25:11,945 --> 00:25:15,360
[Speaker 0]
alerts have flagged the heart condition AFib correctly
510
00:25:15,500 --> 00:25:18,480
[Speaker 0]
around eighty four percent of the time. Wow.
511
00:25:18,860 --> 00:25:21,340
[Speaker 0]
So your body starts shifting in ways the
512
00:25:21,340 --> 00:25:25,100
[Speaker 0]
sensors notice well ahead of any symptom you'd
513
00:25:25,100 --> 00:25:28,605
[Speaker 0]
catch on your own. So this is another
514
00:25:28,605 --> 00:25:32,285
[Speaker 0]
one of those beginning of, you know, we're
515
00:25:32,285 --> 00:25:35,965
[Speaker 0]
starting to see a real actual strong use
516
00:25:35,965 --> 00:25:41,105
[Speaker 0]
case for, AI systems rather than having them
517
00:25:42,150 --> 00:25:46,390
[Speaker 0]
completely drive software systems that they don't need
518
00:25:46,390 --> 00:25:48,790
[Speaker 0]
to be driving. And I think this is
519
00:25:48,790 --> 00:25:51,690
[Speaker 0]
definitely a cool use. Hoping to see more
520
00:25:51,990 --> 00:25:57,305
[Speaker 0]
of this type of work come out. Here's
521
00:25:57,305 --> 00:26:01,225
[Speaker 0]
a interesting one from the Scientific American. They
522
00:26:01,225 --> 00:26:04,525
[Speaker 0]
ran a piece written by philosopher Henry Taylor
523
00:26:04,585 --> 00:26:07,705
[Speaker 0]
at the University of Birmingham on how working
524
00:26:07,705 --> 00:26:11,370
[Speaker 0]
memory might be tied to consciousness itself. Working
525
00:26:11,370 --> 00:26:13,770
[Speaker 0]
memory is the little scratch pad in your
526
00:26:13,770 --> 00:26:16,250
[Speaker 0]
head, and it's the information you're actually holding
527
00:26:16,250 --> 00:26:18,090
[Speaker 0]
on to to do whatever you're doing right
528
00:26:18,090 --> 00:26:21,130
[Speaker 0]
this second. Taylor opens with something we've all
529
00:26:21,130 --> 00:26:24,225
[Speaker 0]
felt, you know, the doorway effect, where you
530
00:26:24,225 --> 00:26:26,385
[Speaker 0]
walk into a room and instantly forget why
531
00:26:26,385 --> 00:26:29,025
[Speaker 0]
you came in. That blank moment is your
532
00:26:29,025 --> 00:26:31,985
[Speaker 0]
working memory clearing out. And the intriguing claim
533
00:26:31,985 --> 00:26:34,065
[Speaker 0]
is that when something drops out of working
534
00:26:34,065 --> 00:26:35,985
[Speaker 0]
memory, it seems to drop out of your
535
00:26:35,985 --> 00:26:39,570
[Speaker 0]
conscious awareness at the very same instant, which,
536
00:26:39,570 --> 00:26:41,250
[Speaker 0]
you know, hints that the two might be
537
00:26:41,250 --> 00:26:44,210
[Speaker 0]
deeply linked. So this idea is getting a
538
00:26:44,210 --> 00:26:48,210
[Speaker 0]
lot more intention recently across psychology, philosophy, and
539
00:26:48,210 --> 00:26:51,090
[Speaker 0]
neuroscience. And I think part of that is
540
00:26:51,090 --> 00:26:53,809
[Speaker 0]
the rise of AI making everybody ask fresh
541
00:26:53,809 --> 00:26:57,165
[Speaker 0]
questions about what memory and awareness even are.
542
00:27:00,184 --> 00:27:02,345
[Speaker 0]
So definitely worth a read. If you want
543
00:27:02,345 --> 00:27:09,080
[Speaker 0]
to hop into this, interesting rabbit hole. Now
544
00:27:09,080 --> 00:27:11,000
[Speaker 0]
this next one I thought was really, really
545
00:27:11,000 --> 00:27:17,660
[Speaker 0]
cool. Scuba diving, cyborg cockroaches. This one is
546
00:27:18,520 --> 00:27:21,340
[Speaker 0]
a little bit wild, but stay with me.
