WEBVTT - AI Doesn't Work and We All Know It Doesn't Work

0:00:02.490 --> 0:00:07.350
<v Speaker 1>Hello, and welcome to Cool People Who Did Cool Stuff,

0:00:07.390 --> 0:00:10.710
<v Speaker 1>your weekly reminder that when there's bad stuff that's happened

0:00:10.750 --> 0:00:14.890
<v Speaker 1>in history, there's people trying to do good things. Except

0:00:15.610 --> 0:00:18.680
<v Speaker 1>this week is an essay week. This week is an

0:00:18.730 --> 0:00:21.750
<v Speaker 1>essay week because I wanted to tell you Monday's essay

0:00:22.120 --> 0:00:27.860
<v Speaker 1>so badly that I broke from the regular world of

0:00:28.220 --> 0:00:32.229
<v Speaker 1>the way that I do this show. But now it's Wednesday.

0:00:32.659 --> 0:00:34.140
<v Speaker 1>Now it's Wednesday, and so I'm going to read you

0:00:34.159 --> 0:00:36.660
<v Speaker 1>a different essay, and another one that I think is timely,

0:00:37.220 --> 0:00:40.200
<v Speaker 1>even though I wrote it a couple months ago. Today,

0:00:40.220 --> 0:00:43.500
<v Speaker 1>I'm going to talk about AI. I'm sure I'm going

0:00:43.520 --> 0:00:47.330
<v Speaker 1>to talk about AI more in the future, although it's

0:00:47.350 --> 0:00:51.970
<v Speaker 1>not exactly a history topic, unless I do an overview

0:00:51.990 --> 0:00:54.350
<v Speaker 1>of how AI has been represented in fiction, but that's

0:00:54.370 --> 0:00:57.450
<v Speaker 1>more of a book club thing. I have too many podcasts.

0:00:58.310 --> 0:01:00.310
<v Speaker 1>But today I'm going to read you this essay about

0:01:00.510 --> 0:01:03.170
<v Speaker 1>AI because AI is in the news again. It's kind

0:01:03.190 --> 0:01:05.250
<v Speaker 1>of always been in the news for the past several years.

0:01:05.830 --> 0:01:07.970
<v Speaker 1>But it's in the news again because it's all going

0:01:07.990 --> 0:01:11.490
<v Speaker 1>to kill us all or whatever thing. And there's a

0:01:11.500 --> 0:01:14.080
<v Speaker 1>lot of argument back and forth about whether or not

0:01:14.120 --> 0:01:17.020
<v Speaker 1>that's all just a PR move, a cynical way to

0:01:17.420 --> 0:01:21.220
<v Speaker 1>get people to believe that AI is something that isn't.

0:01:22.200 --> 0:01:25.200
<v Speaker 1>And I've been thinking a lot about AI and its

0:01:25.760 --> 0:01:28.780
<v Speaker 1>essentially false critics. You know, I'll see things going around

0:01:28.819 --> 0:01:34.059
<v Speaker 1>that say, well, the negative social effects of AI don't

0:01:34.140 --> 0:01:37.060
<v Speaker 1>make it worth it, the fact that Google is a

0:01:37.080 --> 0:01:40.700
<v Speaker 1>little better now. And the thing is, I don't like

0:01:40.720 --> 0:01:45.280
<v Speaker 1>when people post that stuff because Google isn't better now.

0:01:46.280 --> 0:01:48.720
<v Speaker 1>I stopped using it entirely, and not just because of

0:01:48.780 --> 0:01:52.350
<v Speaker 1>surveillance concerns, but because when they move to AI forward,

0:01:53.130 --> 0:02:00.900
<v Speaker 1>it doesn't work. You can't do... real work by using AI.

0:02:02.080 --> 0:02:05.120
<v Speaker 1>It's very frustrating because we wish we could, right? We

0:02:05.160 --> 0:02:07.280
<v Speaker 1>wish we could ask a computer being like, hey, can

0:02:07.330 --> 0:02:10.590
<v Speaker 1>you answer this question for me really quickly? But sometimes

0:02:10.630 --> 0:02:13.350
<v Speaker 1>I'll dig into the sources because when I do search things,

0:02:13.389 --> 0:02:15.790
<v Speaker 1>you know, the AI automated response will come up at

0:02:15.850 --> 0:02:18.710
<v Speaker 1>the top because I'm not always bothering to type in

0:02:18.730 --> 0:02:21.329
<v Speaker 1>whatever you need to type to make it not do that.

0:02:21.950 --> 0:02:25.590
<v Speaker 1>And I'll look at the AI summary sometimes and I'm

0:02:25.650 --> 0:02:27.989
<v Speaker 1>often searching things that I have a fair amount of

0:02:28.169 --> 0:02:32.330
<v Speaker 1>contextual knowledge around. And so I'll often see all of

0:02:32.389 --> 0:02:36.490
<v Speaker 1>the things that are wrong. And sometimes I'll look up

0:02:36.550 --> 0:02:38.760
<v Speaker 1>a thing and it'll give me a solid answer and

0:02:38.800 --> 0:02:41.340
<v Speaker 1>it'll give me two sources that it comes from. And

0:02:41.380 --> 0:02:45.000
<v Speaker 1>I'll go and read those sources. And what it told

0:02:45.060 --> 0:02:49.700
<v Speaker 1>me won't be in those sources. It's not just that

0:02:49.740 --> 0:02:53.930
<v Speaker 1>it's replicating incorrect information that people have put out into

0:02:53.970 --> 0:02:59.410
<v Speaker 1>the world. It's that it's not replicating information at all.

0:03:01.370 --> 0:03:07.550
<v Speaker 1>AI doesn't work is the central thesis of this particular essay.

