Dreaming Against the Machine - Episode 17： Weapons of Math Destruction, with Cathy O’Neil

00:00:00.000 Welcome back to Dreaming Against the Machine. I'm your host, Adam Becker. This week's guest

00:00:07.900 is Kathy O'Neill. She is a mathematician, an algorithmic auditor, and the author of

00:00:15.520 Weapons of Math Destruction, among other books and essays. Kathy is also a friend and

00:00:23.580 And was one of the people who first made me aware of the idea of algorithmic bias and the fact that computer systems do not just neutrally enforce rules, but are tools that are used to exercise power.

00:00:44.380 In fact, I remember back when her book, Weapons of Math Destruction, first came out, I went

00:00:51.600 to see her speak, and she said something that has stuck with me now for over a decade.

00:00:57.580 She said that a friend of hers asked her why she wrote a book about math, and she said,

00:01:03.440 I didn't write a book about math.

00:01:05.280 I wrote a book about power.

00:01:07.880 So with that, here's Kathy O'Neill.

00:01:14.380 kathy welcome to dreaming against the machine thank you adam i'm glad to be here so kathy

00:01:22.900 you i mean how do i even describe you i i say that a lot on this show i should probably

00:01:28.640 stop saying that because i know who my guests are i just i'm happy to introduce myself adam

00:01:34.580 why don't you introduce yourself i should just have my guests do that yeah i know more about

00:01:38.680 me than you do that's true yeah i mean i know a bit about you but yeah if i forget anything that

00:01:44.160 you want to mention then jump on in but um i'm a mathematician turned um finance quant turned

00:01:50.600 data scientist turned occupier turned author and now i have a company that audits algorithms for

00:01:57.720 harm writ large and most of my clients at this point are attorneys general and other like law

00:02:04.640 firms i basically work as an expert for lawsuits that sue companies for harming people using

00:02:11.260 algorithms. Okay, I am going to say something. One of your books is called Weapons of Math

00:02:17.620 Destruction, which in addition to being an excellent pun, is also an excellent book.

00:02:23.280 Thank you. 10 years ago. Yeah. God, really? 10 years ago?

00:02:27.780 Yeah, 10 years ago. Like 10 years ago, minus two months or something.

00:02:32.340 Wow. Okay. And how does that feel? Like the more things change,

00:02:36.680 the more they stay the same kind of feeling. Well, tell us what the book is about. I mean,

00:02:40.720 And again, you know more about it than I do.

00:02:43.040 I mean, I've read it, but it's been a minute.

00:02:45.620 It was a call to arms.

00:02:47.440 Yeah.

00:02:48.380 I was working first as a quant in finance, then as a data scientist.

00:02:54.140 I was also in Occupy at the time.

00:02:56.240 So I was learning from my fellow occupiers about how to think about things

00:03:00.420 through the lens of power and entrenched biases and prejudice.

00:03:09.320 And I was like, oh, yeah, this is exactly what's happening in data science.

00:03:12.860 Like algorithms are just propagating all sorts of, you know, capitalistic tendencies.

00:03:21.200 And they're just like going to eat up the world.

00:03:23.860 But we are all being asked to trust it.

00:03:25.860 We're all, in fact, then it was called big data.

00:03:28.540 It's like such an old term.

00:03:29.900 That's like the oldest thing about my book is like the subtitle is how big data increases inequality and threatens democracy.

00:03:36.620 I remember big data. Yeah. Yeah. But at the time I wrote it, I like every fucking I'm allowed to swear on this podcast, right? Yes. I am from New Jersey. You can swear on this podcast because I'm a big I'm big on it on the swearing words. Yeah, no, that's one of the reasons we're friends. Yeah. Okay, cool. Yeah. No, I mean, like everything was like all the generalistic articles, all the books were just like big data is going to make everything great.

00:04:05.800 and it's going to make everything fair and it's going to be so wonderful um and and you know what

00:04:11.720 like my book was the first of many like absolutely fucking not books um and i've been really happy to

00:04:18.480 see the absolutely fucking not books coming out but but i feel like we're so in some sense lots 0.76

00:04:23.480 of things have changed for the better like we don't trust algorithms but we are hearing the 0.83

00:04:28.700 same kind of like crappy hype stuff marketing like everything's gonna be great with the ai stuff now

00:04:34.920 So I also feel like, really? Again? Not that we trust it. I mean, again, I think the thing I wanted to make sure people were aware of is that these things are not trustworthy. We should not hand over our autonomy and our agency and our big, important moral decisions. We just shouldn't hand that all over to capitalistic machines. And I feel like we still know that, but it's happening anyway.

00:05:02.720 So one of the things I want to talk about is what we can do about that. But before we jump into that, I do want to just ask something I realized I don't know about you, which is you worked as a quant. You worked for an investment firm, and then you ended up in Occupy Wall Street. Can you explain that journey a little bit?

00:05:29.940 Well, I guess the easiest way of explaining it is that I was like unbelievably naive.

00:05:35.500 You know, I didn't know at anything.

00:05:38.400 Like I was a hedge fund quant at D.E. Shaw.

00:05:41.100 And that was back in the time in 2006 is when I applied and got the job.

00:05:45.680 And I started in 2007, early 2007, before the market started getting turbulent, which they did in August, like the month after I actually joined the firm.

00:05:57.500 But I was a professor at the time.

00:05:59.080 So like I delayed starting for a semester to finish off the academic year.

00:06:06.380 But back then, DE Shaw was just emailing random people in math departments being like,

00:06:11.580 you want to do stuff with the market and make markets more efficient?

00:06:15.280 You'd be like making a lot more money than you are now and you'd be improving the world.

00:06:19.960 And I was like, okay, that sounds great.

00:06:24.280 And I remember thinking to myself, I want to have a more impact in the world

00:06:28.800 because right now, like, I work in number theory.

00:06:31.260 Oh, boy.

00:06:32.300 And the number, like, you know, I write a paper,

00:06:35.160 and there's seven people who care about it

00:06:37.100 know about it within a week, as I tell them,

00:06:39.220 and then it takes five years to publish the paper.

00:06:41.580 You know what I mean?

00:06:42.020 It was just, like, unbelievably slow-moving,

00:06:45.520 and, like, I just felt impotent,

00:06:47.360 and I was like, I want to have a little bit more effect on the world.

00:06:51.060 And I thought, the naivete, of course, was, like,

00:06:55.660 I forgot to say positive effect.

00:06:58.220 I just went into it being like, ooh, we're going to make markets more efficient.

00:07:03.060 I totally, 100% was like, whatever Alan Greenspan and Larry Summers are saying must be true because they're so smart.

00:07:11.840 Oh, boy.

00:07:12.360 It was gross. 0.99

00:07:13.920 Looking back on it, I want to slap my face. 0.90

00:07:16.820 But it only took me a couple of years to be like, wow, this is really gross. 0.89

00:07:21.080 This is super gross.

