Wharton AI Researcher: If AI Froze Today, It Would Still Rewrite Work

Forecasters gave AI under 2% odds of helping solve a millennium-level math problem by September 2026. It happened. Ethan Mollick has been measuring what AI actually does to real work with field experiments, not forecasts, and his answer is not what most people expect. Subscribe if you want science with evidence, not speculation. Mollick runs the Generative AI Lab at Wharton. He is the author of Co-Intelligence and Coexistence, his new book on navigating a world where AI tools are already good enough to change almost all intellectual work, even if progress stopped today. The conversation covers the jagged frontier: why AI solved a math olympiad before it could count the Rs in strawberry, what it means that roles blur when everyone on a team starts using AI, and what four human advantages still hold an edge when a model can already write a publishable paper. It also covers what happens when your kids tell you that you sound like Claude. What you’ll hear: What centaur, cyborg, and conjurer work actually look like in a real academic lab Why brainstorming was considered a bad idea shortly after it was invented and what that means for working with AI Why Mollick reduced em dashes in his new book and what that tells you about living alongside these models What co-evolution might look like if the escalator keeps moving “If we just had the models we have right now, that’s enough to change almost all of human work in the next ten years. It’s good enough.” — Ethan Mollick CHAPTERS 00:00 AI writes good papers now. Taste is the weak spot. 03:14 The jagged frontier: math olympiad gold, but not the Rs in strawberry 06:30 Deep knowledge, wide knowledge, taste, and agency 09:44 Escalator or eschaton? 16:02 Does AI simulate wisdom or actually have it? 23:46 The labs keep AI in English. Here is why. 32:28 Roles blur. Everyone does everything now. 36:42 My kids say I sound like Claude. They might be right. 38:12 AI found errors in his sources that his fact-checkers missed 46:34 What to do before giving a child access to AI Get the transcript, fascinating bonus content, and my Monday M. A. G. I. C. Message: https://briankeating.com/yt Enter the debate: https://briankeating.com/ai Have a .edu email and live in the USA? You automatically win a meteorite: https://BrianKeating.com/edu Subscribe: https://www.youtube.com/DrBrianKeating?sub_confirmation=1 Support Into the Impossible on Patreon, get my weekly M. A. G. I. C. Message, unfiltered bonus content, and live monthly Office Hours with me: https://www.patreon.com/drbriankeating Join this channel for perks, monthly Office Hours, and your name in the Member Roster at the end of every episode: https://www.youtube.com/channel/UCmXH_moPhfkqCk6S3b9RWuw/join Featured Guest: Ethan Mollick on Twitter/X: https://x.com/emollick?lang=en Coexistence (book and site): https://co-existence.ai/ One Useful Thing (Substack): https://www.oneusefulthing.org/ My books: Losing the Nobel Prize (memoir): http://amzn.to/2sa5UpA Think Like a Nobel Prize Winner: https://urlgeni.us/amzn/TLANPW Focus Like a Nobel Prize Winner: https://a.co/d/hi50U9U Galileo’s Dialogue (first-ever audiobook): https://a.co/d/iZPi9Un Twitter/X: https://x.com/BrianKeating Substack: https://briankeating.substack.com Blog: https://briankeating.com/blog Audio-only: https://briankeating.com/podcast #EthanMollick #AI #artificialintelligence #AItools #briankeating #intotheimpossible #AIagents Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  The wheel was brilliant, but the road made it iconic. Electricity was powerful, but the grid brought it to every home. Innovation needs infrastructure. AI is no different, especially in finance, where AI can't run without context. But it can run on Plaid's infrastructure. Built over 13 years to understand why money moves. Plaid makes AI work in finance. Visit plaid.com/ai to learn more.

• 0:30Speaker 01

  Stonefire Naan is made with real buttermilk and ghee. It's the perfect flatbread for any occasion.

• 0:35Speaker 02

  I put out Stonefire Naan with dips at every dinner party.

• 0:38Speaker 03

  I make Naan grilled cheese when my daughter has friends over. I use Stonefire Naan to reward the humans who do my bidding.

• 0:45Speaker 02

  Right, it's perfect for camping trips.

• 0:48Speaker 03

  Dance for it, feeble mortal.

• 0:50Speaker 02

  Sorry, is he talking to me or?

• 0:52Speaker 03

  Dance.

• 0:55Speaker 01

  From our tradition to yours, Stonefire.

• 0:58Speaker 02

  Together we Naan.

• 1:00Speaker 04

  On the academic research side, this stuff can write a good paper at this point. I think people are too eager to assign bright lines to AI. Like, it can never do creativity or judgment. It does creativity and judgment. So the AIs are trained to use English because it lets us have some observability into what they're thinking. It's not a lucky thing. It is an actual design decision. If we just had Fable 5.1 right now in Ashram, that's enough to change almost all of human work in the next 10 years. It's good enough.

• 1:24Brian Keating?

  This is Ethan Mollick. He runs a generative AI lab at Wharton, and he measures what AI does to real work with field experiments, not just theoretical forecasts. He's going to name the four human skills he thinks still hold an edge. And one of them is the thing that we scientists claim to have and can almost never define. We start with mathematics, where the experts forecast missed by a megaparse. Getting towards the mathematical angle of things, you know, just in the recent days, a couple of days ago, as we're recording now in mid-October, there was an announcement of 700 papers and different levels of theorems and proofs and axiom development released by OpenAI.