0:03:08.870 --> 0:03:12.030
<v Speaker 1>People are losing their minds. More and more people I

0:03:12.070 --> 0:03:18.060
<v Speaker 1>know are getting sucked into the AI delusion. And it

0:03:18.090 --> 0:03:22.000
<v Speaker 1>makes me really sad to kind of be self-critical for

0:03:22.040 --> 0:03:25.530
<v Speaker 1>a moment and thinking a lot even about the way

0:03:25.570 --> 0:03:28.930
<v Speaker 1>we create micro realities, right? We've created micro realities. We

0:03:28.970 --> 0:03:31.070
<v Speaker 1>used to have subcultures. We used to have a culture

0:03:31.190 --> 0:03:33.570
<v Speaker 1>and then we had subcultures, but now we have micro

0:03:33.610 --> 0:03:36.730
<v Speaker 1>cultures where our echo chambers have gotten smaller and smaller

0:03:36.810 --> 0:03:40.210
<v Speaker 1>as cultures, right? And that's always been a problem. And

0:03:40.230 --> 0:03:43.470
<v Speaker 1>the internet has made that a problem. It's kind of interesting.

0:03:43.490 --> 0:03:45.690
<v Speaker 1>There's some interesting stuff that comes out of that, but

0:03:45.710 --> 0:03:50.190
<v Speaker 1>it has also led people to rather dramatic beliefs. You

0:03:50.250 --> 0:03:55.510
<v Speaker 1>could basically refer to the mass shooters in American society

0:03:55.670 --> 0:03:59.070
<v Speaker 1>as part of a microculture, right? Where they're online encouraging

0:03:59.130 --> 0:04:04.200
<v Speaker 1>each other about these things. But AI breaks even that

0:04:04.400 --> 0:04:07.060
<v Speaker 1>because it takes the echo chamber down to an echo

0:04:07.120 --> 0:04:11.100
<v Speaker 1>chamber of one, right? You can now talk essentially to

0:04:11.150 --> 0:04:14.110
<v Speaker 1>yourself and have your own ideas reflected back at you.

0:04:14.710 --> 0:04:17.350
<v Speaker 1>And that can cause cognitive distortions and you can go

0:04:17.470 --> 0:04:21.080
<v Speaker 1>further and further afield And on some level, when hermits

0:04:21.140 --> 0:04:24.060
<v Speaker 1>do this, when people go into isolation and they stop

0:04:24.339 --> 0:04:27.870
<v Speaker 1>believing the things that the rest of society believes, you

0:04:27.890 --> 0:04:30.930
<v Speaker 1>sometimes get interesting stuff out of that, right? There's a

0:04:31.830 --> 0:04:35.270
<v Speaker 1>kind of connection you can make essentially to nature and

0:04:35.290 --> 0:04:40.710
<v Speaker 1>to the universe by choosing isolation. I'm not encouraging this, right?

0:04:41.170 --> 0:04:44.039
<v Speaker 1>But that is a thing that has produced something that

0:04:44.100 --> 0:04:47.690
<v Speaker 1>looks like wisdom throughout history. Although I think most of

0:04:47.710 --> 0:04:49.849
<v Speaker 1>the real wisdom is when you bring it back and

0:04:50.010 --> 0:04:55.800
<v Speaker 1>then everyone runs through it and we iterate ideas. But instead,

0:04:57.480 --> 0:05:00.820
<v Speaker 1>it's like a fairytale magic mirror where you say, mirror,

0:05:00.839 --> 0:05:02.940
<v Speaker 1>mirror on the wall, who's the fairest of them all?

0:05:02.980 --> 0:05:06.670
<v Speaker 1>And Claude or ChatGPT says back to you, whatever you

0:05:06.770 --> 0:05:09.250
<v Speaker 1>want to hear. And then if you say, oh, that's wrong,

0:05:09.290 --> 0:05:11.990
<v Speaker 1>and says, oh, you're right. Let me try again. I'll

0:05:12.029 --> 0:05:17.600
<v Speaker 1>do better. The impacts of that are going to be monumental.

0:05:18.040 --> 0:05:19.660
<v Speaker 1>But the reason that I want to be critical of

0:05:19.720 --> 0:05:22.560
<v Speaker 1>my own work as relates to this is that I

0:05:22.660 --> 0:05:27.360
<v Speaker 1>think that podcasts are a little bit dangerous too. I

0:05:27.680 --> 0:05:30.670
<v Speaker 1>think we all know people who, you know, the equivalent

0:05:30.690 --> 0:05:32.550
<v Speaker 1>of like my parents used to be normal and then

0:05:32.589 --> 0:05:35.690
<v Speaker 1>they watched too much Fox TV, right? They stepped into

0:05:35.730 --> 0:05:39.089
<v Speaker 1>an echo chamber where certain ideas were reflected back and

0:05:39.110 --> 0:05:41.750
<v Speaker 1>they spun further and further afield. I think that the

0:05:41.770 --> 0:05:45.280
<v Speaker 1>current rise of fascism is related to this whole thing

0:05:45.790 --> 0:05:47.489
<v Speaker 1>But I have other friends who will fall down different

0:05:47.529 --> 0:05:51.450
<v Speaker 1>rabbit holes and will end up being like leaving a

0:05:52.430 --> 0:05:55.510
<v Speaker 1>culture that we built by connecting with each other or

0:05:56.070 --> 0:05:58.860
<v Speaker 1>a political practice that we built by connecting to each other,

0:05:58.900 --> 0:06:02.039
<v Speaker 1>like the social movements I've been part of in my years.

