Arguing about AI with Billy Hollis

WEBVTT
1
00:00:01.199 --> 00:00:04.799
How'd you like to listen to. NET Rocks with no ads? Easy.
2
00:00:05.410 --> 00:00:08.810
Become a patron. For just $ 5 a month, you get
3
00:00:08.910 --> 00:00:11.750
access to a private RSS feed where all the shows
4
00:00:11.810 --> 00:00:14.929
have no ads. $ 20 a month will get you that
5
00:00:15.130 --> 00:00:18.870
and a special. NET Rocks patron mug. Sign up now
6
00:00:18.929 --> 00:00:37.259
at patreon.dotnetrocks.com. NET Rocks Hey, and welcome back to. NET Rocks.
7
00:00:38.280 --> 00:00:39.079
I'm Carl Franklin.
8
00:00:39.320 --> 00:00:40.159
And I'm Richard Campbell.
9
00:00:40.359 --> 00:00:41.439
Billy Hollis is here.
10
00:00:41.460 --> 00:00:41.640
Yay.
11
00:00:41.740 --> 00:00:44.960
He's back with us. And he's got something to say.
12
00:00:45.060 --> 00:00:45.520
He always does.
13
00:00:45.579 --> 00:00:47.500
I know he does. Richard, how are you?
14
00:00:47.840 --> 00:00:50.759
I'm good. You know, summertime. Been home for a few
15
00:00:50.799 --> 00:00:53.020
weeks for a change. You know, trying to enjoy being
16
00:00:53.079 --> 00:00:55.299
up on the coast, living by the ocean. Yeah. A
17
00:00:55.479 --> 00:00:58.780
little smoky, of course, because, you know. British Columbia is
18
00:00:58.840 --> 00:01:01.219
on fire. And it's been a bad one this year.
19
00:01:01.240 --> 00:01:04.879
So there's folks who've lost their homes. There's lots of displaced.
20
00:01:05.079 --> 00:01:06.760
It's been tough.
21
00:01:07.040 --> 00:01:07.519
That's bad.
22
00:01:07.680 --> 00:01:08.060
It's bad.
23
00:01:08.250 --> 00:01:10.010
Probably not as bad as all the things I'm about
24
00:01:10.030 --> 00:01:10.810
to list about 2017.
25
00:01:10.810 --> 00:01:11.269
No, geez.
26
00:01:11.290 --> 00:01:16.849
When we start. If you've been paying attention, I don't know,
27
00:01:16.870 --> 00:01:19.250
for the last 50 shows or so, Richard and I
28
00:01:19.299 --> 00:01:24.060
have been doing what happens in the year of the
29
00:01:24.120 --> 00:01:24.700
show number.
30
00:01:25.099 --> 00:01:25.280
Yeah.
31
00:01:25.549 --> 00:01:28.859
I think it started like in episode 1914. I'm like, hey, look,
32
00:01:28.879 --> 00:01:30.780
World War I, right? And then it got out of
33
00:01:30.799 --> 00:01:31.299
hand from there.
34
00:01:31.359 --> 00:01:33.640
But also, you know, there were important history lessons to
35
00:01:33.680 --> 00:01:36.159
learn because in some ways we saw some of these
36
00:01:36.200 --> 00:01:37.239
things repeating themselves.
37
00:01:37.819 --> 00:01:38.060
Sure.
38
00:01:38.120 --> 00:01:39.219
Or at least the danger of that.
39
00:01:39.879 --> 00:01:40.060
Yeah.
40
00:01:40.299 --> 00:01:40.579
All right.
41
00:01:40.659 --> 00:01:44.400
So, number one, Donald Trump begins his presidency.
42
00:01:44.980 --> 00:01:45.340
First time.
43
00:01:45.609 --> 00:01:49.569
Yeah, first time. Me Too and the Harvey Weinstein scandal.
44
00:01:49.890 --> 00:01:51.349
Yeah, that kind of cemented that.
45
00:01:51.450 --> 00:01:56.769
So, my stepdaughter... Who's actually visiting with her husband and
46
00:01:56.829 --> 00:01:58.329
daughter this week.
47
00:01:58.590 --> 00:01:59.269
Your grandbaby.
48
00:01:59.670 --> 00:02:03.450
Worked for said Harvey Weinstein in New York.
49
00:02:03.609 --> 00:02:04.230
Oh, no kidding.
50
00:02:04.329 --> 00:02:09.099
For a while and left before all this stuff came out. Wow.
51
00:02:09.300 --> 00:02:10.500
She was not affected by him.
52
00:02:10.699 --> 00:02:10.860
Good.
53
00:02:10.879 --> 00:02:13.340
Let's put it that way. Lucky. Just kind of weird.
54
00:02:14.259 --> 00:02:19.020
Hurricanes Harvey, Irma, and Maria. Yeah. It's an exceptionally destructive
55
00:02:19.060 --> 00:02:20.340
Atlantic hurricane season.
56
00:02:20.370 --> 00:02:22.409
Yeah.
57
00:02:22.900 --> 00:02:24.340
Las Vegas mass shooting.
58
00:02:24.530 --> 00:02:24.759
Yeah.
59
00:02:24.870 --> 00:02:29.590
Which has become like not news anymore. Unfortunately, uh, Russian
60
00:02:29.610 --> 00:02:35.090
investigation and Robert Mueller, uh, the North Korean nuclear crisis, which.
61
00:02:35.389 --> 00:02:37.550
Hasn't gone away, but really hasn't gone anywhere either.
62
00:02:37.810 --> 00:02:42.009
Hasn't gone anywhere either. It's contained for now. Uh, the
63
00:02:42.110 --> 00:02:44.389
Charlottesville white nationalist rally.
64
00:02:44.569 --> 00:02:45.210
Yeah. It wasn't good.
65
00:02:45.289 --> 00:02:49.810
August unite the right rally in Charlottesville, Virginia erupted into violence.
66
00:02:50.189 --> 00:02:57.870
It's not good. Um, The Rohingya refugee crisis in Myanmar.
67
00:02:58.840 --> 00:03:04.159
Hundreds of thousands of Rohingya Muslims fled for Bangladesh following
68
00:03:04.180 --> 00:03:09.819
a brutal military crackdown. Brexit formally gets underway. Brexit, the
69
00:03:09.860 --> 00:03:12.379
best idea the Brits have ever had. Oh, geez.
70
00:03:12.539 --> 00:03:17.080
Well, oddly enough, I was in London for the official
71
00:03:17.759 --> 00:03:19.300
Brexit exit.
72
00:03:19.580 --> 00:03:20.509
Okay. Wow. I.
73
00:03:21.919 --> 00:03:26.930
signed up to speak at NDC London, which is in
74
00:03:26.969 --> 00:03:30.449
the Q. You guys have been there, I think. It's
75
00:03:30.550 --> 00:03:34.030
right there in Westminster, right there next to where everything happened.
76
00:03:34.550 --> 00:03:36.250
And I actually went down.
77
00:03:36.289 --> 00:03:37.590
Right where Protest Corner is.
78
00:03:37.870 --> 00:03:40.750
Yeah. And I mean, I had a great hotel, not
79
00:03:40.770 --> 00:03:44.800
very far because I got that before the Brexit celebration
80
00:03:45.400 --> 00:03:48.039
was scheduled. So I got to, I wandered around, took
81
00:03:48.080 --> 00:03:51.780
videos and pictures of That's probably the most historic thing
82
00:03:51.819 --> 00:03:53.419
I've ever managed to get in the middle of.
83
00:03:53.479 --> 00:03:54.120
Yeah, no kidding.
84
00:03:54.300 --> 00:03:54.560
Wow.
85
00:03:54.639 --> 00:03:56.860
You should go back to NBC London and watch now
86
00:03:56.939 --> 00:03:59.379
because what's protesting on that corner is we should rejoin
87
00:03:59.419 --> 00:03:59.879
the EU.
88
00:04:00.340 --> 00:04:03.280
Yeah. Probably not the same people.
89
00:04:03.919 --> 00:04:05.280
No, I'm suspecting.
90
00:04:05.620 --> 00:04:08.919
It was quite an emotional decision, if I could say that.
91
00:04:10.840 --> 00:04:11.060
Yeah.
92
00:04:11.719 --> 00:04:18.269
Okay. And the number 10 story is Catalonia independence. uh,
93
00:04:18.379 --> 00:04:21.939
Catalonia had held a disputed independence referendum, October 1st, despite
94
00:04:21.959 --> 00:04:27.759
opposition from Spain's government and courts, Catalan leaders subsequently declared independence,
95
00:04:29.000 --> 00:04:33.019
prompting Madrid to dissolve regional government and impose direct rule.
96
00:04:34.220 --> 00:04:35.439
See what I did there? Independence.
97
00:04:36.259 --> 00:04:38.009
Yeah. Yeah. Okay.
98
00:04:39.050 --> 00:04:43.149
That's the way they pronounce their asses. Anyway, Richard, tell
99
00:04:43.209 --> 00:04:45.949
us about what happened in 2017 in space and tech.
100
00:04:46.529 --> 00:04:49.459
All right. Uh, this is the year 2017 is the
101
00:04:49.490 --> 00:04:54.949
year that commercial cargo to the space station exceeds government
102
00:04:55.689 --> 00:04:58.389
transport to the station. So of course, the U S
103
00:04:58.430 --> 00:05:00.310
only ever supplied the station with the space shuttle, which
104
00:05:00.350 --> 00:05:03.529
hadn't flown since last flight in 2011. So that had
105
00:05:03.550 --> 00:05:05.470
been six years from that. So they weren't doing any supply,
106
00:05:05.509 --> 00:05:09.290
but the Japanese have vehicle, the HTV for a while,
107
00:05:09.350 --> 00:05:12.370
the Europeans flew a vehicle, the ATV, and of course
108
00:05:12.550 --> 00:05:14.350
the Russians with their progress boosters.
109
00:05:14.689 --> 00:05:14.870
Yeah.
110
00:05:15.350 --> 00:05:19.189
And so, And progress mostly have been sustaining the station,
111
00:05:19.509 --> 00:05:22.990
including bringing all people to it with the Soyuz for
112
00:05:23.029 --> 00:05:25.689
many years and still are. We're still not getting to
113
00:05:25.889 --> 00:05:32.060
commercial crew. But in 2017, SpaceX flies the station four
114
00:05:32.120 --> 00:05:35.259
times with Cargo Dragon and Orbital ATK, which is now
115
00:05:35.279 --> 00:05:39.259
part of Northrop Grumman, flew twice with their vehicle, the
116
00:05:39.300 --> 00:05:42.569
Cygnus vehicle. So that's six compared to three Progress and
117
00:05:42.689 --> 00:05:46.160
one HTV. So the amount of resupply of the stations
118
00:05:46.199 --> 00:05:48.480
remains the same. It's eight to nine missions a year.
119
00:05:49.079 --> 00:05:52.449
But in 2017 is when commercial represents the bulk of it.
120
00:05:52.550 --> 00:05:55.970
And what's the advantage there? It costs less. It costs
121
00:05:56.050 --> 00:06:00.649
a lot less. And so, you know, it'll always be
122
00:06:00.689 --> 00:06:02.050
that way going forward after that.
123
00:06:02.589 --> 00:06:05.170
It's pretty amazing to look at the nine years and
124
00:06:05.350 --> 00:06:10.500
how dramatically different the entire– wait to orbit and cost factors,
125
00:06:10.540 --> 00:06:11.220
et cetera, are.
126
00:06:11.259 --> 00:06:13.040
Well, and the miracle, you know, these are the years
127
00:06:13.060 --> 00:06:15.000
of the miracles because this is also the year where
128
00:06:15.439 --> 00:06:19.579
in March, SpaceX reflies a Falcon 9 booster for the
129
00:06:19.620 --> 00:06:22.439
first time. The one that landed the year before, right?
130
00:06:22.500 --> 00:06:24.639
This is just when this is happening. We take it
131
00:06:24.680 --> 00:06:29.620
for granted now. I don't. It's magic every time, but
132
00:06:29.660 --> 00:06:31.139
it happens a hundred plus times a year.
133
00:06:31.160 --> 00:06:31.939
Yeah.
134
00:06:32.000 --> 00:06:36.610
Uh, 2017, also the year of that great eclipse across
135
00:06:36.670 --> 00:06:38.980
the U.S. from Oregon to South Carolina.
136
00:06:39.250 --> 00:06:39.810
Oh, yes.
137
00:06:39.949 --> 00:06:40.620
I've got pictures.
138
00:06:40.810 --> 00:06:41.060
Yeah.
139
00:06:41.310 --> 00:06:43.360
From my backyard. I didn't even have to travel.
140
00:06:43.379 --> 00:06:45.839
Didn't have to. Well, and we saw it here, too.
141
00:06:45.879 --> 00:06:47.220
In fact, I was on the coast when it happened.
142
00:06:47.259 --> 00:06:49.600
And I did the colander trick where I just pulled
143
00:06:49.639 --> 00:06:52.139
a colander out and shone it on the ground. And so,
144
00:06:52.199 --> 00:06:54.920
I saw several. Several hundred little eclipses?
145
00:06:54.939 --> 00:06:55.899
Little eclipses, yeah.
146
00:06:55.959 --> 00:06:56.279
It's cool.
147
00:06:56.360 --> 00:06:58.819
And the trees. I have lots of trees in the yard. Yeah.
148
00:06:58.879 --> 00:07:00.800
And so, you could just take a picture on Broadway
149
00:07:00.819 --> 00:07:02.100
in these beautiful crescents.
150
00:07:02.339 --> 00:07:02.920
Yeah.
151
00:07:02.959 --> 00:07:04.740
You wimps. I got the glasses.
152
00:07:05.579 --> 00:07:08.230
Too much work. They did. Too much work.
153
00:07:08.250 --> 00:07:10.709
They did work. I put them on. I couldn't see anything.
154
00:07:10.750 --> 00:07:14.110
I'm like, these things are useless. And then I looked
155
00:07:14.149 --> 00:07:18.069
at the sun. I'm like, oh, okay. They're supposed to
156
00:07:18.089 --> 00:07:19.790
do that.
157
00:07:19.810 --> 00:07:24.189
Two more space stories here. In September– After 13 years
158
00:07:24.250 --> 00:07:28.509
of service, Cassini is deorbited into Saturn. The concern, this
159
00:07:28.550 --> 00:07:30.589
is a huge vehicle, one of the largest ever flown.