00:07:22.300 And I feel literally sick to my stomach every morning.

00:07:25.940 it got to the point where like i had this a major internal conflict what's it called um when you

00:07:33.940 like have two thoughts at the same time dissonance yeah cognitive dissonance thank you yeah and where

00:07:39.920 i was just like well this is a good job i'm learning a lot it's really interesting the

00:07:43.020 people i talk to are absolutely fascinating because they're so crazy weird like ayn rand

00:07:48.340 preppers you know like well they were listening to alan greenspan r.i.p uh uh he died what like

00:07:54.460 two days ago or something when we were recording this.

00:07:57.660 RIP to someone who was wrong about basically everything.

00:08:00.740 But yes, anyway, continue.

00:08:02.220 Yeah.

00:08:03.480 So interesting place to work, making good money,

00:08:06.700 learning a lot, mathematically fascinating,

00:08:09.600 like really, truly interesting, like on a daily basis.

00:08:14.020 Also, I was like, I'm trying to front run like CalPERS,

00:08:18.520 which is like the, you know, retirement pension fund for California teachers.

00:08:25.400 And I was just like, I just don't feel good about this.

00:08:27.940 I don't, I feel like a junkyard dog, like a scavenger, like eating, like tearing the flesh out of old teachers, you know, retired teachers.

00:08:39.860 So I would wake up every morning.

00:08:41.760 When you said front run?

00:08:43.760 Mm-hmm.

00:08:44.460 What does that mean?

00:08:45.400 that just means like get in the market and do whatever they're about to do before they

00:08:50.460 did it because like if you can if you can anticipate a very very large funds

00:08:56.900 you know purchases and sales then you can make money right um and they they i mean by the way

00:09:05.980 i don't think that that my model would have necessarily success because succeeded but the

00:09:12.460 model i was working on was basically intended to do that right um anyway the point is that it just

00:09:19.420 the cognitive dissonance got to me i was like my stomach hurt every morning and i at some point i

00:09:25.320 was like i don't care how like on paper how good this job is i have to get at right now yeah and i

00:09:31.020 got out and then i started a blog basically to remind people oh yeah i forgot to mention i'm a

00:09:36.700 blogger i started math babe like as soon as i quit yep finance and i was like i gotta warn other

00:09:43.340 mathematicians because they're still sending these emails to mathematicians being like hey are you

00:09:47.280 naive come work for us and i started math babe and then i joined occupy once it started yeah and

00:09:55.760 and so that's how it worked out i don't know if that answers your question no that that makes

00:10:01.200 that makes sense and i feel like also brings us back to okay we're still in this place where

00:10:08.940 people think that amassing large amounts of data and feeding them through algorithms which you know

00:10:14.500 were called you know it went from big data to data science to machine learning to ai but it's all

00:10:21.740 still like statistics on on a mac which i i don't know if you encountered this joke back then but

00:10:28.520 the joke i remember from back then is that data science is is just stats on a mac i remember i

00:10:33.360 actually coined a line that i've heard repeated through other people which was like you know it's

00:10:38.560 data science if it doesn't fit onto an excel spreadsheet like otherwise it's just statistics

00:10:44.480 you know wow yeah i i think it's you know listen i i can both i can have two two things can be true

00:10:50.520 at the same time adam like it it can be true that we're going through the same political confusion

00:10:56.880 like marketing hype 0.97

00:10:59.180 bullshit propaganda system 0.96

00:11:01.720 which is true 0.98

00:11:02.760 and that the technology is impressive

00:11:05.500 and different

00:11:07.660 and like yeah like data science

00:11:09.700 was different from what then

00:11:11.320 became machine

00:11:13.620 learning and that's in turn different

00:11:15.460 from what now is large language modeling

00:11:17.280 it's really different in particular

00:11:18.880 it's much much worse now

00:11:20.700 it's different because it's worse

00:11:23.160 it's much much much worse

00:11:25.600 it's like much worse for the environment yeah it's much more of a threat to our way of life

00:11:31.860 yeah if we're you know normal people so i you know i just i just think yeah like it it's not

00:11:39.360 we shouldn't poo-poo it as as like oh this is just statistics on a mac it's not it's gotten

00:11:44.840 way bigger than that no yeah that's fair yeah um no i was just being glib as i want to be

00:11:52.820 sometimes but um but yeah it has gotten worse and it feels like it's eaten the entire economy

00:11:59.940 um it's interesting though because like just thinking about what's happened over the last

00:12:05.280 10 years thinking about the the sort of career path that you were just talking about i remember

00:12:10.780 i moved out to the bay when i finished my phd so what 2012 and that was right around the shift

00:12:19.940 from when people were saying big data to data science, I think, right around then. And I had

00:12:26.640 the profile of a data scientist in that I had a PhD in a quantitative field and knew how to code

00:12:36.200 in Python and had some experience working with reasonably large data sets and running code

00:12:43.880 remotely on reasonably powerful computers. And then instead of becoming a data scientist,

00:12:49.800 I became a journalist because I had this premonition that if I became a data scientist, I would

00:12:57.460 wake up feeling nauseated every day, which I guess is what happened to you.

00:13:03.460 But I thought that that was pretty bad already.

00:13:08.480 And I didn't anticipate how much worse it was going to get or how much of the U.S.

00:13:16.320 and global economy it was going to eat.

00:13:18.460 And continues to threaten. I don't know how you feel about it, but I'm like whipsawed on a daily basis between like, oh my God, there won't be mathematicians anymore. And I'll say more about that if you want.

00:13:34.720 But two, this is just so bullshit. 0.98

00:13:38.640 And the companies are going to stop using AI 0.99

00:13:42.720 because no one's asking the right question.

00:13:46.480 Instead of asking, what they're asking is,

00:13:49.520 does this help?

00:13:51.580 But they're not asking, how much does this cost?

00:13:53.700 And is it reasonable to use it instead of humans?

00:13:59.540 So they're replacing their workforces,

00:14:01.780 at least on paper, with AI agents or what have you.

00:14:07.860 But they haven't actually started paying the costs of the AI

00:14:12.480 because the AI companies are all just competing for customers so hard

00:14:16.260 that they're not charging for their costs.

00:14:19.500 So you're just like, well, this isn't cost-efficient yet

00:14:22.640 until you consider costs.

00:14:24.280 All you can say is that it's useful in some way.

00:14:28.000 That's not the same thing as being cost-efficient.