• 2:04Brian Keating?

  Three weeks earlier, there was an announcement that it had solved the millennium problem, which is a huge deal. And the pace of change is just astonishing. I have my arguments against it being as useful in physics and the hard physical sciences as it is in math, because I do think math is much closer to a language. And what these things are are large language models, you know, coupled to GPUs. But I have my doubts about whether or not they can truly do physics in the way that, you know, your colleague at Penn, my friend Mark Devlin does it, or I do it. So my question is, you know, right now, solving Olympiad-level math problems and millennium problems is happening, but also simple mistakes are happening. And I wonder, you know, if the kind of jagged frontier that I want you to define

• 2:49Brian Keating?

  is going to be, you know, kind of a permanent feature, the jaggedness? Or is it going to become a phase transition where we do trust it? It doesn't think there's 10 Rs in strawberry. And I mean, it doesn't do that anymore typically. What is the underlying problem? Is the uniqueness or the unevenness of things that we see now, is that temporary? Or is it a permanent feature of how these machine intelligences work?

• 3:11Speaker 04

  Really great question. So the jagged frontier's idea we came up with in our first academic paper, where we were studying the effects of AI on people in a consulting business. And what we identified was the idea that AI is good at some things you won't expect, bad at some things you won't expect, and you can't necessarily tell in advance what it's good or bad at in a particular way. And that means your expertise matters, right? Like, knowing what this is. Now, math is really interesting because it was the most jagged of parts of the intelligence, right? When my colleague Phil Tetlock, who runs a famous forecaster, the Forecasting Institute, predicted the chance of the AI would get a gold medal at the International Math Olympiad right before ChatGPT came out, the odds were something, you know, were under 5% or something like that. And of course, the AI got a gold medal, but for a year, it was getting the Rs in strawberry wrong.

• 3:54Speaker 04

  There was a bunch of technical breakthroughs that around how AI reasoning works, but also it just turns out the models were better than we thought they were. And then last year, the same forecasting group, including a group of industry experts, were asked to predict last September what the chances that the AI would help solve a millennial problem by September 2026, a year later. I think the answer was 1.7% that that would happen, right? So like, as an indicator of how quickly the jagged frontier changes, right? Leave aside all the other effects on math, that's a really good indicator. I think AI remains jagged. And part of what you got at with your question was one of the reasons why. So AI works better when we have definite things to train on, right? So math coding where there are right answer works great. And by the way, that's why writing lags so much.

• 4:34Speaker 04

  Because good, like Claude writes the same kind of way because there isn't an objective measure of writing to judge against.

• 4:41Brian Keating?

  Well, that's a load-bearing argument that we can.

• 4:44Speaker 04

  Thank you, thank you. Yeah, it's really doing the work here, yes. If it was a year ago, we would have said, that's a rich tapestry of answers. On the other hand, I'm an economic sociologist, I train, I work at a business school, right? I put out those quantitative papers, large-scale data sets. I am getting the latest models to be able to go and say, find an interesting dataset, download it, do all the tests you need, pre-register the result, and I'm getting good papers. I'm not getting, the research taste is the weak spot right now, not the execution. There's no errors, the sites are great, everything is terrific, but it's taste for research, which is kind of an interesting question anyways. Like, should we, as humans, we need to have taste for research because there's only so much work we could do. But if the AI could do infinite research, do we care that it's doing boring but measurably real work? I don't know. We have to reconfigure science around that. So I guess what I'd say is, I think that jaggedness will remain,

• 5:28Speaker 04

  but as capabilities increase, the jaggedness may not matter that much. If you reach superhuman level in something, does it matter whether or not that's weaker than the superhuman level in a different field? At the same time, that jaggedness creates a lot of opportunities for humans to work with AI in important ways.

• 5:42Brian Keating?

  In the book, there's a lot of different characters and maybe archetypes that get portrayed. And it's different than your previous book in that you're expanding upon things that are familiar, the metaphors of the centaur and the cyborg. But now you introduce this concept of a conjurer, you know, some human delegator to AI for completing agentic work. Talk about those three different, you know, creatures, a centaur, the cyborg, and the conjurer.

• 6:07Speaker 04

  So the centaur and cyborg, I talked about in the original book, right? And so those have both been used in different ways. We define, use them in some of my research papers with people as well. So centaur work, think about half-person, half-horse with a division line. So centaur work was, you know, you do the writing and I do the coding, right, with the AI? Or vice versa? And you have a clearly divided line. What most people move to after a little while is what I would call cyborg work, where you blend the work together, right? And this is very common, you know, you ask the AI a question, you go back to your own stuff. Well, you give me a few versions of this, it becomes kind of blended work. That all made sense in the kind of co-intelligence era when you work with chatbots. The biggest change that's happened is AI can now do fairly arbitrarily long lists of work on its own, without human intervention, with a good chance of success. And that means you can assign it very complex work for it to do.

• 6:52Speaker 04

  And I've had to do all kinds of insane things, right? Talking about writing research papers and dating myself as a nerd side. I gave Astro the game Zork, which is one of the original text adventure games, and said, turn this into a graphical adventure game. It built a full graphical adventure game with, like, amazing 3D spaces you move through, but it had to also rebuild every aspect of the adventure to make that work, which is clever stuff, right? So as we can assign work to AIs, that becomes what I called conjurer work. Now, I did make a mistake here. Cyborg and centaur start with C, but they were soft Cs, so it's cyborg and centaur. When I was writing conjurer, like, oh, this will sound great, but it actually looks good on the page, but it's actually terrible because it's centaur work.