160
00:07:31.730 --> 00:07:34.279
Size of a school bus, right? And they don't want
161
00:07:34.300 --> 00:07:36.839
to take the chance with contaminating any moon like Enceladus
162
00:07:36.899 --> 00:07:41.040
or Titan. And so they deliberately fly it into Saturn
163
00:07:41.319 --> 00:07:43.800
and manage to get a couple of pictures of it
164
00:07:43.899 --> 00:07:45.959
as it goes through the first cloud layer. It takes
165
00:07:46.019 --> 00:07:48.639
a few pictures. So getting really close to Saturn, you
166
00:07:48.660 --> 00:07:50.819
can see the shape of it before it's destroyed.
167
00:07:50.899 --> 00:07:51.360
Very cool.
168
00:07:51.379 --> 00:07:54.459
A little sad, but also it's the responsible ending. It's
169
00:07:54.500 --> 00:07:58.449
what we do now is you don't leave junk floating around.
170
00:07:58.550 --> 00:08:02.310
Every time I hear about Cassini, I think of Scott
171
00:08:02.329 --> 00:08:09.759
Guthrie naming the in-memory– Web server in Visual Studio, Cassini.
172
00:08:10.019 --> 00:08:12.439
Right. I always think of arguing with Rob Conner about
173
00:08:12.459 --> 00:08:14.439
the terrible ending to his book, but that's, you know,
174
00:08:14.459 --> 00:08:20.110
apparently I'm the only one who hates that ending. To
175
00:08:20.129 --> 00:08:21.790
the point where he says, listen, if I tell you
176
00:08:21.829 --> 00:08:23.189
I'm not going to fix this, is this going to
177
00:08:23.209 --> 00:08:26.879
break our friendship? No, no, I'm okay with you being wrong, but,
178
00:08:26.920 --> 00:08:28.459
you know, we'll see what we have to do.
179
00:08:28.480 --> 00:08:30.850
Just next time you write a book. Send me the
180
00:08:30.889 --> 00:08:32.159
copy before you publish it.
181
00:08:32.669 --> 00:08:34.580
He hadn't published it yet, but he wasn't changing it.
182
00:08:34.629 --> 00:08:37.970
He liked his ending. He was just wrong. Last but
183
00:08:37.990 --> 00:08:40.149
not least, Oumuamua. Do you remember this?
184
00:08:40.870 --> 00:08:43.429
You know, I usually have that with gravy and some
185
00:08:43.450 --> 00:08:45.289
potatoes in Hawaii.
186
00:08:45.309 --> 00:08:49.730
Oumuamua was the first time we detected an interstellar object. So,
187
00:08:49.809 --> 00:08:53.909
this is an asteroid that came through the solar system
188
00:08:54.090 --> 00:08:56.110
at a sufficient velocity that came from outside the solar
189
00:08:56.129 --> 00:08:59.929
system and exited again. We didn't detect it until it
190
00:08:59.950 --> 00:09:03.750
had already passed around the sun in September. We saw
191
00:09:03.769 --> 00:09:07.659
it in October. It was detected by a land-based telescope
192
00:09:07.720 --> 00:09:12.299
called Pan-STARRS, which is specifically designed to find near-Earth asteroids,
193
00:09:12.659 --> 00:09:16.500
which is why we detected this one. We're sure there's plenty.
194
00:09:16.580 --> 00:09:20.019
We've seen more since. It's just that they're hard to see.
195
00:09:20.179 --> 00:09:23.299
They move fast and they're on funny planes. When we think,
196
00:09:23.379 --> 00:09:25.289
when we're looking for near earth asteroids, cause we don't
197
00:09:25.330 --> 00:09:27.289
want to go the way of the dinosaurs, right? Like
198
00:09:27.330 --> 00:09:30.169
dinosaurs are gone cause they didn't have a space program. Uh,
199
00:09:30.450 --> 00:09:32.250
we are looking along the plane of the ecliptic where
200
00:09:32.289 --> 00:09:35.720
we're at the highest risk of collision. And when Amumuah
201
00:09:35.759 --> 00:09:38.259
passed that roughly in that direction, we were able to
202
00:09:38.299 --> 00:09:41.889
see it there. Uh, Lots of interesting conversation around this.
203
00:09:42.009 --> 00:09:45.970
It appeared to accelerate as it moved away from the sun,
204
00:09:46.250 --> 00:09:48.769
which is weird. That normally happens with comets, but then
205
00:09:48.830 --> 00:09:51.590
comets have a huge amount of emission, like you're superheating.
206
00:09:52.210 --> 00:09:55.129
The water on the comet is heating up and being
207
00:09:55.210 --> 00:09:57.740
jetted out, and that gives additional propulsion. None of that
208
00:09:57.779 --> 00:10:01.039
appeared from Oumuamua. This caused guys like Avi Loeb, who
209
00:10:01.620 --> 00:10:03.200
used to be a credible scientist, but has turned into
210
00:10:03.220 --> 00:10:04.820
a bit of a crackpot these days to say, oh,
211
00:10:04.879 --> 00:10:08.389
it's actually a spacecraft. Pretty sure it's a rock. But
212
00:10:08.429 --> 00:10:10.230
it was a weird rock. Like, near as you could tell,
213
00:10:10.269 --> 00:10:14.009
it was sort of cigar-shaped, and tumbling fairly slowly.
214
00:10:14.049 --> 00:10:16.509
There was another one that came through last year, wasn't it?
215
00:10:16.590 --> 00:10:18.639
Yep. I think we're on a third or fourth now.
216
00:10:19.080 --> 00:10:22.340
And people were freaking out thinking it was a spaceship
217
00:10:22.379 --> 00:10:24.940
and there was also, you know, look at these lights,
218
00:10:25.039 --> 00:10:25.360
you know.
219
00:10:25.539 --> 00:10:28.799
Yeah. It's pretty unlikely, you know. And, you know, you
220
00:10:28.840 --> 00:10:30.399
need really great evidence if you're going to make a
221
00:10:30.440 --> 00:10:31.299
great claim like that.
222
00:10:31.460 --> 00:10:31.679
Right.
223
00:10:31.720 --> 00:10:33.470
But the main thing that's happened and I think it's
224
00:10:33.490 --> 00:10:35.230
the good news is, you know, for the past 10
225
00:10:35.230 --> 00:10:38.230
years or so, we've had good enough sensors to start
226
00:10:38.269 --> 00:10:41.230
to be able to see these things. They've always been there.
227
00:10:41.250 --> 00:10:44.840
We just haven't seen them. All right, on the computing side,
228
00:10:44.860 --> 00:10:47.629
a bunch of interesting stories in 2017. We can nail
229
00:10:47.649 --> 00:10:50.600
the Microsoft ones quickly, some good news, some bad. Visual
230
00:10:50.620 --> 00:10:54.440
Studio 2017, which I don't think is a very remarkable
231
00:10:54.519 --> 00:10:56.820
version of Studio, except that it was really the version
232
00:10:56.860 --> 00:11:00.299
that started embracing the reality of we're in a GitHub
233
00:11:00.559 --> 00:11:04.019
Azure container world where we don't care what the operating
234
00:11:04.059 --> 00:11:07.779
system the container runs in. We get. NET Core 1.1.
235
00:11:07.759 --> 00:11:10.139
They're putting together the tools for the way that we
236
00:11:10.179 --> 00:11:14.639
work today where you're deploying into the cloud using Linux-based containers.
237
00:11:14.659 --> 00:11:14.980
Yeah.
238
00:11:15.320 --> 00:11:17.480
So, you know, much more focused on crossbot for 2017
239
00:11:17.480 --> 00:11:20.320
was kind of the version there. That's also, this is
240
00:11:20.340 --> 00:11:22.340
also the year that Teams launched, which I guess would
241
00:11:22.360 --> 00:11:26.179
be the bad news. Did I say that out loud?
242
00:11:26.230 --> 00:11:26.570
I feel bad.
243
00:11:26.590 --> 00:11:27.299
I hate Teams.
244
00:11:27.779 --> 00:11:29.789
Teams hates you back. It's only fair.
245
00:11:29.809 --> 00:11:31.529
I use it under duress.
246
00:11:31.970 --> 00:11:34.970
Yeah. It's also the year that Coplex finally shuts down
247
00:11:34.990 --> 00:11:37.669
because of course they've moved to GitHub completely. Started back
248
00:11:37.950 --> 00:11:41.509
in 2006. It was never, I mean, it was a
249
00:11:41.669 --> 00:11:44.590
source forge kind of mindset, but it's definitely the place
250
00:11:44.669 --> 00:11:48.340
where Microsoft would put up early code for us to
251
00:11:48.379 --> 00:11:51.220
download and play with and comment on. And we saw
252
00:11:51.240 --> 00:11:56.519
that rapid iteration, right? The AJAX toolkit, Silverlight came through
253
00:11:56.620 --> 00:12:00.279
that channel. So by October, it'll be read-only. And by December,
254
00:12:00.299 --> 00:12:04.179
it'll be gone and replaced with Archive. July is when
255
00:12:04.200 --> 00:12:07.279
Microsoft announced end of support for Windows Phone. And I
256
00:12:07.340 --> 00:12:11.470
think I kept a shrine to my 930 for five years.
257
00:12:11.970 --> 00:12:13.750
I have three of them in the drawer of broken dreams.
258
00:12:13.769 --> 00:12:16.389
You know, light a candle for a phone. Yeah, that's it.
259
00:12:16.649 --> 00:12:17.649
Yeah, there they sit.
260
00:12:17.840 --> 00:12:20.259
Okay, that's enough Microsoft stuff. Let's talk about the other
261
00:12:20.299 --> 00:12:23.279
important stuff that happens here. In June of 2017, the
262
00:12:23.399 --> 00:12:27.700
Google Brain Group publishes a paper called the Attention Paper.
263
00:12:28.679 --> 00:12:32.309
So this is eight Google researchers that were really focused
264
00:12:32.350 --> 00:12:36.789
on trying to improve translation. And they'd recognize that the
265
00:12:36.909 --> 00:12:40.309
way that machine learning was reading language was inefficient. It
266
00:12:40.330 --> 00:12:43.980
was sort of a one word at a time Bayesian
267
00:12:44.059 --> 00:12:47.200
methodology that didn't make for good translation. And they saw
268
00:12:47.220 --> 00:12:51.840
this parallel method they called attention mechanisms. So that every,
269
00:12:52.019 --> 00:12:53.379
the way they described it in the paper, which is
270
00:12:53.419 --> 00:12:56.110
very eloquent, is this is every word examines every other
271
00:12:56.149 --> 00:12:59.850
word based solely on the attention mechanism dispensing with recurrence
272
00:12:59.889 --> 00:13:01.250
and convolutions entirely.
273
00:13:01.429 --> 00:13:02.690
That sounds very familiar.
274
00:13:02.990 --> 00:13:06.750
What they're talking about is transformer architecture. And it's not
275
00:13:06.789 --> 00:13:09.090
a big splashy paper. It didn't make the news and
276
00:13:09.129 --> 00:13:13.049
so forth. But certain people read it. And it's sort of,
277
00:13:13.059 --> 00:13:15.980
you know, we talk about what's going on today. This
278
00:13:16.039 --> 00:13:21.019
is kind of the foundational paper, even more than the
279
00:13:21.019 --> 00:13:25.230
2020 hockey stick paper about how these models will grow.
280
00:13:25.769 --> 00:13:27.830
This is the one that makes the scientists go, hmm,
281
00:13:27.929 --> 00:13:31.610
we should look at processing language this way. 2017 is
282
00:13:31.649 --> 00:13:35.769
also the year of WannaCry. Now, I normally don't talk
283
00:13:35.830 --> 00:13:38.309
about viruses other than like first and so forth. What
284
00:13:38.330 --> 00:13:40.009
was the big deal with WannaCry? Well, it was big.
285
00:13:40.820 --> 00:13:43.679
It hit hundreds of millions of machines. It was a
286
00:13:43.720 --> 00:13:47.419
ransomware attack. The thing that was crazy about it was
287
00:13:47.500 --> 00:13:50.320
that it utilized a vulnerability that had a code name,
288
00:13:50.779 --> 00:13:54.340
Eternal Blue, because it had been developed by the NSA
289
00:13:54.679 --> 00:13:57.840
years before. And rather than getting a patch, they literally
290
00:13:57.940 --> 00:14:01.110
used it as a spying strategy. But when they got
291
00:14:01.169 --> 00:14:04.210
hacked by shadow brokers and leaked it, then it got
292
00:14:04.309 --> 00:14:08.529
widely distributed. Now, Microsoft always had a patch for it.
293
00:14:08.529 --> 00:14:10.259
They just didn't push it very much. But all of
294
00:14:10.279 --> 00:14:13.019
a sudden, they need to a whole bunch. The other
295
00:14:13.080 --> 00:14:17.759
weird part about this is that the attack, the ransomware
296
00:14:17.799 --> 00:14:21.960
attack that is WannaCry, had a kill switch. And the
297
00:14:22.039 --> 00:14:25.340
kill switch was based on every time the worm ran,
298
00:14:25.360 --> 00:14:28.139
it would check for a domain name that had never
299
00:14:28.179 --> 00:14:30.649
been registered. And if it was registered, then it would
300
00:14:30.669 --> 00:14:35.629
shut itself off. And when Marcus Hutchins had decompiled the
301
00:14:35.730 --> 00:14:37.970
ransomware and figured out how it worked, he registered the
302
00:14:37.990 --> 00:14:42.529
domain name and essentially stopped WannaCry. Beautiful. It's crazy, isn't it?
303
00:14:42.899 --> 00:14:45.019
Yeah, it is. But it's neat.
304
00:14:45.100 --> 00:14:45.340
Yeah.