00:14:30.100 so like that

00:14:31.300 consideration

00:14:32.340 which comes to my mind

00:14:33.500 once every other day

00:14:34.620 just makes me feel like

00:14:36.180 yeah maybe this is

00:14:36.800 just gonna go away

00:14:37.640 I mean

00:14:38.900 I think that it's a bubble

00:14:40.040 and it's going to pop

00:14:41.200 but

00:14:41.980 when and how

00:14:44.660 matters

00:14:45.380 right

00:14:46.020 I mean

00:14:48.260 this is

00:14:48.940 currently

00:14:49.940 what

00:14:50.820 I mean

00:14:53.140 I don't remember

00:14:54.300 the figures off hand

00:14:55.560 but you know

00:14:56.600 some massive

00:14:57.380 fraction

00:14:58.040 of the stock market

00:15:00.100 is based on the valuation of of this small set of companies and the amount of money being poured

00:15:07.380 into ai and you know you know both in terms of investments and startups and in infrastructure

00:15:14.580 like data centers is just you know going up and up and up and if that bubble pops

00:15:23.060 i mean if it pops now it would be bad if it pops any you know any significant amount of time

00:15:29.460 further away from now it's going to be catastrophic um and so it's better for it to pop

00:15:35.380 sooner good it is i mean i want it to pop sooner yeah i mean listen i've been i think again both

00:15:42.280 things can be true it can pop it can be a bubble and it can have staying power sure like the

00:15:47.000 railroads right like i'm sure you've heard that guy talk about the railroad bubble the difference

00:15:52.520 being of course that the railroad bubble gave us railroad tracks yeah and those were actually

00:15:57.660 useful and it's not completely clear what the utility of empty data centers like half

00:16:04.460 constructions data centers will be in the future if we don't actually use them i mean a fire sale

00:16:09.980 on gpus right it's it's a lot it's a big bubble i think it's already a big bubble i think my guess

00:16:17.820 is that you know when the open ai and anthropic put out their actual numbers in order to get their 0.95

00:16:25.440 ipos which they're trying to do this year yeah you're gonna see a lot of like oh shit like 0.51

00:16:32.320 happening i think and particularly for open ai i feel like it's just losing now huh and i think 0.54

00:16:39.920 like all the companies like microsoft and all the companies that have huge deals with open ai

00:16:45.080 are going to be like fuck we chose the wrong team well in that well but are they then going to think

00:16:53.120 we chose the wrong team in that like we chose the this direction of development like ai as as 0.97

00:17:01.280 something that we were all in on or are they just going to think that they chose the wrong ai company

00:17:05.540 i think they're going to say they chose the wrong ai company and then the ai company that wins is

00:17:10.520 probably google because they just have access to all the data because they force it down my 0.99

00:17:16.480 fucking throat every time i fucking open my android oh god it's so true it's so gross but then 0.98

00:17:22.540 then we're going to actually, once Google wins, we're going to be like, okay, what is this good 0.99

00:17:28.500 for? How much does it cost? It's just going to become a normal technology. I'm not saying that

00:17:34.760 it's not going to slowly eat up lots of things. So that's the longer term thing. I guess my point

00:17:41.620 is that my shorter term thing is like, this is a fucking huge bubble. My medium term thing is like, 0.97

00:17:46.160 it's still going to take over in ways that we are really going to regret.

00:17:53.740 Adam, what do you think?

00:17:55.060 Well, I mean, I think,

00:17:57.900 I was thinking about what you were saying about no more mathematicians,

00:18:01.780 and I mean, I want to know what you mean by that,

00:18:03.420 but I also think I know what you mean by that.

00:18:05.260 There was an article written, oh, I think earlier this month,

00:18:10.840 which is June when we're recording this,

00:18:12.640 um article for i think it was science by a friend of mine josh sokol who um i'll have on the podcast

00:18:20.560 at some point um about the way that they're using ai in um the field where i did my phd

00:18:29.280 um astrophysics and cosmology and um josh is a very good writer um and a very good journalist

00:18:41.280 and it's a lovely piece, but I also found it to be an incredibly depressing piece.

00:18:46.260 And I actually texted that to Josh and he hasn't written back, Josh, if you're listening

00:18:50.060 to this, text me back.

00:18:51.540 But, um, um, I, I think he hasn't written back cause I, I believe he's closing in on

00:18:57.840 the end of working on a book and that is always a really, uh, nightmarish time for keeping

00:19:02.600 up with anything else in your life.

00:19:03.820 But in any event, it's really, you know, I, it feels like it was a long time ago when I did my PhD, but in absolute terms, it's really not that long ago. It was what, like 14 years ago. It's not that long.

00:19:20.740 um i recognize so much of what josh wrote about in that article about like the actual process of

00:19:30.320 doing that kind of scientific research and i saw what he was saying about you know the way that

00:19:36.880 people were doing astrophysics research and without getting into the details that the thing

00:19:42.060 that i was worried about and the thing that some of the people he was interviewing were worried

00:19:45.940 about was, okay, putting aside concerns about how this technology was created, which I think

00:19:55.340 those concerns are reasonable, right? Like concerns about copyright, concerns about the massive amounts

00:20:02.840 of resources that have to be used, like natural resources that have to be used, the amount of

00:20:07.300 carbon that has to be put into the atmosphere in order to build up and train these models and

00:20:12.260 collect the data in the first place, the sorts of awful biases that are in those data sets and the

00:20:18.900 horrible psychological harms inflicted on the people doing the reinforcement learning with

00:20:23.980 human feedback, mostly in developing countries in Africa. Putting all of that aside, which I don't

00:20:32.320 think we should put aside in the long run or even the short or medium run, putting that aside,

00:20:38.740 I can understand where there are places in research where these tools can be useful if

00:20:46.340 you are using them very carefully and deliberately. The problem is in order to use them that way,

00:20:53.000 you have to know how to do what they're doing without them. And so I worry more about that.

00:20:58.620 I'm not so sure about if I believe you. Okay. I think that you're talking about worst,

00:21:03.260 like best case scenario, best case scenario, like, oh, thanks for doing that more efficiently

00:21:08.020 than I would have done it.

00:21:09.220 I knew how to do it,

00:21:10.260 but best case scenario,

00:21:12.120 like, oh, thanks for doing that

00:21:13.300 more efficiently

00:21:13.880 than I would have done it.

00:21:15.080 I knew how to do it,

00:21:16.100 but cozy description

00:21:17.500 of how scientists might use AI

00:21:20.180 is not the way I think

00:21:21.800 we will see things happen

00:21:23.820 and play out.

00:21:24.960 Wait, okay.

00:21:25.640 Now I'm going to ask you

00:21:26.600 to say more about that.

00:21:27.580 What do you mean?

00:21:28.320 Well, I also interviewed

00:21:30.200 somebody this month.

00:21:31.880 I interviewed Daniel Litt,

00:21:33.680 who is one of the people

00:21:34.920 that wrote up

00:21:36.140 the mathematical piece,

00:21:38.900 the counterexample to the Erdős conjecture.

00:21:41.320 So that was in Scientific American,

00:21:42.860 a write-up about that.

00:21:43.680 I looked at the original paper,

00:21:45.140 which is not that hard to read.

00:21:46.940 Daniel Litt is a number theorist,

00:21:49.220 an algebraic geometer

00:21:49.960 that I used to work at the same place with him

00:21:53.820 at Columbia.