• 7:30Brian Keating?

  I want to know, you know, what are the human advantages that you think are most sort of hopeful or perhaps useful for us in this age? And then I want to drill down into scientists in particular. What are the human scientific and mathematician advantages?

• 7:46Speaker 04

  So I talk in the book about four things, and I want to make it clear, these are not bright lines. Like, I think people are too eager to assign bright lines to AI. Like, it can never do creativity or judgment. It does creativity and judgment. May not be your creativity and judgment, may not be as good as you, but we shouldn't define bright lines as one thing or another. So what does that mean? That means that we should think about ways that we gain long-term advantage, you know, in working with AI. So I call those things deep knowledge, wide knowledge, taste and agency. So deep knowledge is expertise. You have deep knowledge, obviously, in astrophysics, right? And in your fields of interest, probably in podcasting at this point, in a variety of other fields where you can instantly judge, right? Whether or not this podcast is going well or not, or whether it's like, oh, God, this is going to be miserable. And you probably knew it the second I got on, right? Because expertise is all about heuristic and decision-making.

• 8:30Speaker 04

  Just like you can look at a paper and be like, is this BS or not, really quickly, right? So we need expertise because it lets us work with these systems in better ways. And we talk about agentic work. One of the ways you want to be able to work with an agent is just like you would with a team of grad students. You like, know right away if they're in the right direction or wrong direction. And that requires deep expertise. Wide knowledge also is really increasingly important. So wide knowledge is that you know many things, right? You've already made references to the Tower of Babel and to, you know, to philosophy. The more things you know, this is a model trained in all of human knowledge, basically. And that is both fun for using these systems, right? You can explore new topics, but also gives you sort of superpowers. Like, coders are doing design work, designers are doing coding. If you know a little about coding or design, the systems work much better for you. Then there's the idea of taste, and it's something we should talk more about.

• 9:15Speaker 04

  But AI is really good at producing many of a thing. How do you know you should select? How do you pick something that's not just slop? The enemy of slop is taste, right? And then the final thing is agency, which I'm an entrepreneurship professor. So it's something we struggle with a lot. How do you give people the sense of agency of control over their own lives? In the way that, that, you know, in entrepreneurship, but also in AI, how do you not wait for permission? How do you experiment with these things? And I think those four aspects together give you an edge over other people when you're using AI systems.

• 9:40Brian Keating?

  Escatan is a term I first heard from Peter Thiel, talking about the Antichrist and the Escatan. And I have to say, I didn't fully understand the point that he was making. But let's, let's remind the listeners or tell me, the host, what is the Escatan? And what is the escalator? And how are these things in contrast in terms of disruption to this, this possible future of constantly recursively improving superintelligence?

• 10:03Speaker 04

  In my original book, I had four scenarios for the future, right? Maybe AI progress stops, maybe it slows down substantially. Maybe we stay on some sort of exponential curve, or maybe we reach this sort of AGI, ASI, a machine smarter than a human at every intellectual task, right? It'll define sort of future. I, I think scenarios one and two are out. I think we are either on a continual exponential that doesn't reach the level of superintelligence that I think a lot of people in Silicon Valley are obsessed over. Or we, or we do, right? I call those the escalator and the Escatan. So the Escatan is literally the last thing in Greek. I use it as sort of substitute for the singularity, right? The idea of, in the original sense, that we can't predict past this point. There is a point where if you imagine AI gets good enough, we have a superintelligence out there, is it alien to us? It will do what it does, and we don't, can't predict what the world looks like in that, in that case, right?

• 10:48Speaker 04

  And there are people who want to usher in the end of all things. I think the most likely scenario we go through, right? But who knows for sure, is the elevator, a steady increase in abilities of AI. Over time, the speed is going to matter a lot, right? And we're going to muddle through one way or another, hopefully. But there'll be a lot of challenges, a lot of wonders and horrors happening along that way. And I think that's more likely than the end of days, or at least more planable. Like, if the end of days is coming, you know, what are we going to do? You know, all we can do is be nice to the AI and hope that there's a place for us in the, in the world to come. But I think we spend a little less time worrying about how do we adjust well to a situation where this is happening? Calling back to the mathematicians, that's a canary in the coal mine. Other professions are going to face the same kind of problem, right? Maybe we can argue about whether physicists will have the same kind of crisis or not, but I think it's going to come for a lot of us.

• 11:31Speaker 04

  And it's going to come for professions, and we can reconfigure work in different kinds of ways. People have the model of the industrial revolution, they talk about a lot, and like, it worked out great in the end. And hopefully this all works out great, too. I think we can guide it, so it works out great. But I also think that the industrial revolution also sucked to live through. Like, that's what Dickens was about, that's what Marx was about. And I think that we have to think about cushioning the escalator ride so that everyone wins, and it's not just a bunch of AI bros winning this thing.

• 11:56Brian Keating?