305
00:14:45.500 --> 00:14:49.740
Yeah. It's very interesting. Last but not least, we're now
306
00:14:49.799 --> 00:14:52.120
coming into the era of this wave of quantum, which
307
00:14:52.159 --> 00:14:54.019
has faded off now, but this is when this new
308
00:14:54.080 --> 00:14:56.279
ascension happens. This is when Microsoft comes out big and
309
00:14:56.299 --> 00:14:58.659
they say, hey, we're going to start investing in quantum
310
00:14:58.679 --> 00:15:00.850
in a big way. Jason Zander actually led this team,
311
00:15:01.299 --> 00:15:04.730
and they talked about the topological qubit. You can go
312
00:15:04.769 --> 00:15:06.470
listen to Pat Hines' show about that if you really
313
00:15:06.509 --> 00:15:08.470
want to dig into it deeply. But they also then
314
00:15:08.529 --> 00:15:12.529
started immediately announcing toolkits. This becomes Qsharp and Azure Quantum
315
00:15:12.570 --> 00:15:15.850
and all that, and not to be left out. IBM
316
00:15:15.909 --> 00:15:20.240
also announces they have a prototype 50-qubit processor, and they
317
00:15:20.299 --> 00:15:25.639
have their own tool set called OpenQuasm, Q-A-S-M. And they
318
00:15:25.659 --> 00:15:27.460
start with 2.0. I don't know why, but that's what
319
00:15:27.480 --> 00:15:31.429
they do. And also another library called Quizzit. But it's
320
00:15:31.490 --> 00:15:34.730
the beginnings of the platform players saying, we think we're
321
00:15:34.750 --> 00:15:35.870
going to get something here.
322
00:15:35.889 --> 00:15:36.090
Mm-hmm.
323
00:15:36.519 --> 00:15:38.279
And so they started to put the ingredients together. I mean,
324
00:15:38.500 --> 00:15:41.470
you know, we're now looking back nine years and not
325
00:15:41.590 --> 00:15:44.009
much has happened, although we will have a wave. And
326
00:15:44.049 --> 00:15:46.309
over the next few shows, I'll make sure to include
327
00:15:46.330 --> 00:15:48.049
it because that's been certainly been something I've been playing
328
00:15:48.070 --> 00:15:48.710
Ghost in Tension to.
329
00:15:48.889 --> 00:15:49.049
Yeah.
330
00:15:49.230 --> 00:15:50.830
That's the stories of 2017.
331
00:15:50.830 --> 00:15:53.350
All right. Well, I guess it's time for that. fun
332
00:15:53.389 --> 00:15:57.679
little tune that introduces the section called better know a framework.
333
00:15:58.200 --> 00:16:06.139
All right, man, what do you got?
334
00:16:06.279 --> 00:16:10.799
Well, I have something that I'm, I'm doing and I
335
00:16:10.820 --> 00:16:13.440
want to tell everybody about it. And for two reasons, one,
336
00:16:14.240 --> 00:16:18.379
it's an application that I produced and, And I am
337
00:16:18.450 --> 00:16:21.590
a software producer on this project. I am not a
338
00:16:21.649 --> 00:16:22.549
software developer.
339
00:16:22.769 --> 00:16:23.110
Right.
340
00:16:23.169 --> 00:16:27.009
Because I used an AI to do the dirty work.
341
00:16:28.649 --> 00:16:31.470
But it came out of necessity. And the reason that
342
00:16:31.509 --> 00:16:33.409
I want to bring it up is not just because
343
00:16:33.429 --> 00:16:35.769
it's a neat story, but I want some feedback from
344
00:16:35.830 --> 00:16:45.309
our listeners. Especially those who lead bands, musical groups. and
345
00:16:45.429 --> 00:16:49.870
have to wrangle the troops to find dates to play.
346
00:16:51.149 --> 00:16:54.009
So it's called Can We Play It? And you can
347
00:16:54.049 --> 00:17:00.230
read the documentation at canweplayit.app. Okay. It has links to
348
00:17:00.309 --> 00:17:05.970
the stores, but they're not there yet. But anyway, there's screenshots.
349
00:17:06.450 --> 00:17:09.309
There's an explanation of what it does. And I really
350
00:17:09.349 --> 00:17:13.619
would like anybody who manages bands to And especially, you know, bigger.
351
00:17:13.819 --> 00:17:17.099
Mine's a 10 piece, right? So it's very difficult emails
352
00:17:17.140 --> 00:17:19.220
and texts and spreadsheets and trying to figure out who
353
00:17:19.660 --> 00:17:22.940
can play what, when. And then people have questions and
354
00:17:23.009 --> 00:17:25.329
I can't answer them. I have to go back through email.
355
00:17:25.789 --> 00:17:27.490
Who's on this gig again? You know?
356
00:17:27.769 --> 00:17:27.970
Right.
357
00:17:28.210 --> 00:17:31.259
So basically it answers the question, can we play it?
358
00:17:32.049 --> 00:17:35.009
And here's how it works. I, the leader, I have
359
00:17:35.049 --> 00:17:37.630
an account and I'm the only one who pays for
360
00:17:37.670 --> 00:17:41.279
that account, subscribes, has a subscription. And then I invite
361
00:17:41.339 --> 00:17:45.079
members by email or text, depending on their preferred method
362
00:17:45.099 --> 00:17:48.359
of communication. And this is really important because some of
363
00:17:48.400 --> 00:17:51.859
my members never read an email. Some only read text.
364
00:17:51.880 --> 00:17:52.880
Some never read texts.
365
00:17:53.059 --> 00:17:53.279
Right.
366
00:17:53.400 --> 00:17:56.900
So it's an invite and then they install it. And
367
00:17:56.940 --> 00:18:02.240
then once I've invited all the members, their job is
368
00:18:02.299 --> 00:18:06.279
to go through the next three months and pick all
369
00:18:06.319 --> 00:18:08.460
the dates that they can't play.
370
00:18:08.660 --> 00:18:08.859
Right.
371
00:18:09.099 --> 00:18:12.509
Not that they can't. they can't. So they're the exceptions.
372
00:18:13.309 --> 00:18:15.829
And once everybody has done that and I get a
373
00:18:15.869 --> 00:18:18.890
request from a venue, Hey, can you play this date?
374
00:18:19.190 --> 00:18:20.869
All I do is I pick the date and I say,
375
00:18:20.910 --> 00:18:23.569
can we play it? And it says yes or no.
376
00:18:23.589 --> 00:18:28.140
And it takes into account subs substitutes. Like we know
377
00:18:28.160 --> 00:18:30.940
there's a trumpet player has two subs that can fill
378
00:18:30.980 --> 00:18:33.579
in for him. Tenor sax player has a couple of
379
00:18:33.619 --> 00:18:37.069
subs that could fill in for him. Um, And then
380
00:18:37.109 --> 00:18:40.069
basically I can send, if everybody can do it, I
381
00:18:40.109 --> 00:18:44.769
can send a confirmation with one click to everyone. Notifications
382
00:18:44.829 --> 00:18:49.960
in the app, email, text, whatever. And then they reply
383
00:18:50.019 --> 00:18:52.440
yes or no. And if they can't, they can assign
384
00:18:52.500 --> 00:18:56.279
a sub. And I can know in a matter of
385
00:18:56.920 --> 00:19:01.079
half an hour versus weeks of chasing down members if
386
00:19:01.119 --> 00:19:03.750
we can play this gig or not. We've even lost
387
00:19:03.910 --> 00:19:08.150
gigs because people didn't respond in time. And I have
388
00:19:08.250 --> 00:19:10.769
one member in my band who's kind of a senior citizen.
389
00:19:11.470 --> 00:19:15.210
Not that that matters, but his mind is going fast.
390
00:19:15.890 --> 00:19:18.970
He actually said, what's the deal with this gig here?
391
00:19:19.009 --> 00:19:21.690
And I said, yeah, you said you could play it. No,
392
00:19:21.769 --> 00:19:25.230
I got another gig. He double booked because he didn't
393
00:19:25.279 --> 00:19:30.240
remember saying that he could play this gig. Anyway, so
394
00:19:30.279 --> 00:19:33.259
can we play it.app? I wrote it as a Maui
395
00:19:33.279 --> 00:19:37.690
blazer app and it's got an Azure API backend and,
396
00:19:38.230 --> 00:19:42.769
and SQL and all that stuff. So I would like feedback,
397
00:19:42.890 --> 00:19:47.029
go look at the screenshots and tell me if you were,
398
00:19:47.049 --> 00:19:50.670
you know, if you're managing a band or whatever, you know,
399
00:19:50.950 --> 00:19:54.319
what would be some features that we could change or add?
400
00:19:54.339 --> 00:19:55.039
Cool.
401
00:19:55.450 --> 00:20:00.160
So I really wanted one of these Samsung Z Fold,
402
00:20:02.400 --> 00:20:06.180
you know, phones that opens up and it has this thing.
403
00:20:07.019 --> 00:20:10.059
And I went to AI and I said, hey, I
404
00:20:10.099 --> 00:20:12.460
want to write an application that can utilize this. Give
405
00:20:12.500 --> 00:20:14.740
me some ideas. And it said, based on what I
406
00:20:14.799 --> 00:20:17.829
know about you and all this, here's a couple of ideas.
407
00:20:18.640 --> 00:20:23.430
And Kelly said, you can only buy that if you
408
00:20:23.490 --> 00:20:25.269
write an app that can make money.
409
00:20:25.289 --> 00:20:27.089
Nice.
410
00:20:28.430 --> 00:20:30.750
So that's what I landed on. Anyway, that's what I got.
411
00:20:30.789 --> 00:20:32.910
Please send me some feedback if you're a musician or
412
00:20:32.930 --> 00:20:35.529
a band member or a leader. And let me know
413
00:20:35.569 --> 00:20:39.450
if you've got some suggestions. Canweplayit.app. That's what I got.
414
00:20:39.589 --> 00:20:40.490
Who's talking to us, Richard?
415
00:20:40.769 --> 00:20:43.549
Grabbed a comment off show 2014. That's the one we
416
00:20:43.609 --> 00:20:45.519
just recorded with Sean Wildermuth.
417
00:20:45.839 --> 00:20:46.480
Yeah.
418
00:20:46.599 --> 00:20:49.519
Who was talking, you know, with how life as a
419
00:20:49.539 --> 00:20:52.259
senior developer suddenly changed by AI, right?
420
00:20:52.579 --> 00:20:52.859
Yeah.
421
00:20:52.880 --> 00:20:56.240
Which I think is particularly relevant. And another old developer,
422
00:20:56.380 --> 00:20:59.930
Richard Rukima, or Code Pewter, commented on it. So, welcome
423
00:20:59.950 --> 00:21:00.690
to the old gang.
424
00:21:00.869 --> 00:21:01.130
Yeah.
425
00:21:01.289 --> 00:21:03.869
And I like Richard's comment just because it's a troublemaker.
426
00:21:03.930 --> 00:21:04.789
I like Richard.
427
00:21:04.849 --> 00:21:06.609
He said, I had the luxury of starting with AI
428
00:21:06.650 --> 00:21:09.490
harnesses on Greenfield projects where I controlled 100% of the requirements.
429
00:21:09.529 --> 00:21:11.470
It didn't take long before I decided to lean into
430
00:21:11.529 --> 00:21:15.579
it completely. Full YOLO mode. Full access. Full clients. Give
431
00:21:15.619 --> 00:21:17.859
the harness the intent. Give it the boundaries. And let
432
00:21:18.140 --> 00:21:18.779
it run.
433
00:21:18.799 --> 00:21:18.880
Yeah.
434
00:21:19.559 --> 00:21:22.079
My working assumption became simple. Build software the same way
435
00:21:22.099 --> 00:21:25.259
I would with Visual Studio. using the same architectural principles
436
00:21:25.299 --> 00:21:27.880
and standards that have developed over decades, except moved 100
437
00:21:27.880 --> 00:21:31.400
or even 1,000 times faster. And listening to you three,
438
00:21:31.519 --> 00:21:34.039
I hear a much more conservative approach. AI is useful,
439
00:21:34.059 --> 00:21:35.700
but you have to handle it like a grenade with
440
00:21:35.740 --> 00:21:39.410
the pin pulled. I understand the caution. I've deliberately been
441
00:21:39.450 --> 00:21:42.410
testing the other side of the equation. Instead of constantly
442
00:21:42.470 --> 00:21:44.910
watching every action, I've been asking, what happens if I
443
00:21:44.950 --> 00:21:48.369
configure the environment properly, establish the architecture and constraints, point
444
00:21:48.410 --> 00:21:51.250
the harness at the intent, and get out of its way?
445
00:21:52.470 --> 00:21:54.430
The further I push that approach, the more interesting the
446
00:21:54.470 --> 00:21:57.609
results have become. The code increasingly comes out structured the
447
00:21:57.630 --> 00:22:00.259
way I expect because I've spent time configuring the harness
448
00:22:00.430 --> 00:22:03.940
around how I build software. There are failures. There are pitfalls.
449
00:22:04.119 --> 00:22:06.019
You still need to review what gets produced, but the
450
00:22:06.059 --> 00:22:08.519
sheer amount of work one experienced person gets through changes
451
00:22:08.579 --> 00:22:12.569
the economics of the software development. We're approaching a strange problem.
452
00:22:12.660 --> 00:22:14.910
One person with an AI harness produces so much code
453
00:22:14.930 --> 00:22:17.609
so quickly that you start wondering whether the bottleneck becomes
454
00:22:17.670 --> 00:22:21.490
having enough experienced humans to review the output. That's the
455
00:22:21.529 --> 00:22:23.710
edge I'm interested in finding. And the further I push it,
456
00:22:23.789 --> 00:22:26.329
the faster I'm getting work done. It's scary, but it's
457
00:22:26.369 --> 00:22:27.029
also enthralling.
458
00:22:27.170 --> 00:22:30.690
I have a comment. And I love Richard Rukema. Rukema?
459
00:22:30.789 --> 00:22:31.710
Yeah.
460
00:22:31.849 --> 00:22:34.690
He came to Dakota in the castle a couple of
461
00:22:34.710 --> 00:22:38.569
years ago in Tuscany. But Richard, I'm going to defend
462
00:22:38.710 --> 00:22:44.490
our conservative stance, or at least mine. A great person
463
00:22:44.750 --> 00:22:50.079
who shall remain nameless, Patrick Hines, once said that convenience
464
00:22:50.220 --> 00:22:53.980
is the enemy of security. And it's true. And I
465
00:22:54.079 --> 00:23:00.329
think you are experiencing the sheer convenience and to your amazement,
466
00:23:00.789 --> 00:23:04.750
the efficacy of this stuff right now. But the danger,
467
00:23:04.809 --> 00:23:07.390
of course, is if you get too complacent and you
468
00:23:07.410 --> 00:23:14.089
stop checking the code, then you are susceptible to security problems.