00:21:55.960 And that guy who's brilliant

00:21:58.020 and has become really interested in mathematics

00:22:02.080 done with the aid of AI and stuff,

00:22:05.120 he made some predictions that really got under my skin he like really got under my skin i i don't

00:22:12.840 think i've ever gotten so riled up on a podcast and i'll just say briefly and people should listen

00:22:19.060 to it because he's he's i want to be fair to him but i wait hold on time out we haven't even

00:22:24.780 mentioned that you have a podcast yeah yeah we have i have a podcast it's called ai skeptics

00:22:30.020 And I originally invited you to be a co-host of my AI Skeptics podcast, but you decided that was too negative. AI Skeptics is too negative. And a name. And I just think you're wrong about that. But anyway, so I have a different co-host, my friend and colleague, Jake Appel.

00:22:48.860 I am honored that you asked me to do it. And I still probably think that I made a mistake by saying no, but I think we're both having fun doing what we do.

00:22:57.160 We are having fun. I definitely am. It's like my favorite moment of the week. And you were one of my early guests, and I'm on your podcast. So it's nice to do it this way. Anyway, Daniel was on, and his prediction for the future of mathematics was like, we're not going to do this stuff anymore. We're going to ask the computer to do it.

00:23:15.960 And our job is basically going to be like reading the tea leaves of the proofs that it creates.

00:23:21.820 Because that's what he did.

00:23:22.780 Like he, well, Google, like, sorry, OpenAI used a particular version of ChatTBT, which they claim wasn't fine-tuned, but I don't believe them.

00:23:32.020 And they tried a bunch of different Erdős open problems.

00:23:36.820 There's, I think, 600 of them, 100 times each.

00:23:41.380 and like you could multiply the cost of this.

00:23:46.480 It was millions and millions of dollars to get this result.

00:23:49.180 Probably totally worth it to them

00:23:50.660 because it's a huge PR win for them.

00:23:54.760 And they're also just throwing money.

00:23:57.140 They're just burning money with matches anyway.

00:23:59.760 So might as well have something that makes them look smart.

00:24:02.760 But that's just my personal feeling about it.

00:24:05.200 But his point was, it's only going to get better.

00:24:07.180 And in a hundred years, the mathematics will be done by asking the computer, is this true?

00:24:14.060 And then if yes, how do I understand that it's proof?

00:24:18.220 And I was just like, I'm outraged.

00:24:20.380 Like I, you know, we went into math because it's art, because it's beautiful.

00:24:24.340 We're art producers.

00:24:26.100 This is an artistic endeavor.

00:24:28.020 I didn't go into math to become a prompt engineer.

00:24:31.360 And he's like, oh, well, I did.

00:24:33.500 I'm happy with it.

00:24:34.980 He's like totally happy with it.

00:24:36.380 And I was like, whoa, okay.

00:24:38.640 I know.

00:24:39.300 Hold on.

00:24:40.100 Well, actually what he said was, to be fair to him, he said, I went into math because I want to know what's true.

00:24:44.820 And I was like, well, I don't want to just know what's true.

00:24:46.780 I want to know why it's true.

00:24:48.000 Like I want, I feel like handing over proofs to a machine is like seeding the aha moment of discovery.

00:24:57.740 Unfortunately, we still live in capitalism.

00:24:59.760 And that means we need to talk about our partners and sponsors.

00:25:02.880 But fortunately, our partners and sponsors are great.

00:25:05.240 And if you want to support Dreaming Against the Machine directly, you can join us on Patreon, where $5 a month gives you the chance to ask questions of upcoming guests and a dedicated stream of cat pictures.

00:25:18.680 We're a proud member of the Multitude Podcast Collective.

00:25:21.960 Multitude is our home, a group of podcasters creating shows you can count on.

00:25:26.320 They help me and my producer, Nick, with virtually all aspects of getting this show out to you each week.

00:25:31.380 And this month, you can help them.

00:25:33.660 Multitude wants to hear from you.

00:25:35.240 They're running a survey of listeners to better know the audience for all of the shows in the

00:25:39.020 collective, including this one. This information will help us keep this podcast afloat, both

00:25:44.140 financially and emotionally. If you've ever wanted to send us a note about what this podcast means

00:25:49.000 to you, this is a great way to do it. The survey is at multitudeshows.com slash survey, and it's

00:25:54.820 open until August 17th. It's real short. Go check it out. And here's a word from our sponsors.

00:26:00.200 just to back up for a minute for our listeners an erdos problem is one of the many open problems

00:26:09.580 that was left behind by one of the great 20th century mathematicians paul erdash and if you

00:26:14.320 want to hear stories about erdash like listen to that podcast because i was in hungary and i met

00:26:19.940 him oh no shit and he also likes stayed with my parents so my mom had stories about him and i i

00:26:25.540 tell those stories on my podcast. Oh my God. Did you publish with him? My dad did. My dad 0.99

00:26:31.420 had an Erdős number of one and I have an Erdős number of three. So that's close enough to be

00:26:37.580 honest. I also have an Erdős number of three, which is considered low for physicists. And I

00:26:42.680 take more pride in it than I should. Yeah. One of the things that my mom told me, but it didn't get

00:26:49.380 on the podcast. I'll say it here is that she tried to do math with him and he fell asleep. He was

00:26:54.240 much more interested in men oh man he thought of women as like helpful like he handed her his

00:27:00.480 loose button and asked her to fix it that was how he saw women he was just a really messed up guy so

00:27:06.760 i don't i don't idolize him at all but he did leave behind a lot of open questions that's true

00:27:13.100 yeah which is a good thing in in math um not not a bad thing i should also explain an erdash number

00:27:21.840 is like, um, math and, and I guess physics equivalent of like the Kevin Bacon game where

00:27:29.280 like, if you published with Erdos, cause you had many, many coauthors, you have an Erdos number

00:27:33.180 of one. If you publish with someone who published with Erdos, you have an Erdos number of two

00:27:36.880 and so on and so forth. Um, both of your parents were mathematicians, Kathy.

00:27:42.000 I did not know that. Yeah. Wow. Neither of my parents are physicists,

00:27:47.080 but they're both listening. It's actually pretty normal. If you look at women mathematicians that

00:27:50.680 their father was a mathematician. 0.93

00:27:52.920 Really?

00:27:53.400 It's like much, much higher than you'd expect.

00:27:55.800 Huh.

00:27:56.760 I guess that makes some sense.

00:27:58.020 But I'm sorry. 1.00

00:28:00.920 I got completely derailed by your family knowing Erdash. 1.00

00:28:05.160 Yeah. 1.00

00:28:05.440 So, but going back to the point, which is like mathematicians becoming prompt engineers.

00:28:10.220 I'm, it's dark.

00:28:11.740 I mean, it's dark.