  And you have stuff in here from 2026, but, you know, like most books, you probably started writing in a couple of years ago. And one of the things you talk about in the book is that you ask your AI to buy you a gift, a surprise to you for something under a hundred bucks. You give it a little bit of a guardrail, and it comes back and it, they use your Amazon account, I think I remember. Yeah, when it does that, I mean, how did you react to it? I mean, you describe it, and to me, it's a little creepy, it's kind of nice. How did you react to that?

• 12:21Speaker 04

  I am continually slightly creeped out and really excited and anxious and really optimistic all at the same time. A week does not go by where I'm not shocked by something that AI has done that's interesting. I also know that it fails a bunch of times, right? So like, there's all these persistent problems. I start my, my very first book, I started with the idea that you can't use AI without three sleepless nights of anxiety. And I think that, that it kind of just continues. I think I've underestimated something, it was like, actually, there's been a lot of moments like that, you know? And when Astrak came, when Fable came out, like, there's, there's a bunch of moments. Interestingly, right, I think that since I wrote the book, I kind of had jury-rig AI agents together even through early 2026. And now there's Muse and Dot and a bunch of these other systems that I think people will be much more used to using AI agents to do stuff.

• 13:06Speaker 04

  And they're pretty good at what they do. There is kind of an interesting question, what you said, which is also, they haven't messed up that much yet. Like, for a billion people using these systems, I think we were all worried two years ago, would there be mass psychosis? Would the AI be causing all of these issues for everybody? And the answer turned out to be kind of no, like, that doesn't mean bad stuff doesn't happen, like, it actually has happened.

• 13:24Brian Keating?

  You're right. The fact that you hear about it is, means it's noteworthy, which means it doesn't happen that much.

• 13:28Speaker 04

  Most people are using these systems and not like, and there's lots of people making mistakes and people are being hurt. And I have no doubt there's people psychologically being harmed by this. The data is really unclear about how often that happens, but there's definitely people who are being driven to, you know, suicide or psychosis by these things, I'm sure. But it's also the kind of thing where it's like, huh, we've muddled through to some extent. And I think we have to think about how do you cushion the muddling through so people don't get hurt? But it's also worth noting, like, you know, there is some experimentation here. You're right about agentic hygiene, we should all be much more careful. Now, the past is not the future. And again, this sort of singularity awaits us of, I can't predict what happens next. I do think cyber attacks are going to become a very big issue in the very near future. I think we're reaching points where models are much more capable. So the future may not look like the past there, but it hasn't been the disaster yet.

• 14:11Speaker 00

  The wheel was brilliant, but the road made it iconic. Electricity was powerful, but the grid brought it to every home. Innovation needs infrastructure. AI is no different, especially in finance, where AI can't run without context. But it can run on Plaid's infrastructure. Built over 13 years to understand why money moves. Plaid makes AI work in finance. Visit plaid.com/ai to learn more.

• 14:41Brian Keating?

  And your first book was co-intelligence, which I found very helpful. And, you know, to be honest with you, you know, and with love and respect, you know, coexistence sounds like, you know, we're trying to hang on. It was almost a little bit disturbing or depressing before I read it, as I judge the book by cover, as everybody does. It's kind of the thumbnail and the CTR rate that we have to monitor. But I think, you know, co-intelligence was sort of important in a variety of ways. Most, most of which had to do with the fact that, you know, Homo sapien, the word sapien means, you know, one who is wise, that's our name. And what is it wise about? It's, it's wise that it's, you know, according to most tradition that I understand, is that we understand uniquely among all animals or creatures that we're going to die.

• 15:24Brian Keating?

  And that makes me think that, you know, intelligence is almost free, you know, there's so much intelligence, but what we lack is wisdom. We lack the wisdom that to discern between what is, what is useful, what is knowledge, what is, what is to be limited. And so I kind of wish that your next book would be, you know, co-wisdom, but, but it's coexistence. And as I say, you know, with, with respect, it kind of is like, we have to justify ourselves or like, how are these two-state solutions going to coexist? Was that intentional? And what do you make of this knowledge versus wisdom? I mean, it seems to me wisdom is the shorter in supply, and, and we need to sort of maybe lean into that.

• 15:58Speaker 04

  I do want a slight note of panic in this. I mean, I think that the subtitle was going to be before, it's the next phase of AI, how to work with AIs that are sometimes but not always smarter than you. And I think that that, that does require a form of us coexisting. And by the way, that also includes protecting parts of ourselves from AI use, right? Like, it involves like choosing not to use these systems when you shouldn't use them, so you don't de-skill yourself. It involves, you know, in learning situations, you know, knowing when to use AI or not, what we want to maintain humanity of. I think that we have to make choices about this, and I think that this is a force in the world we coexist with. So I didn't mean it so much the AI was forcing us into a world where we had to live in the margins, but more one that we have to negotiate our relationship with this technology now.

• 16:38Speaker 04

  A chatbot sort of sat in the chatbot, you know, in the box, and you'd talk to it, and when you closed the box, it went away. I've got agents right now looking at my emails, and they're popping in, like, I've got a message already from one of them about like something I probably should be paying attention to. They've saved me lots of money, they made my life easier, but I also now have an entity that is actively playing a role in my life, right? So I think that there is that piece. On the wisdom side, I feel like a controversial piece I will say to you is that I still worry about bright lines there. And I don't know if wisdom, there's human wisdom, we could say we value human wisdom, but wisdom writ large? You know, and we only have, we only us have a, you know, visceral sense of mortality, right? But like, the AI simulates wisdom well enough and simulates theory of mind well enough that the question is like, is it wise or not?