469
00:23:14.369 --> 00:23:18.119
And you're also susceptible to not being able to defend
470
00:23:18.240 --> 00:23:21.920
the code. Like somebody says, this is great, but why
471
00:23:21.940 --> 00:23:24.180
did you do that? And you kind of like scratch
472
00:23:24.200 --> 00:23:25.680
your head and say, I don't know, I'm going to
473
00:23:25.720 --> 00:23:28.250
have to check that out. And you'll probably ask your
474
00:23:28.309 --> 00:23:31.430
AI to check it out for you. But so in
475
00:23:31.509 --> 00:23:36.410
this early times, I admit I'm using AI to produce
476
00:23:37.369 --> 00:23:40.829
and I'm being very productive, but I go very slowly.
477
00:23:40.990 --> 00:23:43.089
I take it in steps and in every step I
478
00:23:43.170 --> 00:23:45.930
check and I test and And I look, and I
479
00:23:45.990 --> 00:23:48.269
set breakpoints, and I find out what it's doing, and
480
00:23:48.329 --> 00:23:52.490
I okay it, just like a manager would do that
481
00:23:52.630 --> 00:23:58.250
to software developers under that manager. And I think that's
482
00:23:58.289 --> 00:24:01.430
the smart way to go. You're saving all this time.
483
00:24:02.009 --> 00:24:05.210
Why not just put a little bit back into checking?
484
00:24:05.230 --> 00:24:06.990
All right, Rich, I don't think you need another copy
485
00:24:07.029 --> 00:24:10.269
of Muse to Codeby, but feel free to ping me anyway.
486
00:24:10.549 --> 00:24:13.559
It's certainly great to get in touch. And if you'd
487
00:24:13.599 --> 00:24:14.759
like a copy of Music to Code By, write a
488
00:24:14.779 --> 00:24:17.160
comment on the website at. netrocks. com. We're on the Facebooks.
489
00:24:17.180 --> 00:24:18.700
We publish every show there. And if you comment there
490
00:24:18.720 --> 00:24:20.220
and I read it in the show, we'll send you
491
00:24:20.240 --> 00:24:21.059
a copy of Music to Code By.
492
00:24:21.400 --> 00:24:23.980
And if you want to just get Music to Code By,
493
00:24:24.019 --> 00:24:30.819
go to musictocodeby.net. You can download the 22-track collection in MP3, FLAC,
494
00:24:30.980 --> 00:24:32.359
and WAV formats.
495
00:24:34.029 --> 00:24:34.440
All right.
496
00:24:34.529 --> 00:24:36.809
And now let's bring in Billy. You heard him talking
497
00:24:36.849 --> 00:24:39.970
in the intro there, but he needs no introduction. But
498
00:24:39.990 --> 00:24:42.170
I'm still going to read his bio anyway because it's great.
499
00:24:42.819 --> 00:24:45.319
Billy Hollis is a software designer and developer with a
500
00:24:45.400 --> 00:24:46.720
contrarian streak.
501
00:24:47.839 --> 00:24:48.819
Really? Shock.
502
00:24:48.859 --> 00:24:49.849
Is this a new bio?
503
00:24:50.329 --> 00:24:50.789
Yeah, weird.
504
00:24:52.170 --> 00:24:56.109
This contrarian streak often challenges conventional wisdom in the industry.
505
00:24:56.710 --> 00:24:59.990
He has a consulting practice in Nashville, Tennessee, in the
506
00:25:00.029 --> 00:25:04.940
United States, and his team focuses on user experience design,
507
00:25:05.539 --> 00:25:11.670
advanced user interface development, rules-based architectures, and healthcare systems. He
508
00:25:11.690 --> 00:25:15.470
teaches classes for design thinking and UX design and technical
509
00:25:15.509 --> 00:25:19.549
classes on XAML for WinUI and XAML for WPF. It's
510
00:25:19.710 --> 00:25:20.410
Billy Hollis.
511
00:25:20.710 --> 00:25:23.650
Thank you, Carl. Always nice to be back. And what
512
00:25:23.730 --> 00:25:26.140
is this, 26, 7?
513
00:25:26.140 --> 00:25:27.220
Probably. Something like that.
514
00:25:27.240 --> 00:25:30.759
Something like that at the podcast. I've done something like that.
515
00:25:30.819 --> 00:25:32.859
You're out there in the front runners, my friend, for sure.
516
00:25:32.900 --> 00:25:36.400
Yeah. And I've enjoyed every single one of them. So
517
00:25:36.640 --> 00:25:39.309
we ought to have some fun today. I was listening
518
00:25:39.349 --> 00:25:43.970
to you talk about your use of AI and what's
519
00:25:44.130 --> 00:25:45.460
the person I'm not familiar with?
520
00:25:45.490 --> 00:25:46.150
Richard Rukema?
521
00:25:46.539 --> 00:25:51.940
Yeah. And I completely respect that and what it's doing.
522
00:25:52.019 --> 00:25:55.640
And I hear other people saying similar things. And I'm
523
00:25:55.779 --> 00:26:01.480
coming at it from a different perspective because, first of all,
524
00:26:01.549 --> 00:26:04.089
I don't do as much actual coding as I used
525
00:26:04.109 --> 00:26:05.349
to have a team for that. A lot of the
526
00:26:05.369 --> 00:26:08.970
stuff we're doing is pretty advanced and advanced architectures. And
527
00:26:09.029 --> 00:26:11.369
I don't know that we're at the point that these
528
00:26:11.769 --> 00:26:16.430
all-encompassing systems are accessible to just prompting your way through
529
00:26:16.450 --> 00:26:19.170
it yet. You could do pieces, but I'm not sure
530
00:26:19.210 --> 00:26:21.859
you could do kind of the unified stuff that we
531
00:26:21.920 --> 00:26:25.000
tend to do. In addition to that, I don't find
532
00:26:25.980 --> 00:26:29.980
a lot of the coding AI generation to be very
533
00:26:30.039 --> 00:26:32.519
helpful because I find it doesn't do a very good
534
00:26:32.559 --> 00:26:36.380
job with XAML. Because I guess there's not really a
535
00:26:36.420 --> 00:26:40.440
lot of good training material for it out there.
536
00:26:40.599 --> 00:26:42.609
And if what there is, you probably wrote it.
537
00:26:43.240 --> 00:26:48.829
Well, actually probably not because while I put my stuff
538
00:26:48.910 --> 00:26:51.769
out there sort of freely available, I've never put anything
539
00:26:51.789 --> 00:26:54.549
on GitHub where it was, where the models would, would
540
00:26:54.609 --> 00:26:58.279
normally go to get it. Um, and, and, I've observed
541
00:26:58.339 --> 00:27:01.160
over the years that the quality of the XAML code
542
00:27:01.180 --> 00:27:04.359
that I see written by the typical average corporate developer,
543
00:27:04.579 --> 00:27:09.440
which probably comprises the stuff on GitHub, that quality isn't
544
00:27:09.460 --> 00:27:13.480
very good. They don't use XAML in the interesting way
545
00:27:13.500 --> 00:27:16.259
that my team and I tend to. So it doesn't
546
00:27:16.299 --> 00:27:20.539
really help me much in the coding area because of that.
547
00:27:21.160 --> 00:27:25.119
It doesn't help me in the writing area. Because I've
548
00:27:25.140 --> 00:27:28.720
written several million words, and my brain's trained up to
549
00:27:28.819 --> 00:27:31.019
do that. So that if I'm going to write a
550
00:27:31.039 --> 00:27:34.000
three-paragraph email to somebody, there's no way in the world
551
00:27:34.039 --> 00:27:36.599
I'm going to an LLM to do that. I mean,
552
00:27:36.980 --> 00:27:39.119
I can write that so quickly, and it will express
553
00:27:39.240 --> 00:27:42.960
exactly what I want. And I guess maybe as a writer,
554
00:27:43.279 --> 00:27:47.140
I'm very tuned to the tone and the style of
555
00:27:47.339 --> 00:27:51.529
AI writing. And it just sounds wrong to me. It's
556
00:27:51.569 --> 00:27:53.930
just sloppy. So I don't use it in the writing.
557
00:27:54.190 --> 00:27:58.089
I use it for image generation a fair amount because
558
00:27:58.130 --> 00:27:59.950
if I'm doing slideshows and such and I want to
559
00:27:59.970 --> 00:28:03.809
get illustrated visuals of the things that I do. And
560
00:28:04.210 --> 00:28:06.309
while that is helpful and I get some good stuff
561
00:28:06.369 --> 00:28:08.430
out of it, I also get some of the most
562
00:28:08.599 --> 00:28:12.859
bizarre and incoherent stuff out of it that I've ever seen,
563
00:28:12.890 --> 00:28:16.460
which I posted some of those on my ex Twitter account.
564
00:28:16.920 --> 00:28:20.099
And at times, I mean, I've got an absolute incoherent
565
00:28:20.279 --> 00:28:24.980
gibberish out of A.I. I was doing a search just
566
00:28:25.019 --> 00:28:29.559
to find words that rhyme with something. And I posted
567
00:28:29.619 --> 00:28:32.059
it on my Twitter account. I'll give a link to
568
00:28:32.099 --> 00:28:34.880
that for the show if you like. But what Copilot
569
00:28:34.900 --> 00:28:38.480
came back with, I'm telling you, it was just incoherent.
570
00:28:38.819 --> 00:28:42.059
And that makes me nervous about it. I can't have
571
00:28:42.099 --> 00:28:45.680
that confidence that he had in the comment that everything
572
00:28:45.700 --> 00:28:47.869
is just going to be okay because I see too
573
00:28:47.880 --> 00:28:48.960
many things going wrong.
574
00:28:49.200 --> 00:28:52.220
Yeah, and you said it when you said it doesn't
575
00:28:52.319 --> 00:28:58.029
particularly write good XAML. It does, however, write good Blazor, C-sharp,
576
00:28:58.470 --> 00:29:04.730
web code, really good. And it turns out the Blazor
577
00:29:04.930 --> 00:29:07.410
MAUI stuff it did a really good job on as well.
578
00:29:07.990 --> 00:29:10.319
And I'm perfectly prepared to believe all that, that it
579
00:29:10.380 --> 00:29:13.759
is doing a good job in those areas. Yeah. So
580
00:29:13.819 --> 00:29:16.740
I keep trying new things and I do run into
581
00:29:17.319 --> 00:29:20.759
enough obstacles and blocks that I don't use it routinely,
582
00:29:20.779 --> 00:29:23.519
but I try to investigate things. And I also look
583
00:29:23.559 --> 00:29:26.740
at kind of the bigger picture if I can about
584
00:29:26.789 --> 00:29:29.269
what's going on in the industry and try to understand
585
00:29:30.430 --> 00:29:34.089
how things are going to develop. And there are potential
586
00:29:34.589 --> 00:29:36.849
issues there. I think a lot of us are kind
587
00:29:36.869 --> 00:29:39.130
of sensitive to it. And I think it's important to
588
00:29:39.150 --> 00:29:43.079
talk about that stuff because the proponents of AI, the
589
00:29:43.119 --> 00:29:45.940
guys who have a vested interest in it, are not
590
00:29:46.039 --> 00:29:48.740
going to talk about these things. They don't like to.
591
00:29:49.019 --> 00:29:52.859
There are some fairly well-known people, including AI researchers out there,
592
00:29:53.220 --> 00:29:56.500
like Gary Marcus, if you know who that is, and
593
00:29:56.579 --> 00:29:59.880
a fellow named Ed Zitron. And they kind of take
594
00:30:00.279 --> 00:30:04.430
that negative sort of, this could go wrong in these ways.
595
00:30:04.589 --> 00:30:07.990
And I honestly think it's worthwhile for us to know
596
00:30:08.029 --> 00:30:09.490
about them. There's an old saying.
597
00:30:09.690 --> 00:30:10.230
Absolutely.
598
00:30:10.529 --> 00:30:13.990
That if you see a problem rolling down the hill
599
00:30:14.089 --> 00:30:18.220
at you, that it's nine out of 10 chances that
600
00:30:18.240 --> 00:30:19.759
it's going to roll into a ditch before it gets
601
00:30:19.779 --> 00:30:23.539
to you. And I think that's probably true. But if
602
00:30:23.579 --> 00:30:26.819
you've got enough problems, eventually one of them hits you
603
00:30:27.440 --> 00:30:31.279
and starts to cause the problem. I've got a list.
604
00:30:31.319 --> 00:30:33.680
Let's see if I can go through it and see
605
00:30:33.700 --> 00:30:36.119
which one of these resonate with you guys and you
606
00:30:36.140 --> 00:30:37.549
want to talk about. I have one at the end
607
00:30:37.569 --> 00:30:40.329
that I'd like to go through. What are we going
608
00:30:40.349 --> 00:30:42.089
to do about the people who can't find work because
609
00:30:42.109 --> 00:30:42.890
of AI displacement?
610
00:30:43.029 --> 00:30:44.509
That's a huge issue.
611
00:30:44.869 --> 00:30:47.450
What's that going to do to them psychologically? That's a
612
00:30:47.509 --> 00:30:52.140
huge issue. Are we in a bubble? If we are,
613
00:30:52.599 --> 00:30:55.180
what kind is it? What will the aftermath be? Because
614
00:30:55.299 --> 00:30:58.900
some bubbles don't don't really cause any long-term problems. I
615
00:30:58.920 --> 00:31:01.000
think about the fiber optic bubble in the late 90s.
616
00:31:01.859 --> 00:31:04.579
And there was a huge bubble in it. A lot
617
00:31:04.599 --> 00:31:08.849
of companies that over-invested in it lost everything. Other people
618
00:31:08.930 --> 00:31:12.069
bought those assets for pennies on the dollars and created
619
00:31:12.089 --> 00:31:14.809
the foundation of today's internet out of all that fiber
620
00:31:14.869 --> 00:31:15.589
optic capacity.
621
00:31:16.109 --> 00:31:18.630
Yeah, well, I mean, that was really the web bubble. Yeah,
622
00:31:18.670 --> 00:31:20.650
that was the web bubble. Right, the dot-com boom.