00:28:13.300 And one of the reasons I got particularly riled up is because I had just gotten back when

00:28:18.200 i recorded this with daniel like gotten back from a memorial of the guy who founded my math camp

00:28:24.000 as a teenager i went to a math camp and this is a guy his name was david kelly it was at hampshire 0.62

00:28:30.160 college he's a total hippie feminist gay guy and he's the kind of guy who would like take a stick

00:28:36.920 and teach us math in the dirt with a stick like he was like a you know kind of a cult leader type 0.96

00:28:44.480 Pythagorean, imagine

00:28:47.060 a modern-day Pythagorean cult.

00:28:49.560 Cult leader, positive. 0.70

00:28:52.360 In a positive

00:28:53.340 way, yeah. It was actually cultish.

00:28:55.480 He obsessed over

00:28:57.060 yellow pigs and the number 17.

00:28:59.280 Those were our mascots.

00:29:01.920 And it was very, very

00:29:03.000 sweet and dear.

00:29:05.240 It was where I found my identity

00:29:06.760 as a young person. That's why I called my blog

00:29:09.180 Math Babe, because of my

00:29:10.520 discovery of who I was

00:29:13.120 once I got to that math camp.

00:29:15.440 And before I start crying, let me just say that, like, just the idea of telling him, like, here's what math is going to look like in a hundred years.

00:29:24.520 Like, I'm kind of just like, I'm glad he's not here to see this.

00:29:28.460 Well, first of all, in the same way that this is a podcast where we can swear, this is also very much a podcast where we can cry.

00:29:35.600 That's good.

00:29:36.300 Yeah, I am, you know.

00:29:37.820 Maybe I really do need that glass of wine.

00:29:39.800 Yeah, maybe you do need that glass of wine.

00:29:44.480 oh man um i yeah i mean honestly you're gonna make me cry because now i'm thinking about the

00:29:52.400 nerd counts where i sort of figured out who i was and also thinking about the mentors who meant

00:29:59.400 a lot to me and also i don't i feel obliged to say this even though it might make you cry

00:30:05.340 more but also like hampshire college is going away yeah it was very fitting to be there we

00:30:11.760 were there when everybody at him like all the staff at hampshire college got fired the same

00:30:15.540 day as the memorial it was just brutal it's a brutal ending yeah in so many ways i've never

00:30:22.520 been to hampshire college but i've always known about it and sort of admired it from afar and was

00:30:27.160 happy that it existed it's exactly what you think it's like this bastion of hippie love like there's

00:30:34.220 other ways of thinking there's other ways of prioritizing truth and knowledge and beauty

00:30:40.400 and love and it was really special i mean i guess that's one of the things that kind of like going

00:30:47.720 back to ai and its long-term influence which i you know we did we talked short-term and medium-term

00:30:53.920 but like long-term it upsets me that it's like so homogenizing of thought yeah you know i feel

00:31:02.340 like we're all just gonna we're being very much encouraged just to offload our critical thinking

00:31:06.880 to ai if we even have critical thinking anymore but like offloading our way of thinking and it is

00:31:14.420 scary how how many people around me right now and i consider myself like a little pocket of

00:31:22.540 resistance but even the people around me at being like well claude told me blah blah blah and i just

00:31:27.500 want to punch them in the face every time like stop asking yeah no same the place to me where

00:31:33.820 it's the most visible, and maybe this is just because this is what I do, but I see it in writing. 0.98

00:31:38.500 Right? I mean, I really believe in the old saw that writing is thinking. And I think that when

00:31:46.480 you offload your writing to ChatGPT or, you know, Claude or whatever, you are offloading your

00:31:53.640 thinking. And this gets me to the stuff that I was, you know, sort of in the process of saying

00:32:00.200 that I was worried about regarding astrophysics and cosmology, which is, you know, I'm worried

00:32:04.040 about de-skilling people learning how to do these things only with AI and not learning how to do

00:32:08.480 them for themselves, which is terrifying, but less terrifying than what you were just talking

00:32:14.140 about regarding math. And definitely also something that people are saying about

00:32:18.540 astrophysics and cosmology. And that makes me want to scream and cry. But seeing the way that

00:32:29.500 people's thinking is getting, you know, smoothed out and homogenized. Like you were saying, like,

00:32:35.780 like replacing your own voice with the smeared out averaged voice of the internet.

00:32:44.600 I work hard at my writing. I have worked hard at my writing for a long time since before I

00:32:50.520 became a professional writer. It was an important creative outlet for me. And then like a,

00:32:54.300 Like a fool and many people before me, I let my creative outlet become my job. 0.95

00:33:00.500 So then I had to find a new creative outlet, which these days is photography. 0.89

00:33:03.560 But one of the things that I work hard at and that I think a lot of writers work hard at is finding ways to make your writing express your voice more authentically.

00:33:16.480 Like, I want my writing to sound like me, and I want to make sure that I'm there on the page with the reader, so that when the reader reads what I have to say, it's a conversation between the two of us, even though it's one way.

00:33:34.780 Okay, let me stop you there, Adam.

00:33:36.080 Okay.

00:33:36.680 Not because I don't admire that, which I do.

00:33:39.680 I would even agree with it, but I would just say that most people don't take writing that seriously.

00:33:46.020 Yeah.

00:33:46.880 I would even venture to guess that you don't always take writing that seriously.

00:33:50.780 Like there's some writing that you have to do, not because it's a creative outlet,

00:33:55.540 not because you're having a great thought that you want to get down just so,

00:33:59.620 but because you're obligated to respond to emails.

00:34:03.220 Sure.

00:34:03.880 And I feel like that's the slippery slope.

00:34:06.920 Right.

00:34:07.180 I can imagine like a school of, a future school of writing where you're like,

00:34:13.060 you're going to do this on pieces of paper with pencils and this is the way we do it and it's

00:34:18.280 great and people are like yeah i'm in but they're still going to use ai to respond to emails and

00:34:24.200 you know what i mean i feel like that's that's how they get you but it is a slippery slope i mean

00:34:28.760 that's why i don't do that well one of the reasons i don't do that the other is the other stuff i was

00:34:32.920 talking about before that i was putting to the side for the moment is you know i don't want to

00:34:36.500 use the ai to do it because i think that there are harms that that you inflict by using it and

00:34:41.120 creating this stuff. But I mean, yes, you're right. A lot of people don't take writing that

00:34:46.040 seriously. But first of all, you don't have to take it that seriously to see that writing is

00:34:50.940 thinking. And second, I have seen professional writers start to use AI to do stuff. And it makes

00:34:57.420 their writing worse because it makes them sound like everyone else. And I don't think that AI

00:35:02.420 is a particularly good ventriloquist or will become one. Like when people ask it to, you know,

00:35:07.960 write in the style of so-and-so,

00:35:10.940 I don't actually think it does a great job of that.

00:35:14.320 Usually, it's only doing a particularly good job of that

00:35:18.000 when you are asking it to do something

00:35:21.340 in the style of someone

00:35:24.180 from a particularly different time and place

00:35:26.580 so that the word frequencies are really, really different.

00:35:30.200 Or they're like super stylized.