• 17:22Speaker 04

  And this is where, you know, I try to be pragmatist about this, or, you know, materialist, I guess, which is like, let's judge it by what it does, rather than try and say that it's the same characteristic as a human, right? I don't find it that useful to ask questions about, is it really wise? It simulates wise enough that I don't know, you know, for practical purposes, I mean, I teach at a business school, right? Like, this is, you know, there's a practical aspect to this kind of thing of like, it does what it does. Let's measure it and see if it does these things, let's see if it causes harm, and then we should debate it afterwards. So on the wisdom side, I think we all need more wisdom. I think the AI gives very wise advice. We could think there were things we know about, like finance or medicine, where you absolutely should be asking a frontier model as a second opinion. Not as your first opinion, but as your second opinion, we have enough evidence that it's quite good at those things.

• 18:06Brian Keating?

  Except for that meme that, you know, shows a guy asking ChatGPT, is this mushroom safe to eat? And this is, yes, and then the next scene is a gravestone, and it said, I'm sorry, I should have considered the answer more carefully.

• 18:18Speaker 04

  But absolutely, but also it was worse and more dangerous. That, like, the thing to note is, everyone is worried about long-term safety. Short-term safety has gotten better. The, the AIs are much less likely to guide you into a psychosis than they used to. I mean, it sounds ridiculous, but like, they were releasing models that would like, you would underline, you know, tell you, yes, you're special, you should hide out from humanity. They don't do that as much anymore, right? They used to be bad at health advice, they're better at that. So like, I'm not saying, you know, there's a human wisdom involved too of like, yeah, don't eat the possibly poisonous mushroom off the advice of one AI, right? But on the other hand, if all you have to eat are two mushrooms, I'd rather ask the AI for advice than not to, right? And so I think it's, that's the wisdom we need in some ways.

• 18:57Brian Keating?

  So the founder of your august institution was, of course, Ben Franklin, and he was an experimentalist. He was a physicist of sorts, an amateur physicist, among his many things. You know, in physics, we don't call a theory successful because it Produces things that sound good or resonate with us and look at string theory and other things. They sound good, but we demand a prediction that can survive an experimental conflict with the experiment. Is our inability to confront mother nature? So my question for all AI researchers and specialists, and you're an expert in this, is like, what is the equivalent experiment? What is a decisive experiment demonstrating that electrostatics and lightning and conduction are one unified phenomenon as Franklin did?

• 19:40Brian Keating?

  Or any other empirically verifiable, but most importantly falsifiable, potentially falsifiable equivalent? What is a prediction? What's some crisp thing, a decisive experiment that you could demonstrate that AI actually understands?

• 19:53Speaker 04

  The problem is a definitional one rather than experimental design, right? I mean, I have and built experiments. Does AI affect learning? Does AI affect productivity? Does AI affect creativity? You know, and even then there's definitions. What is productivity? And how do we define creativity? And how do we measure it? Again, empirical researcher here, and I don't know how you define what thinking is. Like, we can look inside the brain of the AI, right? And we know what it's doing. Is that thinking or not? I couldn't tell you. We get a really good simulation of thinking and we can test that. That's where my pragmatism sort of kicks in. It's like, I want people to grapple with this problem. In fact, one of the biggest worries I have about academia is the complex reactions to AI. And I think math was more enthusiastic than most and kind of got burned by it. I mean, not burned, this is where things are going.

• 20:37Speaker 04

  But there's a lot of resistance among fellow academics about this technology. For a wide variety of reasons, good, bad, real and not. To me, I really want to hear people who are philosophers of consciousness dig into this. And biologists have got, like, I am not an expert in those areas. I can tell you about the things I know a lot about and I'm broadly interested in many topics. But this is the problem for humanity is to understand what this tool is. How do we work with it in the right way? How do we use it? Where is it useful as a tool for advancing human work? Where should we never use it? What does this teach us about intelligence and language and meanings of the world? I mean, it's the most exciting opening. It didn't do everything else it was doing. The fact that we have a simulated artificial brain should be the most exciting thing that academia has dealt with in a decade or a century.

• 21:18Brian Keating?

  And I wonder, at what level is everything reducible to language? But I want to give you what I call the Keating test, very modestly named, but it's actually the Einstein test. And it goes back to 1907 when Einstein said he had his happiest thought. It was a thought he claimed that titillated him. And he was participating in what's called a thought experiment, which is something an AI can do, which was that if you were in an elevator and the cable broke and you were in free fall, you experienced no gravitational field. And that's now known as the Einstein equivalence principle. So I always like to ask, in what sense can an AI, an artificial brain, have that visceral sensation that we all know? If we've been on a roller coaster in the backseat of my car as I'm taking my kids to school, going over zero Gs?

• 22:01Brian Keating?

  What sense can an artificial intelligence have a sensation viscerally? A, and then B, can it have a happy thought? Those are things that we can have and a child can have it, but can an AI have it? Are these things imprisoned by not having bodies or not having happiness? Or have those dimensions of their personalities?