623
00:31:20.890 --> 00:31:23.559
And we have a bubble like that in which everything
624
00:31:23.579 --> 00:31:25.259
will turn out fine in the end. It's just that
625
00:31:25.319 --> 00:31:28.210
people got too exuberant early on. Right.
626
00:31:28.329 --> 00:31:31.309
Well, you get into what's the ass, what's the residual
627
00:31:31.410 --> 00:31:33.210
asset you can buy for 10 cents on the dollar.
628
00:31:33.250 --> 00:31:36.750
That's a, yeah. Yeah. And I don't, I'm not really
629
00:31:36.789 --> 00:31:38.990
sure I understand exactly where that's going. We'll talk about
630
00:31:39.019 --> 00:31:40.839
that at home because that's kind of another danger.
631
00:31:40.960 --> 00:31:44.099
Well, I mean, the residual asset is data centers, but
632
00:31:44.180 --> 00:31:48.279
those companies are building nose aren't financially, you know, stressed.
633
00:31:48.500 --> 00:31:51.700
Actually they're starting to, but they're, they're, they're big.
634
00:31:52.039 --> 00:31:54.339
There's also a lot of social pushback to data centers.
635
00:31:54.779 --> 00:31:55.819
all around the United States.
636
00:31:55.839 --> 00:31:57.140
That's a, that's a problem.
637
00:31:57.200 --> 00:32:00.019
Yeah. But that's, that's not the major issue here. They're spending,
638
00:32:00.059 --> 00:32:01.039
they're spending the money on.
639
00:32:00.980 --> 00:32:03.119
It one way or the other. And they're spending the money,
640
00:32:03.200 --> 00:32:06.670
but they have not discovered. That's another problem that the
641
00:32:06.769 --> 00:32:11.789
cost of the compute to service their customers is not
642
00:32:11.829 --> 00:32:12.309
making money.
643
00:32:12.609 --> 00:32:15.130
I think you're getting the bigger issue here is building
644
00:32:15.150 --> 00:32:17.839
a model, just organizing electrons by generating a lot of heat.
645
00:32:18.059 --> 00:32:21.140
And that money's gone. Well, it's what it's been is
646
00:32:21.220 --> 00:32:23.799
paid to the company built the data center, you know,
647
00:32:24.490 --> 00:32:27.529
The joke here is that Microsoft and Amazon and Google
648
00:32:27.569 --> 00:32:29.720
are posting record profits through this whole thing.
649
00:32:29.880 --> 00:32:33.200
Yeah, so they're going to keep going despite what are
650
00:32:33.240 --> 00:32:37.710
potential external effects, such as the negative effects of tying
651
00:32:37.789 --> 00:32:42.900
up so much of the planet's compute. Gaming computers, people
652
00:32:42.940 --> 00:32:46.640
building their own gaming computers has just absolutely gone off
653
00:32:46.660 --> 00:32:46.940
the cliff.
654
00:32:46.960 --> 00:32:47.940
It's impossible, yeah.
655
00:32:48.220 --> 00:32:51.920
Yeah, because they can't find the components at the price
656
00:32:51.940 --> 00:32:54.930
that they need. Now, if that spreads into the wider
657
00:32:54.990 --> 00:32:58.210
corporate world, what does that mean for corporations who need compute?
658
00:32:58.549 --> 00:33:01.029
My advice on Run As is here, my top of
659
00:33:01.069 --> 00:33:03.450
the year show was don't buy machines this year. Extend
660
00:33:03.490 --> 00:33:04.109
your warranties.
661
00:33:04.170 --> 00:33:05.130
Yeah, yeah.
662
00:33:05.170 --> 00:33:07.309
Because you're going to overpay for those machines you've got
663
00:33:07.329 --> 00:33:09.910
to amortize over four or five years. It doesn't make sense. Mm-hmm.
664
00:33:10.650 --> 00:33:12.579
But I was also presuming this was going to end
665
00:33:12.680 --> 00:33:14.480
and it doesn't look like it's ending anytime soon.
666
00:33:14.599 --> 00:33:15.920
It doesn't look like it's going to, no.
667
00:33:16.460 --> 00:33:19.039
I mean, it is deflating now. There's no question about
668
00:33:19.420 --> 00:33:22.119
the stock prices are off, the smart money starting to
669
00:33:22.200 --> 00:33:23.900
pull out of AI as it is.
670
00:33:24.160 --> 00:33:25.920
As it is, yeah. I kind of see some of
671
00:33:25.960 --> 00:33:26.640
that trends as well.
672
00:33:26.660 --> 00:33:28.400
But that doesn't mean there isn't a crash coming.
673
00:33:29.230 --> 00:33:32.809
But the problem I probably spend more time thinking about
674
00:33:32.950 --> 00:33:38.910
than any other is brain rot. The tendency of AI
675
00:33:39.349 --> 00:33:43.109
to get I mean, my philosophy about that has to
676
00:33:43.170 --> 00:33:46.690
do with the fact that thinking is hard work. And
677
00:33:46.730 --> 00:33:49.029
there are a lot of people who would prefer not
678
00:33:49.069 --> 00:33:51.579
to think. I mean, you start to see things in
679
00:33:51.650 --> 00:33:53.880
our industry where everybody just wants to be told what
680
00:33:53.920 --> 00:33:54.220
to do.
681
00:33:54.619 --> 00:33:56.019
Right.
682
00:33:56.220 --> 00:33:59.740
Just I don't want to design it. Just tell me,
683
00:34:00.599 --> 00:34:04.220
give me a set of design guidelines and prescriptive design
684
00:34:04.259 --> 00:34:06.849
that will just do it for me. And because I
685
00:34:06.869 --> 00:34:08.929
don't want to think about that. Well, I understand why
686
00:34:09.190 --> 00:34:11.340
design is kind of a weird discipline. But I think
687
00:34:11.380 --> 00:34:14.900
that that's a bigger, bigger issue is a lot of people,
688
00:34:14.920 --> 00:34:17.780
I think, get into doing their systems with AI because
689
00:34:18.260 --> 00:34:21.380
they don't want to think about the details anymore. Well, OK.
690
00:34:21.539 --> 00:34:23.179
Well, there's a group of people who thought it was
691
00:34:23.199 --> 00:34:25.210
going to be brain rot. We allowed garbage collection to
692
00:34:25.250 --> 00:34:28.269
happen automatically, too. Like you reference your own memory.
693
00:34:28.449 --> 00:34:30.670
That's what else comes to mind.
694
00:34:30.829 --> 00:34:33.730
I guess I'm not sympathetic to that because I always
695
00:34:33.809 --> 00:34:36.969
worked on a system that did garbage collection. I never
696
00:34:37.030 --> 00:34:38.170
worked in the secret. Yeah.
697
00:34:38.269 --> 00:34:42.050
And Socrates got really angry that his students weren't memorizing
698
00:34:42.090 --> 00:34:44.230
their speeches anymore because they had papyrus.
699
00:34:44.389 --> 00:34:44.969
That's right.
700
00:34:45.570 --> 00:34:47.750
It's like, this is brain rot. If you don't have
701
00:34:47.789 --> 00:34:49.630
going to put out the effort to memorize your speech,
702
00:34:49.690 --> 00:34:50.789
how serious are you?
703
00:34:50.989 --> 00:34:54.760
I also think that a software producer using AI to
704
00:34:54.820 --> 00:35:01.619
help them with design and code still has the brain engaged. And,
705
00:35:01.639 --> 00:35:04.380
you know, for example, for this last app that I
706
00:35:04.400 --> 00:35:07.469
worked on, I asked it for a couple of design
707
00:35:08.230 --> 00:35:11.900
options and I picked the one that I thought was good.
708
00:35:12.139 --> 00:35:15.519
And I, and I, and I augmented it, you know,
709
00:35:15.679 --> 00:35:18.300
based on, but the, the difference is I didn't have
710
00:35:18.340 --> 00:35:20.260
to go to Figma and figure out how to use
711
00:35:20.300 --> 00:35:23.500
their interface and all that stuff and CSS and blah, blah, blah.
712
00:35:24.039 --> 00:35:27.199
I just said, this is what I want. But I
713
00:35:27.260 --> 00:35:28.769
still had the designer's eye.
714
00:35:29.679 --> 00:35:29.949
Right.
715
00:35:30.389 --> 00:35:34.050
And I still had that input. And also, when developing
716
00:35:34.090 --> 00:35:36.519
these things, you don't just say, here, write a piece
717
00:35:36.539 --> 00:35:39.860
of software, poof, it's done. You do it in steps.
718
00:35:40.780 --> 00:35:44.239
And I've tested everything. It even helped me test things.
719
00:35:44.400 --> 00:35:48.239
I said, I want a script that I can go
720
00:35:48.340 --> 00:35:51.440
through to exercise all of the features with these three
721
00:35:51.480 --> 00:35:54.639
phones and these three users. Give me that, and I'll
722
00:35:54.679 --> 00:35:57.679
take screenshots along the way for documentation and tell me
723
00:35:57.719 --> 00:36:02.050
what the screenshot should be. So that was really, really helpful. Now,
724
00:36:02.070 --> 00:36:06.619
I could have done that without a script, but having
725
00:36:06.659 --> 00:36:10.219
that script, I could focus on the tests and testing
726
00:36:10.280 --> 00:36:12.199
and figuring out these problems. And in the course of
727
00:36:12.239 --> 00:36:17.460
doing that, I found several bugs that I hadn't found before.
728
00:36:18.219 --> 00:36:19.980
That's a producer's job, right?
729
00:36:20.179 --> 00:36:23.400
I will grant everything that you're saying there. Yeah. But
730
00:36:23.460 --> 00:36:26.449
understand that we're in a special circumstance here. We have
731
00:36:26.469 --> 00:36:29.190
two generations of people who had to learn to think
732
00:36:29.250 --> 00:36:29.989
about these things.
733
00:36:30.030 --> 00:36:30.269
Yeah.
734
00:36:30.750 --> 00:36:32.250
Conceptually and in detail.
735
00:36:32.409 --> 00:36:32.610
Right.
736
00:36:32.710 --> 00:36:37.550
Being leveraged in with these new tools. But, I mean,
737
00:36:37.590 --> 00:36:41.889
there's a pretty big related thing in college. College professors
738
00:36:41.929 --> 00:36:45.769
will tell you that their students don't write anymore. No,
739
00:36:45.949 --> 00:36:49.840
that's true. They just don't. shove it out with AI
740
00:36:49.880 --> 00:36:52.400
and they're not really learning that much. They don't read
741
00:36:52.519 --> 00:36:54.760
in depth anymore. They talk about the fact that they
742
00:36:54.780 --> 00:36:58.980
have students who couldn't possibly get through a 200 page book.
743
00:36:59.059 --> 00:36:59.320
Yeah.
744
00:36:59.599 --> 00:37:01.519
How many 200 page books have you guys written?
745
00:37:01.619 --> 00:37:02.780
That's hundreds.
746
00:37:03.599 --> 00:37:06.610
Written, not just bread. I mean, if you talk about red,
747
00:37:06.909 --> 00:37:10.849
it's probably in the thousands. And so, so that's, I'm
748
00:37:10.889 --> 00:37:15.630
talking about more of the longterm of people not developing
749
00:37:15.789 --> 00:37:19.070
their capacity in a way that And there are some
750
00:37:19.110 --> 00:37:22.030
studies that if you depend heavily on AI, especially for writing,
751
00:37:22.349 --> 00:37:25.789
that it does degrade your ability to then put it aside.
752
00:37:26.889 --> 00:37:29.510
But like you said, it is early times. These are
753
00:37:29.550 --> 00:37:32.789
new methods of learning emerging. The biggest challenge you're having
754
00:37:32.829 --> 00:37:37.019
in schools over this is that The testing apparatus cannot
755
00:37:37.119 --> 00:37:40.360
tolerate these changes. You no longer can test students in
756
00:37:40.380 --> 00:37:41.099
a meaningful way.
757
00:37:41.400 --> 00:37:41.659
Yeah.
758
00:37:41.780 --> 00:37:44.820
And remember, because I said at the beginning, nine out
759
00:37:44.860 --> 00:37:46.539
of ten of the things rolling down that look like
760
00:37:46.559 --> 00:37:48.860
a problem may not end up being a problem. And
761
00:37:49.119 --> 00:37:51.800
so I kind of granted that at the beginning. But
762
00:37:51.880 --> 00:37:55.280
I still think it's worthwhile to kind of understand which ones,
763
00:37:56.019 --> 00:37:59.619
what are the candidates? What are the possibilities? and maybe
764
00:37:59.679 --> 00:38:03.030
a little bit about how do they interact. Then I've
765
00:38:03.050 --> 00:38:07.110
seen a little bit about AI rot, the idea of
766
00:38:07.210 --> 00:38:09.929
AI being trained on AI-generated content.
767
00:38:10.030 --> 00:38:10.869
Yeah, don't do that.
768
00:38:11.210 --> 00:38:13.090
Yeah, well, I'm not sure they're going to be able
769
00:38:13.110 --> 00:38:13.710
to avoid that.
770
00:38:13.730 --> 00:38:17.269
Well, they definitely are because they see exactly that. It's degenerative.
771
00:38:17.409 --> 00:38:19.809
Yeah, it collapses the model there.
772
00:38:19.829 --> 00:38:22.889
They're also now scrounging for new original data sources, like
773
00:38:23.469 --> 00:38:26.650
buying antique books, cutting them apart, and scanning them.
774
00:38:26.889 --> 00:38:30.059
Hey, I hate to interrupt this fiery conversation, but we
775
00:38:30.139 --> 00:38:33.639
do have to break here for these very important messages,
776
00:38:33.679 --> 00:38:39.099
so we'll be right back. Developers are asking a new question.
777
00:38:39.659 --> 00:38:43.320
Do we still need third-party libraries when AI can generate
778
00:38:43.420 --> 00:38:48.340
so much code? For common UI tasks, maybe, but not always.
779
00:38:49.179 --> 00:38:55.969
But deep functionality is different. Document editing, PDF generation... Rendering, conversion,
780
00:38:56.130 --> 00:38:59.969
and enterprise document workflows are the result of years of engineering,
781
00:39:00.190 --> 00:39:05.000
extensive testing, and millions of real-world deployments. AI can generate code, yeah,
782
00:39:05.059 --> 00:39:09.360
but it can't instantly generate decades of proven document technology.