00:35:31.840 Yeah, exactly. Super stylized in some way.

00:35:34.760 Yeah, yeah, yeah, exactly.

00:35:36.520 And yeah, okay, maybe it'll get a little better at that.

00:35:40.040 But I don't want that future.

00:35:45.360 And I feel like it is similar to the sort of future of mathematicians and cosmologists

00:35:50.160 as prompt engineers in a way.

00:35:52.620 I'm not convinced that that's where we're heading.

00:35:56.860 And I also think that we have a choice about that.

00:35:59.820 I mean, because first of all, I don't think that large language models and, you know, the sorts of AI that could be like immediate descendants of large language models are conscious and like having experiences in the world in any meaningful way.

00:36:21.260 And I do think that that's an important part of doing all of this stuff.

00:36:25.500 Adam, I think I might have asked you this when we were talking before, but one of my favorite arguments along these lines is like, if you read a really good review of a restaurant written by AI, would you think that it's a good restaurant?

00:36:39.720 Yeah, exactly.

00:36:41.180 You know, just to put that point on it, like the fine point on it, it's just like, it cannot eat.

00:36:46.660 Yeah.

00:36:46.980 It cannot enjoy itself. How about like a review of a, like a sexy movie? Like this was super sexy.

00:36:53.460 Or a perfume.

00:36:55.140 Or a hotel stay.

00:36:56.440 Yeah, yeah, yeah.

00:36:57.040 A national park.

00:36:59.980 There's so many things.

00:37:01.580 They're embodied experiences.

00:37:03.820 And just as they can't enjoy a spa treatment,

00:37:08.640 we also can't punch it in the face for giving us the wrong advice,

00:37:12.200 which is another thing I keep coming back to.

00:37:14.400 There's nobody to punch in the face.

00:37:16.360 Yeah.

00:37:16.980 I mean, this is a better version of the old,

00:37:20.060 you know, a machine cannot be held accountable.

00:37:21.820 therefore a machine cannot be you know make a management decision or whatever like we can't

00:37:26.060 punch a machine in the face so yeah well this brings this brings me to to something else we've

00:37:32.560 been talking about things that we're depressed about but this is dreaming against the machine

00:37:35.780 and yes it's still early days for the podcast and i'm still figuring out exactly what it is that

00:37:40.960 we're doing here but i want to figure out what it is what it is that you could possibly be

00:37:45.820 optimistic about as well well it's not so much optimistic about uh as it is like how do we fight

00:37:51.900 this oh okay what do we do about it well this might seem like a passive form of resistance but

00:37:59.040 i i'll i'll throw something out there like a friend of mine who just got his phd in computer

00:38:04.280 science so he knows a little bit of from what he speaks it um is obsessed over the the danger that

00:38:11.520 is, in his opinion, currently posed by people creating agents that are running amok on the

00:38:19.340 internet. And they're doing it for all sorts of reasons with all sorts of incentives. A lot of

00:38:26.900 them are just trying to make huge amounts of money on taking advantage of dupes on sports

00:38:33.220 betting sites. And I'm sure they're doing a good job with that. But he's afraid that some of them

00:38:38.000 are going to really fuck things up.

00:38:40.360 And I'm like, what's the worst case scenario?

00:38:42.220 He's like, nuclear codes, everyone's going to die.

00:38:44.780 And I'm like, okay, but that's not realistic. 0.99

00:38:47.300 Like, probably something much stupider 0.94

00:38:49.200 is going to happen before that. 0.96

00:38:51.360 It's just going to, like, jam up the electricity grid.

00:38:54.920 It's going to, like, be a virus and make itself,

00:38:57.700 you know, whatever it's going to do. 1.00

00:38:59.100 It's going to do something really stupid. 1.00

00:39:00.960 Yeah. 1.00

00:39:01.660 And I'm all for it.

00:39:03.940 I'm just like, bring it on.

00:39:05.320 Faster the better.

00:39:06.260 Like, let's bring down the electricity grid in the Northeast and be like, this was AI agents run amok. 1.00

00:39:12.780 And everybody would be like, fuck AI agents. 1.00

00:39:15.100 Fuck that. 1.00

00:39:16.460 And we're going to finally see some, you know, response from policymakers. 1.00

00:39:22.120 Because like, there's my experience with like harm, which is what I do, like AI harm or algorithmic harm, is that nobody cares until someone's dead.

00:39:33.000 Wow.

00:39:34.280 And then even then, how many people have to die?

00:39:38.980 And I'm just being realistic here, Adam.

00:39:40.700 Like, we didn't get seatbelts until a lot of people died and Ralph Nader stepped in.

00:39:46.300 And he was hardcore.

00:39:47.500 I want to be the Ralph Nader of this situation.

00:39:50.360 But I'm just saying, like, you also need to wait till the evidence is in your face.

00:39:56.740 In people's faces.

00:39:58.480 This is dangerous.

00:40:00.040 And I don't think it is there yet.

00:40:01.720 Okay.

00:40:02.240 Okay, I'm going to push back on that a little bit.

00:40:04.960 First, I'm going to make a joke, which is you want to be the Ralph Nader for this,

00:40:07.680 but you don't want to spoil the 2000 election.

00:40:12.760 Spoil the 2000 election?

00:40:14.700 Ralph Nader.

00:40:15.540 Okay, yeah.

00:40:16.380 Sorry.

00:40:16.840 Oh, I see what you mean.

00:40:17.860 Because he also did that.

00:40:19.040 Yes, because he also did that.

00:40:20.680 Ralph Nader causing Al Gore not to be president.

00:40:23.300 No, don't do that, Ralph Nader.

00:40:25.500 Why'd you do that?

00:40:26.740 But yeah.

00:40:27.900 I'm not saying he was an unalloyed good, but I'm saying that.

00:40:30.560 No, I know that.

00:40:31.280 definitely did some good stuff. Absolutely. I'm just saying it was a combination of him doing it

00:40:36.900 at the right time in the right way with really brutal tactics. And the fact that the evidence

00:40:42.940 was there, the evidence was palpable. And I just feel like we are still in the, where, you know,

00:40:49.240 the, the phase of like, this looks like it could be really bad for jobs, for this, for that, for

00:40:54.780 this. And I'm, and when people are like, oh, this could happen. I'm just like, just bring it on.

00:40:59.520 Like, AI agents screwing everything up.

00:41:01.720 I can't wait.

00:41:03.080 But the things people are already dying.

00:41:05.280 I know.

00:41:05.740 I know.

00:41:06.280 And I'm involved in the lawsuits, Adam.

00:41:07.920 I'm involved in the lawsuits.

00:41:08.660 I know.

00:41:09.020 I know.

00:41:09.400 And also, I should have said at the beginning, full disclosure, I am an affiliate of your

00:41:14.080 consulting firm.

00:41:15.540 Yeah.