547
00:27:21,915 --> 00:27:24,395
[Speaker 0]
Nature's Daily Briefing covered this and I just
548
00:27:24,475 --> 00:27:27,115
[Speaker 0]
I really couldn't skip it. Scientists have built
549
00:27:27,115 --> 00:27:31,515
[Speaker 0]
tiny diving suits for Madagascar hissing cockroaches. And
550
00:27:31,515 --> 00:27:34,255
[Speaker 0]
in these suits, the roaches can survive underwater
551
00:27:34,475 --> 00:27:38,309
[Speaker 0]
for up to three hours. A regular roach
552
00:27:38,309 --> 00:27:42,470
[Speaker 0]
without the suit suffocates in minutes. So here's
553
00:27:42,470 --> 00:27:44,630
[Speaker 0]
how it works. The three d printed suit
554
00:27:44,630 --> 00:27:48,950
[Speaker 0]
runs tubes into the insect's breathing hole, which
555
00:27:48,950 --> 00:27:51,815
[Speaker 0]
are called spiracles. And those tubes connect to
556
00:27:51,815 --> 00:27:55,675
[Speaker 0]
a little chemical oxygen generator. Basically, a controlled
557
00:27:55,815 --> 00:27:59,675
[Speaker 0]
chemical reaction that makes breathable air, no electronics
558
00:27:59,815 --> 00:28:04,060
[Speaker 0]
involved. So this lab is already, you know,
559
00:28:04,060 --> 00:28:07,020
[Speaker 0]
had the steerable caught cyborg roach part figured
560
00:28:07,020 --> 00:28:09,740
[Speaker 0]
out. They're, they're guiding the bug around with
561
00:28:09,740 --> 00:28:13,740
[Speaker 0]
electrodes wired to its sensory organs. And the
562
00:28:13,740 --> 00:28:16,380
[Speaker 0]
diving suit is the brand new trick that
563
00:28:16,380 --> 00:28:18,655
[Speaker 0]
they've put on top of it. So the
564
00:28:18,655 --> 00:28:21,215
[Speaker 0]
work comes out of teams in Singapore and
565
00:28:21,215 --> 00:28:23,775
[Speaker 0]
Japan, and the researchers actually pitch it for
566
00:28:23,775 --> 00:28:26,175
[Speaker 0]
search and rescue, sending a bug into flooded
567
00:28:26,175 --> 00:28:30,995
[Speaker 0]
rubble where a regular robot can't fit. But
568
00:28:32,140 --> 00:28:34,380
[Speaker 0]
I gotta say the reality of the spy
569
00:28:34,380 --> 00:28:38,540
[Speaker 0]
bug has, has already arrived. It is here.
570
00:28:38,540 --> 00:28:41,660
[Speaker 0]
They've they've had these cyborg cockroaches. They were
571
00:28:41,660 --> 00:28:43,900
[Speaker 0]
able to control, you know, for a while
572
00:28:43,900 --> 00:28:46,780
[Speaker 0]
now. Now they can, you know, they could
573
00:28:46,780 --> 00:28:52,835
[Speaker 0]
swim. So definitely a really, interesting piece of
574
00:28:52,835 --> 00:28:57,794
[Speaker 0]
science coming out of nature dot com. Now
575
00:28:57,794 --> 00:28:59,554
[Speaker 0]
this next one's a little bit of a
576
00:28:59,554 --> 00:29:04,110
[Speaker 0]
brain bender Physicists at Los Alamos National Laboratory,
577
00:29:04,810 --> 00:29:08,190
[Speaker 0]
led by a researcher named Luis Pedro Garcia
578
00:29:08,490 --> 00:29:11,050
[Speaker 0]
Pintos have found a way to make a
579
00:29:11,050 --> 00:29:13,690
[Speaker 0]
quantum system look like it's running backward in
580
00:29:13,690 --> 00:29:18,534
[Speaker 0]
time. And it's a little bit sensationalized. But
581
00:29:18,534 --> 00:29:22,455
[Speaker 0]
time is not literally being reversed. Apparently, what
582
00:29:22,455 --> 00:29:25,174
[Speaker 0]
they can do is control quantum measurements so
583
00:29:25,174 --> 00:29:29,434
[Speaker 0]
precisely, that these systems evolution appears to unfold
584
00:29:29,654 --> 00:29:34,030
[Speaker 0]
in reverse statistically. So the really useful part