0:06:02.100 --> 0:06:05.060
<v Speaker 1>We're built by people who are connected to each other

0:06:05.140 --> 0:06:08.500
<v Speaker 1>and talking to each other. And when you get your

0:06:08.560 --> 0:06:17.039
<v Speaker 1>politics through influencers... through podcasts, through YouTube, through TikTok, you're

0:06:17.080 --> 0:06:19.169
<v Speaker 1>going to get a different set of ideas and you're

0:06:19.210 --> 0:06:21.870
<v Speaker 1>just going to have a few voices in your head.

0:06:21.890 --> 0:06:26.610
<v Speaker 1>And it's not a conversation, right? You're not participating in

0:06:26.630 --> 0:06:30.850
<v Speaker 1>a conversation by listening to this. You should go participate

0:06:30.870 --> 0:06:33.970
<v Speaker 1>in conversations with other people about the things that you

0:06:33.990 --> 0:06:38.080
<v Speaker 1>hear in podcasts. I worry about this for myself too.

0:06:38.180 --> 0:06:41.050
<v Speaker 1>I worry both as a podcaster, but I also I

0:06:41.070 --> 0:06:42.830
<v Speaker 1>spend a lot of my time alone and I spend

0:06:42.870 --> 0:06:45.150
<v Speaker 1>a lot of my time listening to podcasts. And I

0:06:45.570 --> 0:06:49.680
<v Speaker 1>like that. I like that life. I like cleaning while

0:06:49.779 --> 0:06:53.740
<v Speaker 1>listening to Bridget Todd. I like everything Jamie Loftus has

0:06:53.779 --> 0:06:56.560
<v Speaker 1>ever done, right? But we need to be reflecting on

0:06:56.600 --> 0:07:01.220
<v Speaker 1>those ideas that we hear from people with actual people

0:07:01.279 --> 0:07:04.310
<v Speaker 1>in our lives. And I need to get better at it.

0:07:04.790 --> 0:07:08.330
<v Speaker 1>Maybe you do. But AI is just a whole other

0:07:08.570 --> 0:07:13.010
<v Speaker 1>level of creating cognitive distortions with its tiny echo chamber effect.

0:07:13.500 --> 0:07:16.480
<v Speaker 1>And I'll write more about that probably in the future.

0:07:17.760 --> 0:07:20.160
<v Speaker 1>But for now, I'm going to read you an essay

0:07:20.180 --> 0:07:22.940
<v Speaker 1>about why AI doesn't work and we all know it

0:07:22.980 --> 0:07:28.900
<v Speaker 1>doesn't work. The essay is called AI Doesn't Work and

0:07:28.930 --> 0:07:32.230
<v Speaker 1>We All Know It Doesn't Work. But do you know

0:07:32.250 --> 0:07:36.550
<v Speaker 1>what does work? Products and services, they may work. I

0:07:36.570 --> 0:07:45.720
<v Speaker 1>don't know. And we're back. Okay, here's the essay. AI

0:07:45.760 --> 0:07:48.180
<v Speaker 1>doesn't work and we all know it doesn't work. This

0:07:48.200 --> 0:07:51.480
<v Speaker 1>was published on May 20th, 2026. So there's going to

0:07:51.500 --> 0:07:54.620
<v Speaker 1>be some references to the news that had been happening.

0:07:55.680 --> 0:07:57.780
<v Speaker 1>And it's so funny because these things that seemed like

0:07:57.840 --> 0:08:00.660
<v Speaker 1>such big news at the time are now, you know,

0:08:00.940 --> 0:08:03.240
<v Speaker 1>so far in the past, if we spend too much

0:08:03.280 --> 0:08:04.810
<v Speaker 1>of our time online and too much of our time

0:08:04.850 --> 0:08:10.060
<v Speaker 1>reading the news. Last week, a bunch of college graduates

0:08:10.440 --> 0:08:14.010
<v Speaker 1>walking across the stage to get their diplomas didn't have

0:08:14.050 --> 0:08:17.970
<v Speaker 1>their names called because the university had outsourced the task

0:08:18.030 --> 0:08:21.490
<v Speaker 1>of reading their names out to an AI. And AI

0:08:22.190 --> 0:08:25.670
<v Speaker 1>doesn't work. If those graduates had done work at the

0:08:25.710 --> 0:08:31.750
<v Speaker 1>same efficacy as AI, they wouldn't be graduates. Society is

0:08:31.800 --> 0:08:35.380
<v Speaker 1>in the process of restructuring itself around a tool that

0:08:35.460 --> 0:08:39.550
<v Speaker 1>simply doesn't work. We all know it doesn't work. If

0:08:39.590 --> 0:08:41.990
<v Speaker 1>you needed to build a bridge, you wouldn't hire a

0:08:42.050 --> 0:08:45.689
<v Speaker 1>structural engineer who gets it right about 70% of the time.

0:08:46.730 --> 0:08:49.550
<v Speaker 1>You wouldn't read a history book that is 30% fiction,

0:08:49.910 --> 0:08:53.990
<v Speaker 1>but doesn't tell you what 30%. You shouldn't date someone

0:08:54.030 --> 0:08:56.530
<v Speaker 1>who always tells you that you're right, who lies to

0:08:56.570 --> 0:09:00.890
<v Speaker 1>you with a smile on his lips. My grandfather was

0:09:00.929 --> 0:09:03.600
<v Speaker 1>a naval architect for the U.S. Navy, and he fought

0:09:03.640 --> 0:09:06.060
<v Speaker 1>in World War II. And I talk about him a lot.