00:41:15.980 I'm ready on multiple fronts to be the rail feeder of a given situation, whether it's

00:41:20.160 like chatbot, kid suicide, or the electricity grid going down and people not being able

00:41:25.800 to find their way home on ways and like riding into the Charles River, like whatever it is.

00:41:31.500 No, I'm, I'm with you. I'm just surprised that people don't care more about the people who have

00:41:38.800 already died. Like someone on the internet somewhere, I think maybe Blue Sky made the

00:41:45.500 analogy to the, um, the Tylenol murders back in what the late seventies, early eighties,

00:41:52.900 where somebody um yeah yeah yeah it was early 80s because i was old enough to read the paper

00:41:59.340 yeah so it was uh you know somebody i don't think they ever caught who it was put poison into

00:42:05.020 unopened bottles of tylenol just in in a drugstore somewhere and uh people bought them and a few

00:42:13.420 people died including like at least one kid and immediately there was i mean even before legislation

00:42:20.480 passed, these companies voluntarily started including seals on over-the-counter medication,

00:42:27.560 which we still see to this day and which I believe are now mandated by law. But I'm fairly sure they

00:42:32.540 started including those even before they were mandated by law in response to this.

00:42:37.400 Yeah, certainly Tylenol did.

00:42:38.820 Yeah, exactly. Because otherwise, Tylenol would have stopped being sold.

00:42:43.700 You know what? I think the difference is, I'm going to tell you one of the differences,

00:42:48.140 And I know this because I'm involved in some of these lawsuits.

00:42:52.780 Tylenol, it was like a bad actor causing the deaths of children

00:42:56.600 who were just trying to take Tylenol.

00:42:58.560 Here, it's like kids kill themselves sometimes.

00:43:01.520 The argument from these chatbot companies is like,

00:43:04.060 oh, kids kill themselves.

00:43:05.160 Troubled kids find their way onto the internet. 0.99

00:43:07.340 They would have killed themselves anyway.

00:43:08.860 And they happen to be in discussion with this chatbot,

00:43:11.680 but that's not why they kill themselves.

00:43:13.440 And just kids, think about it, kids kill themselves.

00:43:15.340 And so there's no clear causality yet.

00:43:19.940 And one of the big reasons there's no causality, as we know, is because the big companies will not let us into their data.

00:43:26.960 So that's one of the things I can't wait to get my hands on is the evidence.

00:43:34.680 If we could do that, that'd be amazing.

00:43:37.100 I'm not saying we can do that, but if we could get access to the data in discovery and just be like,

00:43:42.540 Actually, no, there was causation here.

00:43:44.980 Here's evidence that the weirder and more predatory and more groomery this chatbot agent was, the worse off the kid was.

00:43:56.680 And there was a feedback loop, and it got worse and worse, and then the kid killed themselves.

00:44:01.540 That's the kind of thing that a jury would buy.

00:44:05.260 But we don't have that yet.

00:44:07.040 So we don't have the bulletproof case that it was caused by.

00:44:11.900 as like the Tylenol poisoner, you know?

00:44:15.720 Right, right, right.

00:44:16.840 Same goes for the seatbelts in cars.

00:44:19.740 These people, if they hadn't been driving in a fast car,

00:44:23.980 would be alive.

00:44:25.600 There was a huge amount of effort to blame them

00:44:28.320 for crazy driving, but it just happened too often

00:44:30.460 and too statistically consistently to really blame drivers.

00:44:35.180 So I'm just saying, I'm now a student of like,

00:44:39.100 how do things get regulated?

00:44:41.580 And if your listeners have stories to tell me, bring it on.

00:44:45.600 But that's what I'm interested in because I'm like, let's do this.

00:44:48.900 Let's get rid of J.D. Vance because he is owned by Peter Thiel and he's never going to let anything through.

00:44:57.300 Let's get a new administration.

00:44:59.440 Let's bring it on with regulation, but also let's make a good regulation.

00:45:04.840 That's another problem.

00:45:05.880 And then let's pack the court so we can actually have that regulation stand up and not get thrown out by an unelected, partisan, unaccountable supermajority.

00:45:20.260 Oh my god.

00:45:20.840 Yeah, I know.

00:45:21.980 No, but I think we've got to do those things.

00:45:24.680 Which means that I guess what we're saying is that one of the things we've got to do is we've got to vote our way out of this.

00:45:30.540 i one of the things that i've been doing when i've been going around and giving talks about

00:45:36.560 this stuff is i i you know more everything forever my book at the end of the book i close it

00:45:41.680 by calling to tax billionaires out of existence and when i've been giving talks about it one of

00:45:47.040 the things that i've been saying is that taxation is the compromise and and that billionaires should

00:45:51.860 be not just cool with taxation but actually agitating to be taxed which as far as i know

00:45:57.500 the only billionaires doing that are like jb pritzker and warren buffett um but more of them

00:46:04.480 should be doing that because the alternative historically has been guillotines and when i

00:46:09.460 say that people cheer and i worry a little bit that they're cheering for like the mention of

00:46:16.460 guillotines when what i i feel like is no no guillotines are bad um like i don't i i think

00:46:24.980 you know hot take murder is bad and we should be dealing with this through redistribution of wealth

00:46:34.180 like aggressive redistribution of wealth and regulation of markets to make them you know

00:46:39.460 small enough to drown in a bathtub he said i'm with you because i like you was raised to think

00:46:46.400 to have faith in systems of government yeah but if you talk to 20 somethings i know they were raised 0.63

00:46:53.840 in a nihilistic manner

00:46:55.580 and they were like,

00:46:57.560 yeah, maybe that guy

00:46:58.320 who killed the CEO

00:46:59.420 of UnitedHealthcare

00:47:00.260 had a point.

00:47:01.220 Yeah.

00:47:02.060 Like, that's pretty close

00:47:02.820 to the guillotine.

00:47:03.820 But if you ask, 0.84

00:47:04.720 if you ask by age cohort

00:47:07.200 whether that was,

00:47:08.900 that guy should get 1.00

00:47:09.820 the death penalty 1.00

00:47:10.720 or, you know,

00:47:11.980 you'll be surprised.

00:47:12.440 Well, I don't think

00:47:13.080 he should get the death penalty.

00:47:14.380 I think the death penalty 1.00

00:47:15.240 is barbaric.

00:47:16.460 Or even if he should be

00:47:17.500 convicted of a crime.

00:47:19.280 Well, okay.

00:47:21.240 I mean,

00:47:21.840 that was of me

00:47:22.820 by a 24-year-old?

00:47:24.240 Should he be convicted of a crime?

00:47:25.620 And I was like, absolutely.

00:47:26.880 What are you talking about?

00:47:28.400 He killed somebody in broad daylight.

00:47:29.900 And they're like, yeah, but wasn't it reasonable?

00:47:32.860 And I'm like, no, it wasn't. 1.00

00:47:35.320 What the fuck? 0.98

00:47:37.080 But I definitely was like, 1.00

00:47:39.060 I'm talking to a generation right now.

00:47:40.760 I definitely feel that way.