783
00:39:10.159 --> 00:39:14.780
TXTextControl gives developers that foundation so you can focus on
784
00:39:14.820 --> 00:39:19.409
building your application, not rebuilding a document engine. Learn more
785
00:39:19.550 --> 00:39:20.670
at textcontrol.com.
786
00:39:26.539 --> 00:39:27.460
And we're back. It's.
787
00:39:27.429 --> 00:39:27.750
NET Rocks.
788
00:39:27.769 --> 00:39:30.090
I'm Richard Campbell. That's Carl Franglin. Yo. And Taco. with
789
00:39:30.130 --> 00:39:32.570
our friend, Billy Hollis. And I'm going to hijack the
790
00:39:32.610 --> 00:39:34.369
conversation just for a bit to push back on a
791
00:39:34.389 --> 00:39:37.739
few of these things too, because I agree with you, Billy,
792
00:39:37.760 --> 00:39:40.280
that right now, while we have skilled people in the
793
00:39:40.449 --> 00:39:45.599
old methods, they're reigning these tools into some degree because
794
00:39:45.619 --> 00:39:48.719
the tools are mature. You know, you can, I mean,
795
00:39:49.130 --> 00:39:52.949
We're having a revolution in security testing happening with these
796
00:39:52.989 --> 00:39:55.710
tools right now. The rate of patches for repairing a
797
00:39:55.730 --> 00:39:58.820
software is all time, which really counters the argument you
798
00:39:58.880 --> 00:40:02.300
can't make secure software with LLMs. It simply isn't true.
799
00:40:03.039 --> 00:40:05.380
The LLMs are turning into the tool most effective at
800
00:40:05.440 --> 00:40:09.059
finding security flaws. You just have to use them properly.
801
00:40:09.619 --> 00:40:13.170
But the real argument here is as our generation leaves
802
00:40:13.230 --> 00:40:15.309
this industry, don't you think the tools are going to
803
00:40:15.329 --> 00:40:17.750
improve enough that security is just not going to be
804
00:40:17.869 --> 00:40:20.449
optional because that's what the tools require?
805
00:40:20.489 --> 00:40:23.690
It's not now. It was never optional.
806
00:40:24.110 --> 00:40:24.449
No, no.
807
00:40:24.510 --> 00:40:27.309
I'm talking about the tools actually enforcing it, right? We've
808
00:40:27.329 --> 00:40:30.780
been able to ignore it, but as these tools improve,
809
00:40:31.400 --> 00:40:33.320
you think this will be an essential part of the process,
810
00:40:33.579 --> 00:40:34.480
as is architecture.
811
00:40:34.900 --> 00:40:37.539
And again, I grant all those things. And as I said,
812
00:40:38.420 --> 00:40:41.320
not all problems that look like they might come at you,
813
00:40:41.340 --> 00:40:42.420
end up coming at you.
814
00:40:42.980 --> 00:40:44.280
But it's worth talking about.
815
00:40:44.440 --> 00:40:46.610
Well, often you can run in traffic and get away
816
00:40:46.619 --> 00:40:47.989
with it. It doesn't mean you're not going to be
817
00:40:48.090 --> 00:40:49.469
splatted by a 4x4.
818
00:40:49.489 --> 00:40:54.269
That's right. And then I start to think about the
819
00:40:54.389 --> 00:40:58.889
wider sort of social effects. One of the examples from
820
00:40:58.929 --> 00:41:03.119
history that I like to think about is the effect
821
00:41:03.159 --> 00:41:06.440
of the automobile. You can just do all the cascading
822
00:41:06.500 --> 00:41:10.679
things about how the automobile works. See, when people first
823
00:41:11.920 --> 00:41:16.019
had it, it became a part of society. The conventional
824
00:41:16.039 --> 00:41:18.840
thinking was, well, yeah, it's a faster horse carriage.
825
00:41:19.039 --> 00:41:20.239
Yeah, we want faster horses.
826
00:41:20.389 --> 00:41:24.989
We want fast horses. But, I mean, the automobile caused
827
00:41:25.730 --> 00:41:28.679
the drive-in theater. because of other technological changes.
828
00:41:28.699 --> 00:41:29.429
Created suburbs.
829
00:41:29.800 --> 00:41:36.409
Created suburbs. It changed the mating rituals of our whole society.
830
00:41:36.690 --> 00:41:39.070
Well, where else would you go to the submarine races?
831
00:41:39.150 --> 00:41:42.860
Yeah. So when I kind of look at some of
832
00:41:42.920 --> 00:41:47.460
the effects of not just AI, but other related technologies
833
00:41:47.480 --> 00:41:51.900
and how they come together, it concerns me that the
834
00:41:52.260 --> 00:41:55.449
person-to-person communication that has been a part of society since
835
00:41:56.190 --> 00:41:59.880
history began Look, I know young people of my son's
836
00:41:59.920 --> 00:42:03.260
generation who will not answer a phone call.
837
00:42:03.900 --> 00:42:06.460
Dude, I won't answer a phone call unless it's on
838
00:42:06.519 --> 00:42:07.159
my schedule.
839
00:42:07.179 --> 00:42:10.840
Well, it goes beyond that, though. They also have no
840
00:42:10.880 --> 00:42:13.949
experience in actually talking to people in real time. They're
841
00:42:14.070 --> 00:42:17.449
terrified to call a doctor's office and set up an appointment.
842
00:42:17.469 --> 00:42:19.130
All of those things are true, Billy. I'm not going
843
00:42:19.150 --> 00:42:21.090
to argue with a minute, but it's NAI's fault.
844
00:42:21.269 --> 00:42:21.550
I agree.
845
00:42:22.510 --> 00:42:24.389
That was happening before this tool appeared.
846
00:42:24.409 --> 00:42:24.550
Yeah.
847
00:42:25.059 --> 00:42:27.840
I will posit that it's a temporary problem. All my
848
00:42:27.900 --> 00:42:32.320
kids went through this. Kelly's kids wanted their mom to
849
00:42:32.360 --> 00:42:34.519
make the calls on their behalf. And she said, no,
850
00:42:35.019 --> 00:42:36.519
we're going to sit here. You're going to make the
851
00:42:36.559 --> 00:42:39.219
call and you're going to talk to them. And they
852
00:42:39.260 --> 00:42:41.179
got over it. Do you know what I mean? It's
853
00:42:41.239 --> 00:42:45.019
not a permanent problem. I think it's a temporary technologically
854
00:42:45.099 --> 00:42:46.280
disruptive related.
855
00:42:46.300 --> 00:42:51.510
I hope that you are correct about that. And of course,
856
00:42:51.769 --> 00:42:54.550
the fact that they depend on On AI for writing,
857
00:42:54.590 --> 00:42:56.690
someone has to come out of college doing that, I think.
858
00:42:56.710 --> 00:42:57.329
Yeah.
859
00:42:57.429 --> 00:43:02.690
That puts layers in between people's communication. People talk about
860
00:43:02.750 --> 00:43:05.670
work slot, where somebody's in a job and they've got
861
00:43:05.690 --> 00:43:08.150
to generate some report or something. And they just kind
862
00:43:08.210 --> 00:43:09.849
of have AI do it. It doesn't have any depth
863
00:43:09.869 --> 00:43:12.670
or originality or anything in it. But they fulfill the
864
00:43:12.750 --> 00:43:15.699
letter of the job. That's very disrespectful.
865
00:43:15.719 --> 00:43:17.429
That's just because the job's not well described.
866
00:43:17.510 --> 00:43:22.719
It's not. Yeah. Well, yeah. Understood that. But those are all.
867
00:43:22.760 --> 00:43:24.179
This is a transitory effect.
868
00:43:24.219 --> 00:43:26.840
Things that have to be adapted, that we have to
869
00:43:26.900 --> 00:43:33.630
adapt to. Right. So, and, and you, Carl, you may
870
00:43:33.650 --> 00:43:35.949
be right that they're all going to get over it.
871
00:43:36.010 --> 00:43:38.469
But have you seen the clips?
872
00:43:38.670 --> 00:43:40.349
With help. Yeah.
873
00:43:40.530 --> 00:43:46.389
Have you seen the clips of college graduates booing commencement speakers? Yeah.
874
00:43:46.590 --> 00:43:47.690
Who talked about AI.
875
00:43:48.429 --> 00:43:49.210
Yep. Sure have.
876
00:43:49.309 --> 00:43:53.170
Okay. They are, they are not excited about this. They're
877
00:43:53.230 --> 00:43:53.989
negative about it.
878
00:43:54.250 --> 00:43:58.880
Yeah. And especially the Gen Z generation who didn't have
879
00:43:59.039 --> 00:44:02.820
the high school experience that they were, that their friends had,
880
00:44:03.059 --> 00:44:06.519
their older and younger friends. So, because that was COVID time,
881
00:44:06.639 --> 00:44:09.579
they were robbed of my, my youngest daughter was robbed
882
00:44:09.639 --> 00:44:12.800
of her high school graduation, her high school summer, like
883
00:44:13.519 --> 00:44:17.170
all of that stuff. just kind of went away for her.
884
00:44:17.429 --> 00:44:21.630
So they all turned inward into their phones and things
885
00:44:21.690 --> 00:44:26.070
and became very mistrustful of people, I think. And it
886
00:44:26.150 --> 00:44:29.050
took her a long time to, she lives in New
887
00:44:29.090 --> 00:44:32.210
York now, and to go out and pound the pavement
888
00:44:32.329 --> 00:44:35.639
and hit the streets and talk to people and be social. But,
889
00:44:35.820 --> 00:44:38.420
you know, she still doesn't answer my texts for three days.
890
00:44:40.559 --> 00:44:41.480
And I'm her father.
891
00:44:41.500 --> 00:44:47.219
Yeah. So I see a lot of sort of general
892
00:44:47.239 --> 00:44:52.599
stuff like that. The tangible example that I talk about
893
00:44:53.000 --> 00:44:55.480
because I've done some presentations and so I have a
894
00:44:55.480 --> 00:45:00.070
fair amount of thinking about it is where we're at
895
00:45:00.090 --> 00:45:06.309
in recruiting, interviewing HR stuff, because that's a pretty good
896
00:45:06.409 --> 00:45:11.289
one that that HR people. I've met some really good
897
00:45:11.369 --> 00:45:14.150
ones and I've also met some really lazy ones. And
898
00:45:14.269 --> 00:45:16.690
the lazy ones were some of the very earliest ones
899
00:45:16.809 --> 00:45:20.690
to get into automation of things that to me are
900
00:45:20.750 --> 00:45:25.219
fundamentally people to people. Right. And shouldn't be automated. So
901
00:45:25.260 --> 00:45:29.159
that we spent probably 15 years or so that resumes
902
00:45:29.239 --> 00:45:32.480
went to every resume in the tech industry has this
903
00:45:32.639 --> 00:45:36.260
big list of acronyms of the stuff you're supposed to,
904
00:45:36.619 --> 00:45:39.840
the technologies you're supposed to know. I call that alphabet
905
00:45:39.940 --> 00:45:45.760
soup in my life. my course on interviewing developers. Um, but,
906
00:45:45.820 --> 00:45:48.559
and they had, they did, people did that because the
907
00:45:48.679 --> 00:45:55.420
HR people automated resume screening on a keyword basis. Uh,
908
00:45:55.619 --> 00:45:58.280
and so people had to put stuff in there just
909
00:45:58.300 --> 00:46:02.139
to get through that first screen of, of the resumes or,
910
00:46:02.219 --> 00:46:05.420
or that they didn't go any further down the path. Um, and,
911
00:46:05.559 --> 00:46:09.079
and that was, it was understandable from the standpoint that
912
00:46:09.260 --> 00:46:11.119
if you got thousands of resumes, you got to find
913
00:46:11.199 --> 00:46:14.019
some way, to automate get through. On the other hand,
914
00:46:14.539 --> 00:46:18.010
I mean, I used to read a couple dozen resumes
915
00:46:18.090 --> 00:46:21.590
a day when I was running a Microsoft partner in
916
00:46:21.630 --> 00:46:23.710
the late 90s. That was a big part of my
917
00:46:23.750 --> 00:46:26.250
job and doing the interviewing. That's where I got the
918
00:46:26.289 --> 00:46:29.059
experience to draw on for some of these courses I did. Actually,
919
00:46:29.079 --> 00:46:32.159
the first session I did on interviewing developers was at
920
00:46:32.340 --> 00:46:38.119
DevConnections 2014, I think, Richard. And it was a well-intended session.
921
00:46:38.380 --> 00:46:38.920
It went really well.
922
00:46:38.940 --> 00:46:41.559
That's why I kept developing in that area. And there
923
00:46:41.599 --> 00:46:43.090
was a guy that went to that session that I
924
00:46:43.159 --> 00:46:46.010
ran into a couple of years later. And he said
925
00:46:46.530 --> 00:46:50.730
that that approach of asking questions that were more about
926
00:46:50.750 --> 00:46:55.849
the human side of being a developer, that he used
927
00:46:55.889 --> 00:46:58.030
my process and some of the questions that I used
928
00:46:58.070 --> 00:47:02.519
and such. And he said it improved retention. Sure. And
929
00:47:02.639 --> 00:47:05.050
so I think there are getting into the human to
930
00:47:05.130 --> 00:47:09.480
human ways has has tremendous benefit. But HR seems to
931
00:47:09.519 --> 00:47:10.199
be going the other way.
932
00:47:10.300 --> 00:47:13.380
Well, I mean, they're also being– they are getting inundated
933
00:47:13.400 --> 00:47:16.239
with thousands of resumes. What do you do, right? And
934
00:47:16.559 --> 00:47:19.000
I remember triaging off resumes the same way where the
935
00:47:19.099 --> 00:47:21.059
slightest flaw in the resume, you just put it aside
936
00:47:21.079 --> 00:47:22.699
because you had to triage.
937
00:47:22.840 --> 00:47:23.280
Yeah.
938
00:47:24.030 --> 00:47:26.829
And now this has just been amplified by these tools
939
00:47:27.070 --> 00:47:27.829
both ways.
940
00:47:28.070 --> 00:47:28.670
Both ways.