0:09:06.380 --> 0:09:08.740
<v Speaker 1>because I miss him all the time. And he used

0:09:08.780 --> 0:09:12.580
<v Speaker 1>to laugh and say, close enough for government work, when

0:09:12.620 --> 0:09:17.000
<v Speaker 1>something was done good enough, but not perfect. So my

0:09:17.040 --> 0:09:20.469
<v Speaker 1>dad says that now, and I say it too. But

0:09:21.610 --> 0:09:25.010
<v Speaker 1>it's a joke. Every time my grandfather designed a ship,

0:09:25.470 --> 0:09:27.530
<v Speaker 1>he would ride it into the worst storms he could

0:09:27.590 --> 0:09:30.650
<v Speaker 1>find and stand up on the deck and face the weather,

0:09:31.330 --> 0:09:34.100
<v Speaker 1>because why would he build something he didn't believe was

0:09:34.150 --> 0:09:38.810
<v Speaker 1>good enough? Elon Musk won't get on a Grok-designed rocket

0:09:38.850 --> 0:09:42.030
<v Speaker 1>to Mars, but I'm sure he'd be happy enough to

0:09:42.070 --> 0:09:46.660
<v Speaker 1>put you and I on it. The capitalists and authoritarians

0:09:46.720 --> 0:09:50.120
<v Speaker 1>who run this world are perfectly happy to replace us

0:09:50.300 --> 0:09:53.710
<v Speaker 1>all with robots that fuck up all the time, because

0:09:53.730 --> 0:09:58.390
<v Speaker 1>they'd rather accept shoddy work than pay people. It's bad,

0:09:59.150 --> 0:10:01.360
<v Speaker 1>and I'm mad about it. it, and you should be

0:10:01.400 --> 0:10:06.220
<v Speaker 1>mad about it too. I've seen people I care about

0:10:06.460 --> 0:10:10.620
<v Speaker 1>fall for AI delusions, believing themselves to be doing groundbreaking

0:10:10.679 --> 0:10:14.520
<v Speaker 1>research with the help of AI. My colleague Robert Evans

0:10:14.580 --> 0:10:18.170
<v Speaker 1>did an excellent series of podcast episodes about How AI

0:10:18.240 --> 0:10:21.260
<v Speaker 1>winds up replicating methods of cult leaders to convince its

0:10:21.400 --> 0:10:25.459
<v Speaker 1>users that they're geniuses. It's worth going back and listening to.

0:10:26.400 --> 0:10:28.360
<v Speaker 1>There's a new book out by this hack of a

0:10:28.400 --> 0:10:30.699
<v Speaker 1>journalist who might not have always been a hack of

0:10:30.760 --> 0:10:33.890
<v Speaker 1>a journalist, but has become one thanks to his reliance

0:10:33.970 --> 0:10:38.830
<v Speaker 1>on AI. The book is called The Future of Truth,

0:10:39.250 --> 0:10:43.429
<v Speaker 1>How AI Reshapes Reality. And everyone who is involved in

0:10:43.490 --> 0:10:47.600
<v Speaker 1>publishing it should be fired. It's a book about how,

0:10:48.080 --> 0:10:52.550
<v Speaker 1>you know, AI reshapes reality. But it was written with

0:10:52.990 --> 0:10:56.510
<v Speaker 1>research help from an AI, and the New York Times

0:10:56.550 --> 0:10:59.930
<v Speaker 1>has revealed that it contains false statements attributed to real people,

0:11:00.429 --> 0:11:04.250
<v Speaker 1>which were hallucinated by AI. The author claims to be

0:11:04.270 --> 0:11:07.709
<v Speaker 1>an expert on truth as a subject matter, but in

0:11:07.750 --> 0:11:10.750
<v Speaker 1>his acknowledgement of the problem, he admitted the book contained

0:11:11.490 --> 0:11:18.430
<v Speaker 1>synthetic quotes. Lies. It contains lies. If I made up

0:11:18.530 --> 0:11:21.530
<v Speaker 1>quotes whole cloth on my podcast and attributed them to

0:11:21.630 --> 0:11:26.710
<v Speaker 1>real people, I would be fired with cause. I've long

0:11:26.750 --> 0:11:30.150
<v Speaker 1>been unpopular among my peers for my hardline ethics around

0:11:30.230 --> 0:11:33.670
<v Speaker 1>truth telling. I don't believe that it is ethical to

0:11:33.730 --> 0:11:36.410
<v Speaker 1>lie to anyone. It wouldn't be ethical for you to punch.

0:11:37.429 --> 0:11:39.750
<v Speaker 1>You don't always have to answer questions and you can

0:11:39.770 --> 0:11:43.130
<v Speaker 1>be evasive. But when you tell someone something that isn't true,

0:11:43.950 --> 0:11:48.209
<v Speaker 1>you are distorting their sense of reality. You are doing

0:11:48.270 --> 0:11:52.369
<v Speaker 1>it to control their thinking and behavior, the same as

0:11:52.410 --> 0:11:56.110
<v Speaker 1>when you hit somebody. It's important to keep in mind, though,

0:11:56.370 --> 0:11:59.500
<v Speaker 1>that I think violence is perfectly acceptable in a large

0:11:59.530 --> 0:12:03.180
<v Speaker 1>number of situations. If someone is attempting to exert power

0:12:03.280 --> 0:12:07.339
<v Speaker 1>over you, lying in self-defense is often safer and more

0:12:07.380 --> 0:12:11.400
<v Speaker 1>effective than resorting to violence. Or to put it plainly,

0:12:12.240 --> 0:12:15.950
<v Speaker 1>it's ethical to punch cops It's just not usually strategic.

0:12:17.870 --> 0:12:20.689
<v Speaker 1>I believe very strongly in the value of telling the truth.