585
00:29:34,030 --> 00:29:36,670
[Speaker 0]
is the energy angle in their setup, the
586
00:29:36,670 --> 00:29:39,230
[Speaker 0]
act the act of measuring the system becomes
587
00:29:39,230 --> 00:29:41,790
[Speaker 0]
a resource you can actually pull work out
588
00:29:41,790 --> 00:29:45,410
[Speaker 0]
of something they call a continuous measurement engine
589
00:29:45,705 --> 00:29:48,365
[Speaker 0]
that harvests energy right out of the monitoring
590
00:29:48,505 --> 00:29:52,505
[Speaker 0]
process. Their papers in physical review x and
591
00:29:52,505 --> 00:29:55,544
[Speaker 0]
the potential payoffs they name are more powerful
592
00:29:55,544 --> 00:30:01,870
[Speaker 0]
quantum computers and quantum batteries. Yeah. So way
593
00:30:01,870 --> 00:30:04,610
[Speaker 0]
over most of our heads, mine included, but
594
00:30:06,110 --> 00:30:08,110
[Speaker 0]
it seems to be a breakthrough. Definitely an
595
00:30:08,110 --> 00:30:10,190
[Speaker 0]
exciting one. I just thought time was running
596
00:30:10,190 --> 00:30:14,525
[Speaker 0]
backwards and reading in the article, it got
597
00:30:14,585 --> 00:30:18,025
[Speaker 0]
more and more out of touch with my
598
00:30:18,025 --> 00:30:20,905
[Speaker 0]
reality the more I read. So if you
599
00:30:20,905 --> 00:30:23,545
[Speaker 0]
wanna read some mind bending quantum stuff, head
600
00:30:23,545 --> 00:30:25,545
[Speaker 0]
on over to science daily dot com to
601
00:30:25,545 --> 00:30:29,810
[Speaker 0]
check this one out. But really cool stuff.
602
00:30:30,990 --> 00:30:33,410
[Speaker 0]
Now this next one I found really interesting,
603
00:30:34,030 --> 00:30:36,850
[Speaker 0]
came out of science dot org, tall trees
604
00:30:36,990 --> 00:30:41,345
[Speaker 0]
shrug off drought. And it's one of those
605
00:30:41,345 --> 00:30:44,805
[Speaker 0]
studies that's overturning a bit of textbook wisdom.
606
00:30:45,025 --> 00:30:47,265
[Speaker 0]
So for a long time, the prevailing idea
607
00:30:47,265 --> 00:30:50,065
[Speaker 0]
was that taller trees should suffer more in
608
00:30:50,065 --> 00:30:52,385
[Speaker 0]
a drought, because they've got to haul water
609
00:30:52,385 --> 00:30:55,285
[Speaker 0]
so much higher against gravity, which makes sense.
610
00:30:56,110 --> 00:30:59,169
[Speaker 0]
But a new study in the journal Science
611
00:30:59,389 --> 00:31:04,690
[Speaker 0]
says, not so fast. Researchers looked at a
612
00:31:04,830 --> 00:31:10,365
[Speaker 0]
dipterocarps, dipterocarps, you know, the giant trees that
613
00:31:10,365 --> 00:31:13,805
[Speaker 0]
dominate the rainforests of Southeast Asia, some of
614
00:31:13,805 --> 00:31:16,285
[Speaker 0]
them topping seventy meters and during the big
615
00:31:16,285 --> 00:31:18,205
[Speaker 0]
El Nino drought of twenty twenty three and
616
00:31:18,205 --> 00:31:20,845
[Speaker 0]
twenty twenty four. The tall one slowed their
617
00:31:20,845 --> 00:31:22,925
[Speaker 0]
growth by about the same amount as the
618
00:31:22,925 --> 00:31:26,320
[Speaker 0]
short ones. So it turns out height wasn't
619
00:31:26,320 --> 00:31:30,160
[Speaker 0]
exactly the handicap that everyone had assumed. You
620
00:31:30,160 --> 00:31:32,080
[Speaker 0]
know, part of the trick is that a
621
00:31:32,080 --> 00:31:35,280
[Speaker 0]
seventy meter tree grows water carrying vessels at
622
00:31:35,280 --> 00:31:37,840
[Speaker 0]
its base that are roughly twice as wide
623
00:31:37,840 --> 00:31:41,255
[Speaker 0]
as those in a ten meter tree. Wow.