941
00:47:28.809 --> 00:47:31.389
Both ways. In the end, this is a self-correcting system
942
00:47:31.409 --> 00:47:34.070
because you can't hire the people you need and the
943
00:47:34.110 --> 00:47:36.429
people who want to get hired can't get hired. Then
944
00:47:36.469 --> 00:47:37.340
they adapt.
945
00:47:37.409 --> 00:47:37.550
Yeah.
946
00:47:37.909 --> 00:47:41.389
So I think we're just in a weird time. I
947
00:47:41.429 --> 00:47:42.230
would agree that it's a.
948
00:47:42.210 --> 00:47:44.769
Weird time because at this particular point in time, most
949
00:47:44.829 --> 00:47:46.260
resumes are being written by AI.
950
00:47:46.510 --> 00:47:46.889
Sure.
951
00:47:46.949 --> 00:47:49.889
And they're being read by it too. And they're being
952
00:47:49.969 --> 00:47:52.070
read by AI. And I don't know if you saw
953
00:47:52.110 --> 00:47:55.119
the research on this. Somebody did some research to test this,
954
00:47:55.400 --> 00:48:01.260
and they found out that the AI's reading these resumes
955
00:48:01.679 --> 00:48:06.389
have a strong preference for for resumes that were written
956
00:48:06.710 --> 00:48:07.449
by the same AI.
957
00:48:07.489 --> 00:48:07.789
Of course.
958
00:48:08.010 --> 00:48:08.269
Yeah.
959
00:48:08.469 --> 00:48:10.269
Yeah. Yeah.
960
00:48:10.429 --> 00:48:10.679
Well.
961
00:48:11.179 --> 00:48:11.760
Figure that one.
962
00:48:11.949 --> 00:48:13.800
That's kind of a danger zone to me.
963
00:48:13.920 --> 00:48:18.159
There's evidence of this phenomena happening. I saw a commercial
964
00:48:18.199 --> 00:48:19.980
and I've seen it a couple of times from Chase,
965
00:48:20.039 --> 00:48:23.320
you know, the Chase Bank, right? And it's all about
966
00:48:23.480 --> 00:48:26.900
how robots have gone wrong, right? It starts out with
967
00:48:26.940 --> 00:48:29.739
a robot saying, coffee, how would you like your coffee?
968
00:48:29.780 --> 00:48:31.460
And the lady says, no sugar. And he takes the
969
00:48:31.539 --> 00:48:34.030
sugar thing and dumps it in there. And then it's
970
00:48:34.070 --> 00:48:37.889
just like, here's your copy. Enjoy it. And other things
971
00:48:37.969 --> 00:48:41.360
like that. And then the tagline is at Chase, we
972
00:48:41.389 --> 00:48:44.679
don't spend time figuring out how to remove people from
973
00:48:44.719 --> 00:48:47.860
the equation, but we're opening more centers where you can
974
00:48:47.900 --> 00:48:50.820
talk to us face to face. And that resonates with
975
00:48:50.860 --> 00:48:55.880
people for this very reason. But if a bank is saying, hey,
976
00:48:56.860 --> 00:49:00.070
you come to us, we're real people. Jeez, where were
977
00:49:00.110 --> 00:49:02.190
those banks 10 years ago when I used to work?
978
00:49:03.119 --> 00:49:05.500
Because there wasn't as much last batch against technology as
979
00:49:05.519 --> 00:49:08.909
there is right now. Going back to the graduates booing
980
00:49:09.130 --> 00:49:12.329
AI people, can you blame them? No. These guys have
981
00:49:12.369 --> 00:49:15.769
been marketing to investors on the back of, we're destroying
982
00:49:15.809 --> 00:49:18.329
all work, so you better invest in us. These are
983
00:49:18.349 --> 00:49:22.579
the guys trying to enter the workplace. Like, no surprise.
984
00:49:22.980 --> 00:49:25.219
And I think this is also a feedback of the pandemic.
985
00:49:25.300 --> 00:49:27.969
We're all a bit tired of screens. Once upon a time,
986
00:49:28.010 --> 00:49:30.630
we really liked the automation because it was more efficient.
987
00:49:31.190 --> 00:49:34.409
And now we're starting to remember, because we lost enough
988
00:49:34.449 --> 00:49:37.360
of it, that face-to-face communication sucks. kind of awesome.
989
00:49:37.480 --> 00:49:37.860
Yeah.
990
00:49:38.000 --> 00:49:42.110
And we're craving it, right? Even though these young kids
991
00:49:42.150 --> 00:49:45.699
or whatever are getting more depressed because they're not What
992
00:49:45.719 --> 00:49:48.739
they really want is human contact. And I think we
993
00:49:48.800 --> 00:49:49.119
all do.
994
00:49:49.400 --> 00:49:49.860
I do too.
995
00:49:50.139 --> 00:49:52.320
And so I think that it's bouncing back.
996
00:49:52.559 --> 00:49:55.039
And I'd like to think that somebody will figure out
997
00:49:55.079 --> 00:49:57.039
how to leverage AI to make that happen.
998
00:49:57.239 --> 00:50:00.559
Well, I think exactly right. These are just tools. It
999
00:50:00.599 --> 00:50:03.050
has a terrible name that makes everybody stupid.
1000
00:50:03.610 --> 00:50:04.239
Yeah.
1001
00:50:04.280 --> 00:50:07.110
But in the end, there are just new automation tools.
1002
00:50:07.570 --> 00:50:10.650
And the fact that we could have a face-to-face conversation
1003
00:50:11.090 --> 00:50:14.309
while the tool is doing a transcript and double-checking facts
1004
00:50:14.369 --> 00:50:17.400
and collecting useful information so that we get good results
1005
00:50:17.420 --> 00:50:19.800
from that conversation, that's a hell of an asset.
1006
00:50:19.920 --> 00:50:22.420
Well, to me, a useful sort of way of looking
1007
00:50:22.440 --> 00:50:24.699
at it is, from what I see, AI tends to
1008
00:50:24.739 --> 00:50:28.090
be good at things that are measurable, that are quantifiable
1009
00:50:28.670 --> 00:50:31.909
in some sense. And there are human things that aren't.
1010
00:50:33.150 --> 00:50:37.690
And so, for example, in the whole recruitment process, trust
1011
00:50:38.179 --> 00:50:41.000
is– People don't even think about this, I think, when
1012
00:50:41.019 --> 00:50:43.969
they interview. One of the things that they're zeroing in
1013
00:50:44.110 --> 00:50:47.389
on intuitively is, can I trust this person to be
1014
00:50:47.469 --> 00:50:52.429
on my team? And I mean, I don't know that
1015
00:50:52.449 --> 00:50:54.369
we're ever going to have a tool that says, yes, well,
1016
00:50:54.389 --> 00:50:58.760
this person ranks 92.6 on the international trust measurement scale.
1017
00:50:59.920 --> 00:51:02.360
I'm not sure that we're close to anything like that.
1018
00:51:02.860 --> 00:51:04.840
So now you're talking about things that have to be
1019
00:51:04.900 --> 00:51:09.139
established person to person. And the people who are going
1020
00:51:09.719 --> 00:51:12.619
whole hog into applying AI to all this stuff don't
1021
00:51:12.639 --> 00:51:14.340
seem to get that. So I think there's going to
1022
00:51:14.360 --> 00:51:16.579
be damage. I think there are going to be companies.
1023
00:51:17.719 --> 00:51:21.179
And again, a lot of stuff in the long term,
1024
00:51:21.469 --> 00:51:23.639
I hope, will fall out. But I think there will
1025
00:51:23.659 --> 00:51:28.210
be companies that will degrade themselves to the point of
1026
00:51:28.409 --> 00:51:34.909
failure because they over-depend on AI to pick their people.
1027
00:51:35.630 --> 00:51:36.050
Remember, AI.
1028
00:51:37.010 --> 00:51:40.869
It's all about the measurable stuff, but it also sort
1029
00:51:40.909 --> 00:51:44.369
of brings all the world's knowledge and puts it into
1030
00:51:44.389 --> 00:51:47.730
a blended sludge. Yeah.
1031
00:51:47.769 --> 00:51:49.929
Well, that is the superpower, right?
1032
00:51:50.309 --> 00:51:50.389
Yeah.
1033
00:51:50.409 --> 00:51:52.690
We've always had too much data. We certainly– and it's
1034
00:51:52.710 --> 00:51:54.929
been a crisis for a long time. And now we
1035
00:51:55.030 --> 00:51:59.909
finally found a way to JPEG language. It's a lossy
1036
00:51:59.949 --> 00:52:02.769
compression format, but it allows you to get through a
1037
00:52:02.809 --> 00:52:05.969
lot more data. The problem is that once it presents
1038
00:52:06.050 --> 00:52:08.610
information to you, you got to fact check it because
1039
00:52:09.150 --> 00:52:11.719
it's a lossy compression format.
1040
00:52:11.860 --> 00:52:16.480
And the concern about that for me being coming from
1041
00:52:16.500 --> 00:52:20.039
the design world, et cetera, is that if you fall
1042
00:52:20.079 --> 00:52:22.460
into those patterns of thought of kind of everything average,
1043
00:52:22.500 --> 00:52:27.809
everything sort of, if you don't work consciously, to free
1044
00:52:27.829 --> 00:52:31.179
yourself from that and come outside of it. Now, how
1045
00:52:31.199 --> 00:52:33.039
are you going to innovate? How are you going to
1046
00:52:33.059 --> 00:52:39.639
come up with completely new ways of doing things? And
1047
00:52:39.699 --> 00:52:45.949
companies need that. They need, look, a corporate organization can
1048
00:52:46.010 --> 00:52:49.210
only have a certain maximum number of people like me
1049
00:52:49.650 --> 00:52:52.750
who are always questioning everything. But you do need some
1050
00:52:52.809 --> 00:52:55.150
of them. They're the yeast in the bread, so to speak.
1051
00:52:55.190 --> 00:52:55.369
Sure.
1052
00:52:55.389 --> 00:53:00.320
That sort of makes things And AR tends to downgrade people.
1053
00:53:00.449 --> 00:53:02.340
If I could shift gears at the end of the
1054
00:53:02.380 --> 00:53:04.500
show here for a minute. You guys know who Mike
1055
00:53:04.579 --> 00:53:07.840
Rowe is, right? Dirty Jobs, Mike Rowe. He has been
1056
00:53:07.880 --> 00:53:11.059
an advocate for, and he even has a foundation for
1057
00:53:11.119 --> 00:53:13.119
supporting education for the trades.
1058
00:53:13.940 --> 00:53:15.960
You mean the least automatable class of work.
1059
00:53:16.260 --> 00:53:21.440
Yeah, exactly. And the whole argument goes, there are some
1060
00:53:21.760 --> 00:53:24.550
odd 40 million or 4 million. There's some multiple of
1061
00:53:24.550 --> 00:53:29.420
4 million people. open trade jobs right now and nobody
1062
00:53:29.559 --> 00:53:34.800
to fill them. And it's because people have, and going
1063
00:53:34.860 --> 00:53:38.260
back to when they took shop out of high school
1064
00:53:38.440 --> 00:53:43.739
and middle school, there has been this look down upon
1065
00:53:43.820 --> 00:53:47.000
the trades attitude. And you must go to college. You
1066
00:53:47.019 --> 00:53:49.280
must get a four-year degree. And everybody knows where that
1067
00:53:49.320 --> 00:53:50.460
got most people, right?
1068
00:53:50.619 --> 00:53:52.940
I just talked to a kid, a friend of mine's son,
1069
00:53:53.159 --> 00:53:57.320
who just graduated from a nice university in UK with
1070
00:53:57.360 --> 00:53:59.909
a mathematics degree. And you know what he's doing? He's
1071
00:53:59.929 --> 00:54:00.889
studying to be an electrician.
1072
00:54:00.989 --> 00:54:01.170
Yeah.
1073
00:54:01.469 --> 00:54:03.630
Not because he didn't like his mathematics degree, but he
1074
00:54:03.670 --> 00:54:06.690
wants to go make something. He thinks he'll go back
1075
00:54:06.730 --> 00:54:08.590
to school afterwards, but he's going to go apprentice.
1076
00:54:08.630 --> 00:54:11.199
One of my stepdaughters went to NYU for a very
1077
00:54:11.239 --> 00:54:14.300
expensive education, and she's a wedding planner. She's got her
1078
00:54:14.360 --> 00:54:16.500
own wedding planning business. It has nothing to do with
1079
00:54:16.780 --> 00:54:17.920
what her degree was in.
1080
00:54:18.059 --> 00:54:19.179
That's not that unusual.
1081
00:54:19.480 --> 00:54:23.159
But I wouldn't say that a degree isn't useful if
1082
00:54:23.199 --> 00:54:25.750
you think of it as a way to improve yourself
1083
00:54:25.809 --> 00:54:28.690
as a human, not necessarily to prepare for a job.
1084
00:54:29.380 --> 00:54:33.219
But the thing about the trades is that, yeah, like
1085
00:54:33.400 --> 00:54:38.840
Richard said, completely AI, they don't care. Your job isn't
1086
00:54:38.860 --> 00:54:41.849
going to be replaced if you're a carpenter, at least
1087
00:54:41.869 --> 00:54:42.530
for a long time.
1088
00:54:42.889 --> 00:54:45.230
Well, and I'm not convinced that anybody's job is going
1089
00:54:45.250 --> 00:54:48.909
to be replaced. It's just automation tools. You just move faster.
1090
00:54:49.599 --> 00:54:52.679
Well, we have talked a lot about in the past
1091
00:54:52.739 --> 00:54:55.190
about junior developers being screwed, right?
1092
00:54:55.269 --> 00:54:58.110
I disagree, actually, because junior developers embrace these new tools
1093
00:54:58.150 --> 00:54:58.829
better than anybody.
1094
00:54:58.909 --> 00:55:01.230
But do they have the experience to use them effectively?
1095
00:55:01.349 --> 00:55:04.030
Does anybody know if they don't? If they get good results,
1096
00:55:04.090 --> 00:55:04.769
how does it matter?
1097
00:55:05.010 --> 00:55:08.900
Who are they replacing exactly? Because they're not replacing anybody.
1098
00:55:09.179 --> 00:55:09.360
Yeah.
1099
00:55:09.480 --> 00:55:09.739
Okay.