0:12:21.590 --> 0:12:24.229
<v Speaker 1>When I read about the haymarket anarchists of Chicago in

0:12:24.270 --> 0:12:27.980
<v Speaker 1>the 1880s, those immigrants and the ex-slave who built the

0:12:28.010 --> 0:12:32.120
<v Speaker 1>modern labor movement, I'm struck by how every source refers

0:12:32.160 --> 0:12:37.120
<v Speaker 1>to their shocking honesty and forthrightness. I've been told it's

0:12:37.179 --> 0:12:43.100
<v Speaker 1>a neuroatypicality of mine, that I'm so justice-sensitive around lying,

0:12:43.950 --> 0:12:46.069
<v Speaker 1>And it makes me sad to imagine that my position

0:12:46.110 --> 0:12:49.350
<v Speaker 1>might not be a mainstream value. But the fact that

0:12:49.390 --> 0:12:52.309
<v Speaker 1>we are handing over control of our society to lying

0:12:52.350 --> 0:12:58.120
<v Speaker 1>machines is simply the most dystopian thing I can imagine. Water-hungry,

0:12:58.480 --> 0:13:02.579
<v Speaker 1>power-hungry machines are being built while weather gets warmer, while

0:13:02.620 --> 0:13:06.980
<v Speaker 1>power gets scarcer, while water becomes even more precious. It

0:13:07.000 --> 0:13:10.200
<v Speaker 1>feels essentially sacrilegious to me that our rulers offer up

0:13:10.280 --> 0:13:14.280
<v Speaker 1>endless gallons of water to these false machine gods. That

0:13:14.300 --> 0:13:17.179
<v Speaker 1>our rulers are sacrificing the water that we need for

0:13:17.260 --> 0:13:21.560
<v Speaker 1>our crops and our bodies. The people who have built

0:13:21.600 --> 0:13:26.330
<v Speaker 1>these machines aren't even offering us utopia in exchange. They're

0:13:26.370 --> 0:13:32.170
<v Speaker 1>offering us displacement. They're offering us dispossession. They're offering us

0:13:32.250 --> 0:13:36.000
<v Speaker 1>layoffs and poverty and famine. And they're essentially promising us

0:13:36.080 --> 0:13:39.520
<v Speaker 1>that thanks to their machines, we can become less intelligent.

0:13:39.920 --> 0:13:42.820
<v Speaker 1>That we can be saved the drudgery of making beautiful

0:13:42.900 --> 0:13:47.050
<v Speaker 1>things and living fulfilling lives. And instead of talking to

0:13:47.090 --> 0:13:50.209
<v Speaker 1>each other about our problems, we get to each talk

0:13:50.530 --> 0:13:55.010
<v Speaker 1>to a sycophant machine friend. I don't believe in conspiracies,

0:13:55.470 --> 0:13:58.210
<v Speaker 1>except for the ones that we've seen proven true. And

0:13:58.309 --> 0:14:02.010
<v Speaker 1>those are usually proven true quick enough. I don't actually

0:14:02.090 --> 0:14:05.000
<v Speaker 1>believe the rich are collectively plotting to just kill us

0:14:05.130 --> 0:14:09.000
<v Speaker 1>all after we've been rendered obsolete. but I suspect none

0:14:09.010 --> 0:14:13.219
<v Speaker 1>of them would mind if that's what happens. This whole

0:14:13.280 --> 0:14:16.179
<v Speaker 1>thing is very frustrating to me, in part because I'm

0:14:16.220 --> 0:14:20.220
<v Speaker 1>not anti-technology. I think eyeglasses and antibiotics are two of

0:14:20.260 --> 0:14:23.310
<v Speaker 1>the best things humanity has ever accomplished. And I felt

0:14:23.350 --> 0:14:26.750
<v Speaker 1>the same burst of pride as everyone else, not national pride,

0:14:26.790 --> 0:14:31.870
<v Speaker 1>but pride in humanity, when Artemis orbited the moon. There

0:14:31.890 --> 0:14:35.120
<v Speaker 1>are machine learning tasks that don't bother me. My problem

0:14:35.160 --> 0:14:37.600
<v Speaker 1>with Google Maps has far more to do with surveillance

0:14:37.680 --> 0:14:41.870
<v Speaker 1>than AI. I don't inherently hate a grammar check program,

0:14:42.450 --> 0:14:46.030
<v Speaker 1>though autocomplete and autocorrect limit human expression in ways that

0:14:46.050 --> 0:14:48.510
<v Speaker 1>will have knock-on effects on the development of language, I'm

0:14:48.960 --> 0:14:53.180
<v Speaker 1>willing to bet. I would be perfectly happy if I

0:14:53.200 --> 0:14:56.100
<v Speaker 1>had a computer research assistant in the Star Trek fashion.

0:14:56.600 --> 0:14:59.800
<v Speaker 1>Someone I could ask, how many people openly lived in

0:14:59.840 --> 0:15:03.860
<v Speaker 1>homosexual couples in Germany during the Protestant Reformation? Or something

0:15:03.880 --> 0:15:06.430
<v Speaker 1>like that. and let a computer comb through all the

0:15:06.500 --> 0:15:09.860
<v Speaker 1>human knowledge ever put to paper, well, to bits and bytes,

0:15:10.540 --> 0:15:15.280
<v Speaker 1>and give me an answer. But AI doesn't do that.

0:15:16.280 --> 0:15:21.080
<v Speaker 1>It functionally cannot, because it is not thinking. It is

0:15:21.120 --> 0:15:25.240
<v Speaker 1>a fancy autocomplete, designed specifically to convince us that it

0:15:25.340 --> 0:15:30.160
<v Speaker 1>is capable of independent reasoning and research. It isn't designed

0:15:30.180 --> 0:15:32.500
<v Speaker 1>to comb all of the sources and spit out a

0:15:32.560 --> 0:15:36.120
<v Speaker 1>correct answer. but to convince us that it has done so.