• 22:21Speaker 04

  There's two questions there. One is about subjective experience of AIs. And the other is, to the extent that AIs are language models trapped in a box, does that hurt their ability to do things in the world or understand the world? So on the first one, anthropic is convinced that there is a consciousness inside these systems, right? And the way they measure that is they can see that vectors within the sort of the math that makes up these AIs that light up in response to things that are pleasurable, the same sort of way across all sorts of pleasurable things. So the AI simulates it. Like, I have a suspicion, I have to test at some point, that if you give the AI a task it likes to get better results. And not, the math alone would be enough to indicate that. You don't need it to have a subjective experience. So we sort of have this weird place, which is, I have no idea.

• 23:05Speaker 04

  There are things that you could, if you squint at them, pretend they work like the brain does. We don't know how the brain works and this set of stuff. But it's also really easy to delude yourself and believe, like, the AIs trained to act like us. It is given a role and that role is a human role. We train the language model to respond the way a human would and a helpful, happy human. So the fact that it responds that way, are we just being an illusion? I don't really know. I'm skeptical that current models are conscious in any way. And I think that to the extent that they do have a subjective experience, it's an alien subjective experience. And we're arguing over definitions at some point again. Like, is the fact that it has a happiness thing light up is the same thing as a dog waving its tail? Same thing as a human being happy? I couldn't tell you the answer to that question. On the wider issue of, what are they missing?

• 23:50Speaker 04

  There's a lot of money going to that, right? Part of this is robots are a much bigger deal than LLMs. And increasingly, it looks like LLMs might be able to drive robots without having to go through all the steps we thought we'd have to do. There's this push to like, what if we put AI training in worlds rather than training the real environment learned things? So I think there's a lot of controversy over all of this. What I would say at the end about this is large language models tend to be unreasonably effective technology. Like, they shouldn't be as good as they are. Like, they shouldn't be able to do vision as well. I mean, there are slight variations at LLM, but regardless, that they shouldn't be able to do vision and solve problems in robotics and do math and write poetry. Like, it's weird they do all these things. We don't even have a basic answer to that about why it's so good, right? And Stephen Wolfram's angle on mathematician was that in creating a model of human language,

• 24:32Speaker 04

  we created a model of the world because human language models the world. Otherwise, how could we describe quantum physics unless we could describe it with language? And I find that a somewhat satisfying answer, but it's still hard to answer the questions about consciousness or anything else.

• 24:46Brian Keating?

  I talked to Stephen many times and I agree that's somewhat useful, but like, not all languages are the same. In fact, this brings up my next question, which you talk about a lot in the book, is scaling and scaling laws. But there's Hawaiian, which has 13 letters. And there's Cyrillic, which has 32 letters. And there's English. Not all languages are the same. There's binary code. And there may be languages that these entities can communicate in that we don't understand. And we are just lucky for now that they use English.

• 25:16Speaker 04

  That's actually a big worry of AI researchers, because if you've ever used, especially Claude Fable 5 has gotten better at this. If you've ever watched it talk to other AI agents, it talks in a language called Claudish, which is a variation of what's called neuralise, which is, yes, of course, the AIs will come with a better language for talking to other AI agents. And a huge amount of effort and fighting inside the labs is that's more efficient, but we lose observability. We can't see what the AI is doing. So the AIs are trained to use English in this kind of case because it lets us have some observability into what they're thinking rather than not English. It's not a lucky thing. It is an actual design decision that the labs are under pressure, the AI labs are under pressure to relax because the AIs will be much more efficient if they didn't use English.

• 25:58Speaker 00

  The wheel was brilliant, but the road made it iconic. Electricity was powerful, but the grid brought it to every home. Innovation needs infrastructure. AI is no different, especially in finance, where AI can't run without context. But it can run on Plaid's infrastructure. Built over 13 years to understand why money moves. Plaid makes AI work in finance. Visit plaid.com/ai to learn more.

• 26:28Brian Keating?

  With the scaling laws that you mentioned, the only pushback I have is that they're just empirical. They're phenomenological. They're not immutable laws of nature like F equals MA. So again, what observation would convince you that scaling is going to hit some fundamental limit? I mean, Moore's law is sort of saturating. There's Jebbins paradox, there's Wright's law, all these things. But at a certain level, like, I don't know about you, Ethan, but if AIs were future lobotomized, we're stuck at Fable 5.5 and Astra. I mean, 99% of what I do, except for the 1% in physics, which is important, but 99% of what most people do is completely solved. I mean, it's Turing complete at some level for ordinary human beings.

• 27:10Brian Keating?

  So what level do you think that scaling is bound to continue or will it saturate?

• 27:14Speaker 04

  It's so great because you always give two questions to each question. So I'll try and pull those apart. Question one is about scaling saturate. Scaling is always saturated multiple times. There's multiple scaling laws. I talk about a couple of them in the book, but the truth is we keep hitting walls with AI and then we figure out pathways forward. There's enough money and enough smart researchers, right? I mean, Moore's law is the same way. My first academic paper was on Moore's law. And I interviewed Gordon Moore at the time and everything. And Moore's law just turns to be a succession of different stuff that everybody had the pacing that we need to keep this pace going. So people would try many technologies and some would fail, some would succeed. Same thing's happening with AI. We thought we hit a limit with the amount of human language in the world and synthetic language would make the AI break down. But it turns out, oh, actually, the AI does a good job with synthetic language because we figured out ways to make that work.