00:47:42.640 And not everyone is actually that far,

00:47:44.820 but I mean, I'm just saying,

00:47:47.720 we were raised in a different time.

00:47:49.980 And can you imagine growing up right now?

00:47:52.820 I know. I get it.

00:47:54.780 Like, they literally, these 20-somethings, they have spent their entire conscious life with Donald Trump.

00:48:03.520 Yeah.

00:48:05.300 No, I know.

00:48:06.360 So, they see the kind of reality, like the mafia boss reality that's going on.

00:48:13.320 Like, it's not out of the question to get rid of the people that you don't like.

00:48:17.280 I mean, look, I think that billionaires should be a bit, quite a bit more afraid.

00:48:26.220 Well, I'm going to take that back, actually.

00:48:28.560 I think that the billionaires, or many of the worst of the billionaires, like Musk and

00:48:33.900 Thiel and Drees and so on, actually are quite terrified.

00:48:38.360 And if you look at what they say, it's-

00:48:40.100 They have huge security details.

00:48:41.500 Yeah, they have huge security details.

00:48:42.980 uh their their public statements are just dripping with paranoia and persecution complexes

00:48:47.540 so it's not that they should be more afraid but i think that

00:48:53.860 i'm going to tell you a story about these people yeah and i'm not going to name any names but a

00:49:00.080 friend of mine had a friend so it's somewhat third hand yeah but he was asked to go to a

00:49:08.020 undisclosed location in the middle of like utah and talk to a bunch of billionaires and their

00:49:12.740 consultants on how to do what i know who this is this is this is in the public record this is

00:49:19.120 douglas rushkov okay i don't know if this is the same story because i actually don't know the name

00:49:23.780 okay finish the story but i think it's the same story okay tell me if it's the same story but

00:49:28.240 they're basically like we want to know how do we when like the apocalypse comes and we're living

00:49:33.880 on our little island how do we like keep the guards from killing us yep and the the advice

00:49:40.600 was like make friends with the guards

00:49:42.820 and they're like no that's not it

00:49:44.420 we're thinking shock collars

00:49:45.680 is this the same story? 1.00

00:49:47.680 it's the same fucking story yeah 0.99

00:49:49.120 so Douglas Rushkoff wrote about this experience 0.99

00:49:52.540 that he had he was flown out to an undisclosed

00:49:54.780 location and some billionaires

00:49:56.460 asked him that question and his response was

00:49:58.480 you have a lot of power and

00:50:00.340 influence and money right now you should use that

00:50:02.480 to prevent the collapse of civilization

00:50:04.580 and they laughed that off

00:50:06.660 wow

00:50:07.760 yeah I know

00:50:08.860 No, I mean, and this is, I think, also a large chunk of why they're so, so enamored of AI.

00:50:17.480 They see it as a potential solution to this problem.

00:50:20.160 You don't need the guards if you can just do it all with robots.

00:50:25.580 Now, I think that violence is bad.

00:50:30.280 Sorry, I'm realizing soundbites from this episode include murder is bad, Adam Becker,

00:50:34.660 and let's take down the Northeastern power grid, Kathy O'Neill.

00:50:38.860 Um, but, um, I'm not suggesting we actually do that.

00:50:43.620 No, I, I understand.

00:50:44.980 I understand.

00:50:45.620 And by the way, my friend who's worried about this is like, I'm going to put on out an army

00:50:49.020 of agents that do good instead of evil.

00:50:51.720 And I'm like, don't do that.

00:50:53.040 You're like, you're walking into a literally like a, like Star Trek plot.

00:50:58.600 Yeah.

00:50:58.960 You are the one who causes the thing you're worried about.

00:51:01.440 Yeah.

00:51:01.640 Yeah.

00:51:01.940 Yeah.

00:51:02.160 Exactly.

00:51:02.900 Yes.

00:51:03.820 That's exactly what it is.

00:51:05.300 Oh my God.

00:51:06.840 So many Star Trek plots.

00:51:08.040 are happening. There are so many Star Trek plots

00:51:10.020 happening, and also I keep thinking

00:51:12.080 about, like, early 90s

00:51:14.220 computer movies, like Hackers

00:51:16.100 and Sneakers. Yeah, totally.

00:51:18.420 Sneakers in particular has, like, a

00:51:20.120 very special place in my heart. 0.99

00:51:22.140 We are fighting

00:51:23.560 a class war. Yeah.

00:51:26.420 And people

00:51:27.940 are dying, and so I can understand

00:51:30.020 where people might feel that violence is

00:51:32.020 justified.

00:51:34.400 I just don't believe

00:51:36.260 that we're there yet. Maybe I'm wrong.

00:51:38.040 Listen, I'm a big fan of Momdani, and I love the fact that, you know, and also, by the way, going back to 20-somethings, 20-somethings are fans of socialism.

00:51:50.460 And the idea that the democratic establishment is like, oh, that's fringe and far too left and we can't possibly.

00:51:56.980 What they're asking for is like affordable college and affordable health care.

00:52:01.080 It's not crazy.

00:52:03.260 No.

00:52:03.480 It is absolutely reasonable requests from their government.

00:52:08.040 They're asking for stuff that our parents had.

00:52:10.160 Yeah, exactly.

00:52:10.940 My parents went to Brooklyn College and it basically cost nothing.

00:52:14.040 Yeah.

00:52:14.860 But yeah, Kathy, this has been a delight.

00:52:18.160 Thank you for coming on the show and we will have you back sometime.

00:52:22.200 Great.

00:52:22.760 I'll have you back on my AI Skeptics podcast.

00:52:25.620 Sounds good.

00:52:27.440 Thanks again to this week's guest, Kathy O'Neill.

00:52:29.700 Next week, I talk with Alondra Nelson, who is the former head of the White House Office of Science and Technology Policy under Joe Biden.

00:52:40.140 And we talked about how to make scientific research work better and move on from the horrible damage that is being done by the current administration.

00:52:52.220 So tune in. It's a good one.

00:52:54.780 To submit questions for future guests and to suggest other guests,

00:52:59.620 and to see more pictures of Babka, join the conversation on Patreon.

00:53:04.480 You can also find us on YouTube, on Instagram at DATMPod,

00:53:08.800 on the web, and on Blue Sky at DreamingAgainstTheMachine.com,

00:53:12.940 or just find us wherever you get your podcasts.

00:53:16.280 Dreaming Against the Machine is a proud member of Multitude Productions.

00:53:20.300 Our executive producer is Nick Karisimi.

00:53:22.580 our theme music is by Jared Emerson Johnson

00:53:25.620 our show logo is by Nick James

00:53:28.380 and our fearless leader is Babka

00:53:30.740 the greatest cat in the observable universe

00:53:33.040 I'm Adam Becker and I'll see you next week

Dreaming Against the Machine - August 04, 2026

Episode 17： Weapons of Math Destruction, with Cathy O’Neil

Episode Stats

Transcript