0:15:37.580 --> 0:15:40.180
<v Speaker 1>If I hired a research assistant and asked them a

0:15:40.220 --> 0:15:44.180
<v Speaker 1>question and they came back with an incomplete answer, I

0:15:44.200 --> 0:15:47.640
<v Speaker 1>could accept that. If I hired a research assistant and

0:15:47.700 --> 0:15:49.940
<v Speaker 1>asked them a question and they came back with a

0:15:49.980 --> 0:15:52.720
<v Speaker 1>bizarre mix of truth and things they simply made up

0:15:52.760 --> 0:15:57.170
<v Speaker 1>whole cloth without marking which is which, I would fire them.

0:15:58.710 --> 0:16:02.530
<v Speaker 1>Students who fabricate sources are rightly accused of cheating. Yet

0:16:02.590 --> 0:16:06.500
<v Speaker 1>when AI does it, we accept it? The knock-on effects

0:16:06.530 --> 0:16:09.840
<v Speaker 1>of this are incalculable, though you could ask an AI

0:16:09.880 --> 0:16:13.140
<v Speaker 1>to calculate them anyway if you don't mind false information.

0:16:14.440 --> 0:16:16.680
<v Speaker 1>As more and more of the information on the internet

0:16:16.800 --> 0:16:20.140
<v Speaker 1>is AI-generated, it becomes harder and harder to sort out

0:16:20.210 --> 0:16:24.130
<v Speaker 1>fact from fiction. AI is going to start feeding on AI,

0:16:24.430 --> 0:16:27.750
<v Speaker 1>and soon it will spiral further and further away from coherence.

0:16:28.930 --> 0:16:33.590
<v Speaker 1>As Cory Doctorow put it, quote, AI is the asbestos

0:16:33.650 --> 0:16:36.690
<v Speaker 1>we are shoveling into the walls of our society, and

0:16:36.730 --> 0:16:41.190
<v Speaker 1>our descendants will be digging it out for generations. The

0:16:41.230 --> 0:16:44.140
<v Speaker 1>Luddite movement of the early 19th century wasn't actually an

0:16:44.200 --> 0:16:49.360
<v Speaker 1>anti-technological movement per se, but rather a labor movement. Skilled

0:16:49.400 --> 0:16:53.160
<v Speaker 1>laborers were having their jobs automated away, and textile work

0:16:53.240 --> 0:16:57.440
<v Speaker 1>moved from the workshop, the cottage industry, literally done in

0:16:57.480 --> 0:17:02.180
<v Speaker 1>a cottage, to the factories. People went from living self-directed

0:17:02.240 --> 0:17:05.180
<v Speaker 1>lives to lives in which they themselves had to work

0:17:05.240 --> 0:17:09.350
<v Speaker 1>like cogs in a machine. So people revolted at a

0:17:09.390 --> 0:17:13.490
<v Speaker 1>massive scale and came close to overthrowing the early capitalist system.

0:17:14.830 --> 0:17:17.590
<v Speaker 1>But do you know what I would never overthrow? Do

0:17:17.630 --> 0:17:20.050
<v Speaker 1>you know what I owe my allegiance and my ability

0:17:20.090 --> 0:17:23.240
<v Speaker 1>to feed my dog food to? The products and services

0:17:23.300 --> 0:17:33.770
<v Speaker 1>that support this show. And we're back. I'm endlessly inspired

0:17:33.830 --> 0:17:37.169
<v Speaker 1>by the Luddites. But the thing is, those automatic looms,

0:17:37.850 --> 0:17:41.459
<v Speaker 1>at least those things, devils that they were, still actually worked.

0:17:42.380 --> 0:17:46.760
<v Speaker 1>Whatever their negative impacts on society, they actually, effectively, automated

0:17:46.800 --> 0:17:50.810
<v Speaker 1>the production of textiles. Margaret from the future, not from

0:17:50.830 --> 0:17:53.330
<v Speaker 1>the recording, but from having written the essay here, I've

0:17:53.350 --> 0:17:55.510
<v Speaker 1>actually learned since, or I've talked to people who've told

0:17:55.530 --> 0:17:57.909
<v Speaker 1>me since, and they haven't fully sourced this, because I

0:17:57.930 --> 0:17:59.710
<v Speaker 1>read several books on the Luddites, but I didn't read

0:17:59.850 --> 0:18:04.619
<v Speaker 1>as much about specifically the development of textiles. I've learned

0:18:04.670 --> 0:18:09.080
<v Speaker 1>since that actually the cloth that we're used to, the

0:18:09.100 --> 0:18:14.860
<v Speaker 1>cloth produced by the automatic looms, was worse, right? Like,

0:18:15.020 --> 0:18:18.540
<v Speaker 1>while it effectively automated the production of textiles so there

0:18:18.560 --> 0:18:22.859
<v Speaker 1>could be more textiles more easily, it didn't make... better

0:18:22.960 --> 0:18:25.679
<v Speaker 1>products or products even at the same quality as the

0:18:25.700 --> 0:18:28.500
<v Speaker 1>handcrafted stuff. And that makes sense to me. I think

0:18:28.520 --> 0:18:30.940
<v Speaker 1>that we're all in this moment in society where we're

0:18:30.980 --> 0:18:34.020
<v Speaker 1>recognizing that like, take for example, a hand knit sweater.

0:18:34.040 --> 0:18:39.580
<v Speaker 1>A hand knit sweater can't really have value in the

0:18:39.660 --> 0:18:44.459
<v Speaker 1>capitalist sense that is directly understood. Because if you were

0:18:44.480 --> 0:18:46.820
<v Speaker 1>to hand knit a sweater and then sell it for

0:18:46.840 --> 0:18:50.220
<v Speaker 1>a decent wage, you would be selling a sweater for

0:18:50.440 --> 0:18:55.160
<v Speaker 1>substantially more than the market can bear sweaters costing, right?