624
00:31:41,255 --> 00:31:43,255
[Speaker 0]
So you the plumbing, I guess keeps up
625
00:31:43,255 --> 00:31:46,135
[Speaker 0]
with what the tree needs. So it's so
626
00:31:46,135 --> 00:31:49,815
[Speaker 0]
much isn't a height issue. The researchers aren't
627
00:31:49,815 --> 00:31:51,815
[Speaker 0]
tearing down the old theory so much as
628
00:31:51,815 --> 00:31:54,135
[Speaker 0]
saying that the real driver of drought stress
629
00:31:54,135 --> 00:31:58,530
[Speaker 0]
is something more subtle than raw height. Maybe
630
00:31:58,530 --> 00:32:02,290
[Speaker 0]
the microclimate up in the canopy. So this
631
00:32:02,290 --> 00:32:04,210
[Speaker 0]
is pretty cool science, you know, chasing the
632
00:32:04,210 --> 00:32:07,830
[Speaker 0]
actual cause instead of letting the easy correlation
633
00:32:08,130 --> 00:32:11,065
[Speaker 0]
went out. You know, I guess we've not
634
00:32:11,065 --> 00:32:13,225
[Speaker 0]
challenged this assumption in a little while. So
635
00:32:13,225 --> 00:32:17,065
[Speaker 0]
definitely cool to see some research come out
636
00:32:17,065 --> 00:32:23,065
[Speaker 0]
that challenges some of these preconceived notions. Now
637
00:32:23,065 --> 00:32:25,305
[Speaker 0]
this next one's a little bit crazy. The
638
00:32:25,305 --> 00:32:30,340
[Speaker 0]
energy department, purged roughly six thousand web pages
639
00:32:30,800 --> 00:32:35,680
[Speaker 0]
about energy con conservation. And the timing is
640
00:32:35,680 --> 00:32:38,960
[Speaker 0]
pretty brutal. We've literally just talked about the
641
00:32:38,960 --> 00:32:42,254
[Speaker 0]
heat wave hitting New York and mayor Dom
642
00:32:42,254 --> 00:32:45,294
[Speaker 0]
Donnie's recommendations, but, you know, he they posted
643
00:32:45,294 --> 00:32:47,794
[Speaker 0]
out on x to, you know, manage your
644
00:32:48,335 --> 00:32:50,894
[Speaker 0]
electricity so that we didn't destroy the grid.
645
00:32:50,894 --> 00:32:55,149
[Speaker 0]
But, you know, it's crazy to kind of
646
00:32:55,149 --> 00:32:58,990
[Speaker 0]
see this follow-up from the government, you know,
647
00:32:58,990 --> 00:33:04,669
[Speaker 0]
maybe unrelated, but still insane timing to delete
648
00:33:04,669 --> 00:33:10,105
[Speaker 0]
these six thousand pages on energy conservation. So
649
00:33:12,005 --> 00:33:13,524
[Speaker 0]
it's really sad to kind of see it,
650
00:33:13,524 --> 00:33:19,365
[Speaker 0]
but luckily the pages were still saved by
651
00:33:19,365 --> 00:33:24,120
[Speaker 0]
our classic protectors, the Internet archive, and the
652
00:33:24,120 --> 00:33:28,680
[Speaker 0]
Wayback Machine. So I guess the Internet Archive
653
00:33:28,680 --> 00:33:32,840
[Speaker 0]
will eventually or I guess they got what
654
00:33:32,840 --> 00:33:35,480
[Speaker 0]
they could, but the Wayback Machine still has
655
00:33:35,480 --> 00:33:37,800
[Speaker 0]
all of it. So if you're looking at
656
00:33:37,800 --> 00:33:42,495
[Speaker 0]
see some destroyed energy department recommendations, I I
657
00:33:42,495 --> 00:33:45,615
[Speaker 0]
just don't understand why why we delete those
658
00:33:45,615 --> 00:33:47,955
[Speaker 0]
pages. It doesn't make a lot of sense.