1100
00:55:10.400 --> 00:55:13.500
Well, anyway, so I just wanted to throw that out
1101
00:55:13.539 --> 00:55:16.960
there that, you know, if you're thinking about a career,
1102
00:55:17.300 --> 00:55:20.989
young people, consider the trades, the trade school. You don't
1103
00:55:21.010 --> 00:55:23.659
need a degree. And sometimes you can make even more
1104
00:55:23.719 --> 00:55:26.539
than a couple hundred grand a year if you get
1105
00:55:26.579 --> 00:55:27.380
into the right job.
1106
00:55:27.599 --> 00:55:30.960
Well, the way I think about this, because I'm kind
1107
00:55:30.980 --> 00:55:34.320
of exposed to that world a lot. I took shop
1108
00:55:34.659 --> 00:55:37.219
when I was in junior high school. Took two years
1109
00:55:37.280 --> 00:55:40.050
of it. Partially because I'm kind of a generalist and
1110
00:55:40.090 --> 00:55:42.070
I want to be able to kind of understand all
1111
00:55:42.110 --> 00:55:45.889
the things. But I also come from a background. My brother,
1112
00:55:46.539 --> 00:55:50.099
is head of HVAC installation and support for one of
1113
00:55:50.119 --> 00:55:54.320
the local companies. And he has no college. He just
1114
00:55:54.400 --> 00:55:58.219
graduated from high school. But him and those people that
1115
00:55:58.260 --> 00:56:00.570
I know, the way I kind of describe it is
1116
00:56:01.309 --> 00:56:04.550
they live in a world of things. We live, for
1117
00:56:04.570 --> 00:56:07.869
the most part, developers, media, live in a world of
1118
00:56:08.110 --> 00:56:12.889
words and symbols and code. And people that live in
1119
00:56:12.949 --> 00:56:16.010
that world love AI. And people that live in the
1120
00:56:16.050 --> 00:56:21.769
world of things, it doesn't really resonate with them. I
1121
00:56:22.570 --> 00:56:25.070
think it will in some ways. I mean, I can
1122
00:56:25.110 --> 00:56:28.500
certainly imagine an AI that you take a picture of
1123
00:56:28.539 --> 00:56:32.300
this part of an HVAC system to get the exact
1124
00:56:32.360 --> 00:56:37.500
model number of whatever. There are certainly going to be
1125
00:56:37.559 --> 00:56:41.179
places where you develop pieces of that use the capabilities
1126
00:56:41.219 --> 00:56:44.429
of AI to help those guys do their job. Absolutely.
1127
00:56:44.469 --> 00:56:46.269
But at the end of the day, they're going to
1128
00:56:46.289 --> 00:56:48.929
be putting a hand on a thing and taking it
1129
00:56:49.050 --> 00:56:52.550
off and putting another thing on. And that's, I think
1130
00:56:52.590 --> 00:56:54.590
there's a fundamental difference in the way people look at
1131
00:56:54.610 --> 00:56:55.050
the world.
1132
00:56:55.199 --> 00:56:57.679
Before I started an exercise routine, I took a picture
1133
00:56:57.719 --> 00:56:59.760
of every machine in the gym and uploaded it to
1134
00:56:59.780 --> 00:57:03.139
chat GPT and said, make me a PDF of what
1135
00:57:03.199 --> 00:57:08.199
this machine does and how to use it with the pictures. So,
1136
00:57:08.360 --> 00:57:10.380
cause they, I asked them at the desk, they're like, no,
1137
00:57:10.420 --> 00:57:13.320
we don't have any kind of documentation or whatever. You
1138
00:57:13.340 --> 00:57:15.239
get what you get. Like you're expected to know how
1139
00:57:15.260 --> 00:57:16.000
to use these things.
1140
00:57:16.039 --> 00:57:17.360
But at the end of the day, you got to
1141
00:57:17.380 --> 00:57:19.219
get on the machine and start pushing some weights.
1142
00:57:19.320 --> 00:57:19.880
Of course.
1143
00:57:19.980 --> 00:57:23.969
Yeah. So, and that's, if you don't have that mindset
1144
00:57:23.989 --> 00:57:26.050
going in, it doesn't matter how much information you've got
1145
00:57:26.110 --> 00:57:27.050
on what the machines do.
1146
00:57:27.110 --> 00:57:27.369
Yeah.
1147
00:57:27.409 --> 00:57:29.309
Right. And so people that live in that world of
1148
00:57:29.369 --> 00:57:32.590
things that, that, that things are tangible for them. It,
1149
00:57:34.440 --> 00:57:36.000
At the end of the day, you work on the
1150
00:57:36.059 --> 00:57:38.559
air conditioner. It either blows cold air or it doesn't.
1151
00:57:38.699 --> 00:57:41.800
And no amount of words to the person that's living
1152
00:57:41.820 --> 00:57:44.199
in that house is going to change the fact that
1153
00:57:44.239 --> 00:57:46.530
you failed to repair their air conditioner.
1154
00:57:47.010 --> 00:57:49.550
Guys, do you really want to end a software development
1155
00:57:49.570 --> 00:57:51.449
show on a, hey, you should have entered the trades?
1156
00:57:51.769 --> 00:57:51.829
No.
1157
00:57:52.429 --> 00:57:53.570
Because I don't think you think that.
1158
00:57:53.690 --> 00:57:56.190
I just had to make that point.
1159
00:57:56.449 --> 00:57:56.650
Yeah.
1160
00:57:56.750 --> 00:57:59.309
Because, you know, people are afraid of their jobs and stuff.
1161
00:57:59.369 --> 00:58:01.929
And I'm going into high schools a few times a
1162
00:58:01.969 --> 00:58:04.449
year and I talk about careers. And I always mention
1163
00:58:04.590 --> 00:58:07.929
the trades for exactly that reason. Like, don't believe the hype.
1164
00:58:07.989 --> 00:58:08.170
No.
1165
00:58:08.190 --> 00:58:10.010
Because the four-year programs are so broken.
1166
00:58:10.929 --> 00:58:14.010
I know people who gave up on software development in
1167
00:58:14.030 --> 00:58:14.559
their 40s.
1168
00:58:14.739 --> 00:58:14.920
Sure.
1169
00:58:15.199 --> 00:58:20.510
And went and did trades-based hands-on stuff because they... They
1170
00:58:20.530 --> 00:58:22.449
got burned out of all the constant changes.
1171
00:58:24.369 --> 00:58:27.050
If you don't like that, you're having a tough time.
1172
00:58:27.070 --> 00:58:28.550
I think the people that are having the toughest time
1173
00:58:28.690 --> 00:58:31.329
in this current wave are the folks that jumped on
1174
00:58:31.369 --> 00:58:35.739
a stack like. NET 10, 15 years ago and been
1175
00:58:35.780 --> 00:58:38.719
able to live in that stack alone the whole time.
1176
00:58:39.599 --> 00:58:41.119
Now we're being disrupted in a way they've never been
1177
00:58:41.139 --> 00:58:46.050
disrupted before. We're older. We were disrupted by. NET and
1178
00:58:46.110 --> 00:58:48.380
jumped into it. So the fact that we're being disrupted again,
1179
00:58:48.400 --> 00:58:52.099
not as big a deal because we're used to being disrupted.
1180
00:58:52.340 --> 00:58:56.619
Yeah. Yeah. It's one of the metaphors that I tend
1181
00:58:56.639 --> 00:58:59.059
to use. It's probably a good thing to kind of
1182
00:58:59.380 --> 00:59:01.219
end up on since I know we're getting close to
1183
00:59:01.260 --> 00:59:04.829
the end of the time. Our entire industry has always
1184
00:59:04.849 --> 00:59:09.929
been about change. But there's different degrees of change.
1185
00:59:10.429 --> 00:59:10.630
Sure.
1186
00:59:11.030 --> 00:59:15.170
Normal change in our industry is like, I compare it
1187
00:59:15.230 --> 00:59:19.789
to waking up one day and your spouse has been
1188
00:59:19.829 --> 00:59:22.550
replaced by a different person. Right. Yeah. And you just
1189
00:59:22.590 --> 00:59:24.739
have to get to learn to love the new spouse
1190
00:59:26.539 --> 00:59:29.739
and learn their foibles. But you do it and you
1191
00:59:29.780 --> 00:59:33.239
adapt and you go on. And then every once in
1192
00:59:33.260 --> 00:59:37.789
a while, every 10, 15, 20 years, We get a
1193
00:59:37.809 --> 00:59:40.500
change that's more like you wake up one day and
1194
00:59:40.539 --> 00:59:42.539
you're married to a reptilian alien.
1195
00:59:42.840 --> 00:59:44.579
Yeah, I've heard that before.
1196
00:59:44.619 --> 00:59:47.739
You're not even sure what species it is. And AI
1197
00:59:47.780 --> 00:59:49.369
is one of those changes. And we went through one
1198
00:59:49.389 --> 00:59:53.070
of those changes in the 90s where everything changed at once,
1199
00:59:53.489 --> 00:59:57.840
where you went to GUIs, event-driven architectures, relational databases, and
1200
00:59:58.329 --> 01:00:00.750
And all of the things that the people that were
1201
01:00:00.809 --> 01:00:02.130
coming into that world. Right.
1202
01:00:02.269 --> 01:00:03.550
And then got slammed with the internet.
1203
01:00:03.639 --> 01:00:05.300
And some of them couldn't do that.
1204
01:00:05.340 --> 01:00:06.539
Yeah. Right after that.
1205
01:00:06.599 --> 01:00:09.300
And then the internet hit just as that was just
1206
01:00:09.320 --> 01:00:13.170
getting established. So that was the 90s was that reptilian
1207
01:00:13.250 --> 01:00:15.010
alien change. And we're in it again.
1208
01:00:15.250 --> 01:00:18.130
I just like to compare it to the ship for Agile, where, no,
1209
01:00:18.170 --> 01:00:21.139
the way you're thinking is incorrect about how you're building software.
1210
01:00:21.179 --> 01:00:23.139
You can tell about these shifts by looking at the
1211
01:00:23.179 --> 01:00:25.619
history of. NET Rocks and finding the threads that go
1212
01:00:25.639 --> 01:00:29.159
through different eras and different periods. We're always talking about
1213
01:00:29.239 --> 01:00:30.619
Agile there for a while, right?
1214
01:00:30.980 --> 01:00:33.159
And then it just became software development. Right.
1215
01:00:34.019 --> 01:00:34.159
Yeah.
1216
01:00:34.840 --> 01:00:37.440
Billy, it's been such a delight having you back on
1217
01:00:37.460 --> 01:00:39.760
the show and talking to you and seeing you in person.
1218
01:00:39.840 --> 01:00:41.460
It's been great.
1219
01:00:41.699 --> 01:00:43.900
I've loved it as I always do because, I mean,
1220
01:00:43.920 --> 01:00:45.800
I walk around the house ranting about stuff all the
1221
01:00:45.820 --> 01:00:46.460
time to myself.
1222
01:00:46.780 --> 01:00:48.440
It's good to rant with us then.
1223
01:00:48.500 --> 01:00:51.380
It's having a public venue in which to get some
1224
01:00:51.420 --> 01:00:54.059
of these rants out into the wider world is a pleasure.
1225
01:00:54.079 --> 01:00:56.949
Well, when I go out into the wider world, I
1226
01:00:57.170 --> 01:01:00.230
often have those rants of yours in my head when
1227
01:01:00.269 --> 01:01:01.650
I see dumb UX.
1228
01:01:01.889 --> 01:01:02.130
Nice.
1229
01:01:02.670 --> 01:01:07.960
Elevator buttons, you know. towel dispensers over the toilet.
1230
01:01:09.860 --> 01:01:11.679
I used to be perfectly happy with the ways the
1231
01:01:11.699 --> 01:01:13.400
world was laid out. Then Billy told me how it
1232
01:01:13.440 --> 01:01:16.000
was all wrong. All the time.
1233
01:01:16.239 --> 01:01:17.800
Well, it's not so much that it was all wrong,
1234
01:01:17.840 --> 01:01:22.360
but it could be a lot better. And, and, People
1235
01:01:22.380 --> 01:01:26.340
that sort of think everything is just good enough. No,
1236
01:01:26.380 --> 01:01:27.889
I've never been like that. I've always wanted the world
1237
01:01:27.909 --> 01:01:28.320
to get better.
1238
01:01:28.639 --> 01:01:28.969
Awesome.
1239
01:01:29.130 --> 01:01:31.309
Well, I'm going to call the show Arguing About AI.
1240
01:01:31.570 --> 01:01:31.809
Okay.
1241
01:01:31.869 --> 01:01:32.070
Good.
1242
01:01:32.309 --> 01:01:32.869
That sounds good.
1243
01:01:32.909 --> 01:01:34.090
Because I think that's the right name.
1244
01:01:34.150 --> 01:01:35.949
All sides were represented.
1245
01:01:35.969 --> 01:01:39.150
I don't know if that's true, but we all made
1246
01:01:39.170 --> 01:01:40.150
some noises anyway.
1247
01:01:40.190 --> 01:01:42.110
All right, Billy. Thank you very much. And we'll talk
1248
01:01:42.130 --> 01:01:44.210
to you again. And we'll talk to you, dear listener,
1249
01:01:44.250 --> 01:01:44.940
next time on.
1250
01:01:45.190 --> 01:01:46.010
NET Rocks.
1251
01:02:06.960 --> 01:02:09.679
NET Rocks is brought to you by Franklin's Net and
1252
01:02:09.719 --> 01:02:14.440
produced by Plop Studios, a full-service audio, video, and post-production
1253
01:02:14.500 --> 01:02:18.480
facility located physically in New London, Connecticut, and, of course,
1254
01:02:18.539 --> 01:02:22.139
in the cloud, online at pwop.com.
1255
01:02:22.889 --> 01:02:30.789
Visit our website at dotnetrocks.com for RSS feeds, downloads, mobile apps, comments,
1256
01:02:31.210 --> 01:02:33.690
and access to the full archives going back to show
1257
01:02:33.730 --> 01:02:36.460
number one, recorded in September 2002.
1258
01:02:36.460 --> 01:02:39.760
And make sure you check out our sponsors. They keep
1259
01:02:39.840 --> 01:02:40.480
us in business.
1260
01:02:41.079 --> 01:02:43.460
Now go write some code. See you next time.