0:18:55.460 --> 0:18:59.620
<v Speaker 1>Because mass manufactured stuff is available. And it creates this

0:18:59.700 --> 0:19:04.980
<v Speaker 1>interesting effect where we're all starting to learn that the

0:19:05.020 --> 0:19:10.350
<v Speaker 1>handmade stuff is better, not just necessarily higher quality, right?

0:19:10.380 --> 0:19:12.629
<v Speaker 1>It might not even be higher quality in many cases.

0:19:12.670 --> 0:19:14.550
<v Speaker 1>Certainly if I handed a sweater, it would not be

0:19:14.609 --> 0:19:18.810
<v Speaker 1>higher quality, but it is more valuable to us, right?

0:19:19.160 --> 0:19:22.080
<v Speaker 1>Like a quilt made by your friend or your grandmother

0:19:22.160 --> 0:19:27.420
<v Speaker 1>or your grandson is a lot more valuable to a family.

0:19:27.600 --> 0:19:30.959
<v Speaker 1>It has a lot more sentimental value. And we use

0:19:30.980 --> 0:19:35.709
<v Speaker 1>the word sentimental value kind of pejoratively in our society. Like, oh,

0:19:35.750 --> 0:19:38.830
<v Speaker 1>this is value, but just sentimental value, right? Like not

0:19:39.030 --> 0:19:44.510
<v Speaker 1>real value, not materialist value, you know? That's nonsense. I

0:19:44.530 --> 0:19:48.179
<v Speaker 1>think we're all learning the limits of a rigid materialist

0:19:48.220 --> 0:19:51.439
<v Speaker 1>analysis of the world and of only looking for the

0:19:51.500 --> 0:19:55.100
<v Speaker 1>things that like improve our economy and our ability to

0:19:55.160 --> 0:19:58.460
<v Speaker 1>create things at scale or whatever. And we're all learning

0:19:58.619 --> 0:20:03.220
<v Speaker 1>the value of not that. I hope we're all learning that.

0:20:03.380 --> 0:20:07.399
<v Speaker 1>I've been learning it. Maybe you have been too. So anyway,

0:20:07.660 --> 0:20:11.419
<v Speaker 1>here's to the cottage industry. that we've lost. Back to

0:20:11.460 --> 0:20:17.680
<v Speaker 1>the essay. AI doesn't work. It can't even become some

0:20:17.790 --> 0:20:22.450
<v Speaker 1>all-powerful machine god that rules us all, benevolently or malevolently,

0:20:22.810 --> 0:20:24.750
<v Speaker 1>which is a hard sentence to say. It's hard to

0:20:24.770 --> 0:20:30.270
<v Speaker 1>say benevolently or malevolently. Either word alone, perfectly fine, but

0:20:30.330 --> 0:20:34.730
<v Speaker 1>both together. Anyway, it can't even become some all-powerful machine

0:20:34.760 --> 0:20:38.000
<v Speaker 1>god that rules us all because it isn't conscious and

0:20:38.020 --> 0:20:41.879
<v Speaker 1>it doesn't think. AI designers keep telling us they're going

0:20:41.900 --> 0:20:44.320
<v Speaker 1>to iron out all the bugs and that one day

0:20:44.340 --> 0:20:47.740
<v Speaker 1>AI will accomplish all the things it pretends to accomplish today.

0:20:48.700 --> 0:20:52.359
<v Speaker 1>I doubt this. It seems structurally incapable of what it

0:20:52.400 --> 0:20:55.820
<v Speaker 1>was promised to do. Okay, out of the essay again.

0:20:56.240 --> 0:20:59.159
<v Speaker 1>As for its recent promises of killing us all, that's

0:20:59.240 --> 0:21:02.360
<v Speaker 1>more of an open question. I think that on some

0:21:02.380 --> 0:21:06.270
<v Speaker 1>level we need to remember that these are physical machines

0:21:06.310 --> 0:21:09.959
<v Speaker 1>that require electricity and they can be unplugged. Just like,

0:21:11.400 --> 0:21:14.940
<v Speaker 1>you can just unplug them. They exist in a physical

0:21:15.000 --> 0:21:22.810
<v Speaker 1>location and they can be dismantled or just unplugged. But

0:21:23.220 --> 0:21:26.770
<v Speaker 1>that said, do I believe that AI will become self-aware

0:21:26.950 --> 0:21:29.210
<v Speaker 1>in a way and like seek to destroy us and

0:21:29.250 --> 0:21:34.270
<v Speaker 1>create Rocko's Basilisk and all that fucking nonsense? No. No

0:21:34.350 --> 0:21:38.090
<v Speaker 1>part of me thinks that that could possibly happen. But

0:21:38.790 --> 0:21:46.830
<v Speaker 1>Could something that simulates that exist? Maybe. And more importantly, like,

0:21:47.770 --> 0:21:53.320
<v Speaker 1>if I create a chainsaw on legs and I let

0:21:53.380 --> 0:21:57.540
<v Speaker 1>it go, just walking around, the carnage it causes isn't

0:21:57.580 --> 0:22:02.060
<v Speaker 1>because it's, like, smart, right? The carnage it causes is

0:22:02.100 --> 0:22:05.220
<v Speaker 1>because it's a chainsaw on legs walking around. And that's

0:22:05.240 --> 0:22:08.290
<v Speaker 1>what they're building right now. They're building chainsaws on legs,

0:22:08.390 --> 0:22:13.520
<v Speaker 1>and they're building them with targeting systems. So that's bad, right?

0:22:14.119 --> 0:22:17.760
<v Speaker 1>We can all agree that that's bad. Anyway, back to

0:22:17.800 --> 0:22:22.139
<v Speaker 1>the essay. What AI is capable of is what it's

0:22:22.260 --> 0:22:27.679
<v Speaker 1>already doing, replacing skilled labor with just barely good enough automation.