659
00:33:50,175 --> 00:33:52,415
[Speaker 0]
We're gonna finish up the articles tonight in
660
00:33:52,415 --> 00:33:57,170
[Speaker 0]
space, but Webb's puzzling new universe from Quantum
661
00:33:57,170 --> 00:34:00,850
[Speaker 0]
Magazine. This is a pretty interesting one. Quantum
662
00:34:00,850 --> 00:34:02,690
[Speaker 0]
Magazine has a piece on how the James
663
00:34:02,690 --> 00:34:05,570
[Speaker 0]
Webb Space Telescope keeps finding things in the
664
00:34:05,570 --> 00:34:08,734
[Speaker 0]
early universe that by our best model simply
665
00:34:08,734 --> 00:34:11,535
[Speaker 0]
should not be there. Black holes that got
666
00:34:11,535 --> 00:34:15,375
[Speaker 0]
enormous way too fast, galaxies far brighter and
667
00:34:15,375 --> 00:34:18,494
[Speaker 0]
more massive than the timeline allows, and hundreds
668
00:34:18,494 --> 00:34:21,970
[Speaker 0]
of mysterious objects nicknamed little red dots showing
669
00:34:21,970 --> 00:34:24,630
[Speaker 0]
up around six hundred and fifty million years
670
00:34:24,690 --> 00:34:27,250
[Speaker 0]
after the Big Bang. The twist is a
671
00:34:27,250 --> 00:34:29,490
[Speaker 0]
fun one for scientists. The problem used to
672
00:34:29,490 --> 00:34:33,410
[Speaker 0]
be too few explanation, and now it's the
673
00:34:33,410 --> 00:34:35,735
[Speaker 0]
opposite. And let me give you a little
674
00:34:35,735 --> 00:34:39,915
[Speaker 0]
quote from astrophysicist Rachel Somerville at the Flatiron
675
00:34:40,135 --> 00:34:43,975
[Speaker 0]
Institute. We almost have gone from having too
676
00:34:43,975 --> 00:34:47,035
[Speaker 0]
many early galaxies to having too many theories,
677
00:34:47,175 --> 00:34:50,940
[Speaker 0]
Rachel Somerville. So the hard work now is
678
00:34:50,940 --> 00:34:53,820
[Speaker 0]
figuring out which of these actually competing theories
679
00:34:53,820 --> 00:34:57,180
[Speaker 0]
is right. You know? So definitely a really
680
00:34:57,180 --> 00:35:00,380
[Speaker 0]
exciting time to be diving into this type
681
00:35:00,380 --> 00:35:02,700
[Speaker 0]
of science, but I'm excited to kind of
682
00:35:02,700 --> 00:35:05,414
[Speaker 0]
see more results come out, you know, in
683
00:35:05,414 --> 00:35:07,734
[Speaker 0]
the future, but nice to see that we
684
00:35:07,734 --> 00:35:12,694
[Speaker 0]
have gone from clueless to completely exploring, you
685
00:35:12,694 --> 00:35:15,174
[Speaker 0]
know, all these different types of theories, all
686
00:35:15,174 --> 00:35:18,820
[Speaker 0]
about the early universe, Kind of understand how
687
00:35:18,820 --> 00:35:24,520
[Speaker 0]
everything came to be. NASA's emergency telescope rescue.
688
00:35:24,740 --> 00:35:28,180
[Speaker 0]
This one came out of the verge. And
689
00:35:28,180 --> 00:35:30,820
[Speaker 0]
another story on a space telescope, but NASA's
690
00:35:30,820 --> 00:35:34,225
[Speaker 0]
got a rescue mission underway. The Swift Observatory,
691
00:35:34,525 --> 00:35:36,525
[Speaker 0]
which launched back in two thousand four and
692
00:35:36,525 --> 00:35:40,465
[Speaker 0]
studies gamma ray bursts, the incredibly bright explosions
693
00:35:40,525 --> 00:35:42,525
[Speaker 0]
that help us peer back into the air
694
00:35:42,685 --> 00:35:46,410
[Speaker 0]
early universe is in trouble. A recent run
695
00:35:46,410 --> 00:35:49,609
[Speaker 0]
of solar storms, solar storms puffed up Earth's
696
00:35:49,609 --> 00:35:52,650
[Speaker 0]
atmosphere, and that extra drag has been dragging
697
00:35:52,650 --> 00:35:56,089
[Speaker 0]