• 27:58Speaker 04

  So if large language models stop working, there are world models. There are other kinds of approaches to thinking about these things that would step in as a result. What you're seeing now is, especially now that the logic of capitalism is at work, there's so much money at stake. Researchers find ways through, right? Like, that's the one thing you find, like, business hits dead ends, you go forward. Imagine so much infinite more funding going in. Our revenue and it's the only thing America cares about is like, we need to keep building, improving our knowledge of physics at that kind of level, people will find a way. And so I think that that's what you're kind of seeing here is like scaling laws are hitting a wall all the time. But the reason why I don't think there's a stop to this stuff is so they'll figure out something else or will gain efficiency. So you can use a smarter model or you use 20 smart models in a swarm.

• 28:41Speaker 04

  Right now, there's so many pathways forward that it's not like there's like, oh no, this cuts off, we're done, right? It's not like photolithography where we're kind of reaching the images of lithography in Moore's law. Like, we don't have a replacement technology, really. There's 50 or 100 different next paths for LLMs and no one's even bothered with them. They're not bothering to increase efficiency because they're just unreasonably effective as they are and they don't need to do it yet. So I think we will hit walls, but I don't think they're walls we can't overcome.

• 29:21Brian Keating?

  We need to judge this book by its cover. For the 10 people out there who don't know your work, I like to have you go through the title, the subtitle and this cover art, which I can kind of guess at. But take us through it. And the one thing you can't say is my publisher chose everything.

• 29:35Speaker 04

  My publisher did not choose everything. My whole family did help me, right? And we've helped mark up some of the cover art just by not being an artist. My sister helped me come up with the title. So it was a family affair. I have to give credit where it's due. First of all, I want to avoid shining white robots, right? Like part of what I think we have a problem with AI is most of our views of AI are cliche. And most of the images we use of AI are either drawn from like utopian science fiction or Terminator, right? And like, that's not what the AI we ended up with. We didn't end up with the shiny white robots while they glowing over things like. So I think that, given that, right? Like, what do we want to invoke? Something kind of human about our own sort of, you know, our relationship to something other, other intelligence. So first one hand, reaching for the tree of knowledge from a classic painting. And this is obviously from, from the Sistine Chapel, right?

• 30:20Speaker 04

  And a close up of the hands. Although there is an Easter egg. I don't know if you spotted the Easter egg on the cover.

• 30:24Brian Keating?

  The fingers, extra digits.

• 30:26Speaker 04

  I actually debated over which hand to give the extra digits to, but it worked better. It's not supposed to be that big a statement. But the idea is like, there is like something going on. We're contacting, it's not white intelligence. I don't get too bogged down in questions of intelligence or consciousness, but there is an intelligence here that we're in contact with. It's an other. And I think there's something interesting about we're creators, but we're also being shaped by it. And I think that's an important thing to keep in mind.

• 30:47Brian Keating?

  Tell me something. There's a thought that, you know, that Roman has, which is that the fact that these things are provably, you know, unprovable, like that we can't prove that they're safe. Therefore, they're going to have these capabilities and cause great danger and therefore they should be regulated, et cetera. When I try to, you know, corner him, I'm like, well, I'm smarter than the spider that lives near my studio. And whenever I go there, the spider has raised his web. Roman says to us, you know, can a squirrel tell what the humans are going to do? Yeah, you see the spider, he told me, and you don't kill the spider, even though you could, but that doesn't mean you wouldn't kill them all if you could. His point is that we tend, we tend to anthropomorphize these things. And that can lead us to underestimate. Because Brian doesn't kill every spider that gets in his hair.

• 31:31Brian Keating?

  That means that AI won't do it to us. Do you think that the bigger problem is, is underestimating via anthropomorphization or overestimating what these things can do?

• 31:41Speaker 04

  So I'll briefly side answer this, which is, if you want to use AI to do stuff, you have to anthropomorphize it because that it's just, it's trained to work like a human. And in fact, there's a great study out of Anthropic that shows that managers get better coding results out of AI than coders do on their first attempt, because they're better at specifying what they want, right? And they treat the AI like a person that does work. I think anthropomorphizing is a risk for everything large, for the reasons we've talked about here, though, which is like, it's not a person. It's trained to act like a person. And so the arguments that are built on chains of speculation about what a person would do are hard to apply to an AI. I mean, I think, again, there's value in realizing that there's a lot of very serious people very worried about the risk that comes from, existential risk that comes from AI. It's hard to pick one argument out of that set that is like,

• 32:26Speaker 04

  yes, this clearly the logical chain makes sense, but they're knowledgeable in this field that we should be spending policy effort to avoid mitigating some of that risk. But I do have trouble, once we get down to the level of like, what about this particular argument from how humans work? And hypothetically, if you're an evil human who would do this, what would you do? I don't know how to make that connection to AI, right? Like, we're sort of in a spot where it's like, we probably should be very careful about building a super intelligence that can do its own thing without human intervention. That seems like a very reasonable first step to take. In terms of what its motivations are and can we align it? You know, I'll go back to what you said before. If we just had Fable 5 right now, 5.1 right now in Ashram, that's enough to change almost all of human work in the next 10 years on the internet. Like, it's good enough. We haven't even begun to start. The AI labs make the products.

• 33:10Speaker 04

  They haven't even begun to be swallowed by companies and organizations and fields and institutions. So, you know, I think there's reason to be cautious about the front, but I just don't have an argument about what's in front of us, about that anthropomorphization makes one way or another, right? Like, you don't need that. You could just say, super intelligent machines have inherent danger in them without having a logical chain about why they might kill us all. The fact that we don't have a good reason why they would or wouldn't seems like a reasonable, you know, thing to say, let's be cautious.