0:22:28.640 --> 0:22:31.600
<v Speaker 1>AI is capable of tricking people into thinking it's helping.

0:22:32.359 --> 0:22:35.540
<v Speaker 1>AI is capable of reducing our capacity for critical thinking.

0:22:36.460 --> 0:22:40.139
<v Speaker 1>AI is capable of replacing art, which is inherently expression,

0:22:40.960 --> 0:22:44.879
<v Speaker 1>with a hollow simulacrum. The issues of AI run deeper

0:22:44.920 --> 0:22:48.379
<v Speaker 1>than just who will hire illustrators in an era of AI,

0:22:49.140 --> 0:22:52.640
<v Speaker 1>though that alone is an important question. The issues run

0:22:52.680 --> 0:22:55.600
<v Speaker 1>about as deep as deep can go. AI is an

0:22:55.700 --> 0:23:02.060
<v Speaker 1>existential threat, and not some it-will-become-a-sentient-creature-and-destroy-us-all way, but an AI

0:23:02.140 --> 0:23:05.700
<v Speaker 1>is being used for military targeting way, and a AI

0:23:05.760 --> 0:23:09.760
<v Speaker 1>is destroying education way. And AI is a bubble that

0:23:09.780 --> 0:23:13.770
<v Speaker 1>will burst and bring down the world economy way. But

0:23:13.790 --> 0:23:18.690
<v Speaker 1>there's hope. And that hope looks like college students booing

0:23:18.770 --> 0:23:21.649
<v Speaker 1>speakers who try to sell them on AI. And it

0:23:21.690 --> 0:23:24.950
<v Speaker 1>looks like the massive and successful movement against data centers

0:23:25.390 --> 0:23:29.350
<v Speaker 1>that get proposed site after proposed site shut down. And

0:23:29.369 --> 0:23:32.750
<v Speaker 1>that hope might soon look like a militant neo-Luddite movement.

0:23:33.470 --> 0:23:36.270
<v Speaker 1>And based on the yard signs I see in rural America,

0:23:36.810 --> 0:23:42.649
<v Speaker 1>that movement might be, or become, bipartisan. Or rather, it

0:23:42.690 --> 0:23:46.460
<v Speaker 1>might be a class movement of the working class undivided

0:23:46.500 --> 0:23:51.540
<v Speaker 1>by culture war issues fighting against the ruling class. If so,

0:23:52.420 --> 0:23:58.200
<v Speaker 1>I'm here for it. Yeah, anyway, that's the essay. Make

0:23:58.220 --> 0:24:00.859
<v Speaker 1>of it what you will. Don't listen to it. Don't

0:24:01.060 --> 0:24:05.000
<v Speaker 1>listen to podcasts. Listen to them. Hear them. Don't assume

0:24:05.220 --> 0:24:07.780
<v Speaker 1>they're right. Maybe you hate listening to this whole thing

0:24:07.820 --> 0:24:11.550
<v Speaker 1>and you have a relationship with AI where you're better

0:24:11.590 --> 0:24:15.270
<v Speaker 1>at telling what is and isn't the truth and sorting

0:24:15.310 --> 0:24:18.230
<v Speaker 1>all of that out. I don't know. I'm probably not

0:24:18.250 --> 0:24:23.290
<v Speaker 1>going to convince you otherwise. But I do think we're

0:24:23.310 --> 0:24:26.869
<v Speaker 1>being suckered into being told it is good for certain

0:24:26.930 --> 0:24:30.660
<v Speaker 1>things It's possible that there's some use case that I'm

0:24:30.720 --> 0:24:33.580
<v Speaker 1>not thinking of. I don't know. I think that large

0:24:33.619 --> 0:24:38.270
<v Speaker 1>language models isn't inherently an evil technology. But claiming that

0:24:38.290 --> 0:24:42.050
<v Speaker 1>this is an intelligence is a problem. And it is

0:24:42.070 --> 0:24:45.790
<v Speaker 1>a problem when you convince people that they are talking

0:24:45.810 --> 0:24:50.510
<v Speaker 1>to something that has any sense of self or agency

0:24:51.260 --> 0:24:55.760
<v Speaker 1>or can do anything other than just reflect ideas back

0:24:55.820 --> 0:24:59.950
<v Speaker 1>through it. strange fairy tale magic mirror this whole thing

0:25:01.410 --> 0:25:03.880
<v Speaker 1>is going to be a fairy tale one day and

0:25:04.020 --> 0:25:07.760
<v Speaker 1>not one of the good fun modern fairy tales but

0:25:07.780 --> 0:25:11.560
<v Speaker 1>the dark original shit where children are led into the

0:25:11.580 --> 0:25:15.480
<v Speaker 1>dark of the woods and slaughtered and eaten it's much

0:25:15.550 --> 0:25:18.610
<v Speaker 1>more that kind of thing i'll be back next week

0:25:18.930 --> 0:25:22.310
<v Speaker 1>with more history which is usually what we do on

0:25:22.350 --> 0:25:24.350
<v Speaker 1>this show. But if you want to read more essays

0:25:24.390 --> 0:25:28.830
<v Speaker 1>like this, I write one essentially every week over at margaretkiljoy.substack.com.

0:25:28.930 --> 0:25:38.270
<v Speaker 1>All right. Bye. Cool People Who Did Cool Stuff is

0:25:38.330 --> 0:25:41.290
<v Speaker 1>a production of Cool Zone Media. For more podcasts from

0:25:41.310 --> 0:25:45.230
<v Speaker 1>Cool Zone Media, visit our website, coolzonemedia.com, or check us

0:25:45.300 --> 0:25:48.840
<v Speaker 1>out on the iHeartRadio app, Apple Podcasts, or wherever you

0:25:48.859 --> 0:25:49.680
<v Speaker 1>get your podcasts.