Swift's orbit down. It's currently circling about two
698
00:35:56,089 --> 00:35:59,325
[Speaker 0]
twenty four miles up. And if nothing enough,
699
00:35:59,325 --> 00:36:01,165
[Speaker 0]
nothing's done, it could burn up in the
700
00:36:01,165 --> 00:36:04,445
[Speaker 0]
atmosphere as soon as this year. So NASA
701
00:36:04,445 --> 00:36:07,744
[Speaker 0]
brought in a company called catalyst space technologies
702
00:36:08,285 --> 00:36:11,405
[Speaker 0]
out of Flagstaff, Arizona, and their spacecraft called
703
00:36:11,405 --> 00:36:15,060
[Speaker 0]
link launched this past Friday. The plan is
704
00:36:15,060 --> 00:36:17,460
[Speaker 0]
to catch up swift and boost it back
705
00:36:17,460 --> 00:36:19,720
[Speaker 0]
up to around three hundred seventy three miles,
706
00:36:20,020 --> 00:36:22,180
[Speaker 0]
roughly about one hundred and fifty miles higher
707
00:36:22,180 --> 00:36:25,460
[Speaker 0]
than it is now, which is much safer
708
00:36:25,460 --> 00:36:30,345
[Speaker 0]
than burning up in our atmosphere. So definitely
709
00:36:30,345 --> 00:36:33,545
[Speaker 0]
a genuinely a little, you know, dramatic little
710
00:36:33,545 --> 00:36:35,625
[Speaker 0]
save happening right over our heads, but it's
711
00:36:35,625 --> 00:36:40,505
[Speaker 0]
for science. The last one is just a
712
00:36:40,505 --> 00:36:45,120
[Speaker 0]
gorgeous picture out of ESA. They brought a
713
00:36:45,180 --> 00:36:49,040
[Speaker 0]
wonderful image of a gorgeous Aurora from orbit,
714
00:36:49,660 --> 00:36:53,340
[Speaker 0]
and it's, it's just a fantastic picture. If
715
00:36:53,340 --> 00:36:54,460
[Speaker 0]
you have a little bit of time to
716
00:36:54,460 --> 00:36:55,820
[Speaker 0]
check it out, go and check out the
717
00:36:55,820 --> 00:36:59,515
[Speaker 0]
page. But, you know, I love to see
718
00:36:59,515 --> 00:37:01,595
[Speaker 0]
more and more of these amazing photos of
719
00:37:01,595 --> 00:37:03,835
[Speaker 0]
the planet or if deep space come out
720
00:37:03,835 --> 00:37:06,335
[Speaker 0]
just keeps us wondering what's really out there.
721
00:37:08,155 --> 00:37:10,335
[Speaker 0]
Well, that's all I got for headlines tonight.
722
00:37:11,130 --> 00:37:13,369
[Speaker 0]
A little bit of housekeeping, but become a
723
00:37:13,369 --> 00:37:16,250
[Speaker 0]
GNC insider at geek news central dot com
724
00:37:16,250 --> 00:37:19,210
[Speaker 0]
forward slash insider. If you want to email
725
00:37:19,210 --> 00:37:20,890
[Speaker 0]
us about the show, hit us up at
726
00:37:20,890 --> 00:37:24,785
[Speaker 0]
geek newsgmail dot com. You're gonna want to
727
00:37:24,785 --> 00:37:27,424
[Speaker 0]
subscribe via your favorite podcast app and sign
728
00:37:27,424 --> 00:37:29,424
[Speaker 0]
up for the newsletter at geek news central
729
00:37:29,424 --> 00:37:32,464
[Speaker 0]
dot com. And grab yourself a modern podcast
730
00:37:32,464 --> 00:37:35,025
[Speaker 0]
app at podcast apps dot com. There are
731
00:37:35,025 --> 00:37:37,184
[Speaker 0]
some pretty cool ones out there. I use
732
00:37:37,184 --> 00:37:39,860
[Speaker 0]
pocket casts. I usually forget the name, but
733
00:37:39,860 --> 00:37:44,580
[Speaker 0]
I remember it now. Finally, big thanks to
734
00:37:44,580 --> 00:37:46,420
[Speaker 0]
GoDaddy for over twenty years of keeping this
735
00:37:46,420 --> 00:37:48,740
[Speaker 0]
show on the air. Support us at geek
736
00:37:48,740 --> 00:37:52,927
[Speaker 0]
new central dot com forward slash GoDaddy. Every
737
00:37:52,927 --> 00:37:54,847
[Speaker 0]
click counts. Thank you guys so much for