• 33:36Brian Keating?

  Just a quick interruption from today's sponsor, me, Brian Keating. Since it's Nobel Prize season, I can't resist advertising my three books. First book I wrote is called Losing the Nobel Prize. It's a cosmology story mixed with a memoir of what it feels like to do science at the highest level and spoiler alert, fail. But my next book, Think Like a Nobel Prize Winner, tells the story of nine Nobel laureates who will reveal their insights and actions how to take your science or just your job in general to the next level. It's full of insights and tips that I never would have thought of. And lastly, my book published last year, Focus Like a Nobel Prize Winner, is written for students of all ages to learn the tools and the tricks to compete with yourself,

• 34:21Brian Keating?

  which sometimes could be the hardest competitor of all. Get all three of them. Amazon affiliate links in the description below. And if you've read them already, I'd love it if you leave a review. Every single review helps. Now, back to the episode. One thing that didn't quite make it into the book because of timing was a recent paper that you and your colleagues wrote in Organization Science, which I subscribe to. This article is called Cybernetic Teammate. And I love it because it has, where's experiment? I'm like experimental physicist. So anyway, it's called A Field Experiment on Generative AI and Teamwork. The result that, you know, I gleaned from my AI summary of it is that like, basically, AI is worth two people. Like, a good AI in the hands of the right person adds like a, like a couple people. You know, it was focused on, obviously, financial business. Do you think that that's sort of would regress out

• 35:06Brian Keating?

  across all different types of discipline? Like, knowledge work in physics? Could I have my experiment? When Mark and I lead an experiment with, with others that has 450 people on it? Like, should we replace half of them with, you know, a third? Or, you know, replace two thirds of them with AI?

• 35:19Speaker 04

  All right, so another couple of great questions. Actually, I do mention this in the book because the thing that's interesting to me about it is not actually that individuals with AI perform as well as teams without AI. That's obvious. There are people who get 100 times performance gains with systems. It's a little weird about who gets the performance gains or not, because the systems are still sort of single player. Like, I use AI and it does on my own. Use AI on your own. We haven't yet built the systems that are like, what does it mean to, as a team, to work with AI? And that's as much organizational process as it is anything else. And so we're still working on that idea of how do you do multiplayer AI in that kind of way? But the productivity gains are very clear. Like, there's, there's studies and lawyers of doctors. You throw it at anything, programming consultants. The interesting thing about that paper that I actually do cite, because I don't cite the productive impact,

• 36:03Speaker 04

  because again, GPT-4, prior era, before AGENT work. But the thing that I don't, that is really interesting and I mentioned, is the roles blur. So business people on their own do business work. Technical people do technical work. As soon as they use AI, they step on each other's toes. And one thing that's happening in every organization we've talked to is the blurring of lines in parts of the organization. So coders are doing design work, right? Designers are doing coding work. That is a really big change. And by the way, super interesting, my lab, a generative AI lab at Warham that I run, like, the productivity gains have been insane, right? And people do everything now. You have someone who actually can write up an interactive, you know, explainer about, about the paper. And we could do the paper and we can have, you know, the AI check the paper against 15 other values and do the, like, it's a crazy expansion of what we could do.

• 36:48Speaker 04

  But it also makes it hard to manage because I used to know what everyone did in the top and everyone does everything.

• 36:54Brian Keating?

  So I want to turn again to the world's second oldest profession, being a professor now. Walk me through your daily life. You know, you wake up, you, you have your coffee. And where do you go next? How are you using it on a daily basis as a professor?

• 37:07Speaker 04

  First of all, I have to make choices. So the choices, I have lines, right? I will not send anything AI written to anyone, to anyone who I'm actually communicating with. Like, I'm happy to turn down stuff with AI written comments, but like, students, fellow faculties, senior people who write me. And it's a challenge, by the way, because I could write to everybody. I get hundreds of messages. I try and make it so my messages are human. I always write my first drafts and everything without talking to AI at all first, right? To just, all my own writing, my tweets are all my own. I don't have a staff doing it.

• 37:34Brian Keating?

  Are you like whisper flow or are you using dictation?

• 37:36Speaker 04

  Usually I'm typing. I do do messy notes where the AI does have. Now, okay, around all that, I, I wake, you ask when I wake up with a coffee. I've woken up, I've already had two different agents read through everything overnight that I've received. Give me a punch list of things to worry about. We have a secure U Penn account, right? And then I also have stuff for my Gmail account, right? And they're in communication with each other around that set of stuff.

• 37:57Brian Keating?

  Because I don't give it my UCSD email.

• 37:59Speaker 04

  Again, legal rules are often really unclear. So for Penn kinds of email, right? You want to, like, we have a version of ChatGPT Enterprise that is safe for HIPAA use and for student use. I have my emails scanned by the AI and this kind of is telling me what things to pay attention to. Like, if I miss stuff, because I get lots of messages in. So I've become much more responsive since I built these systems. I also have others scanning the research and like, are there papers I should be thinking about looking at or something along those lines that I also get on a regular basis? Along with like, my financials get scanned by the AI and tells me if there's kinds of issues there. And then usually, you know, I'll do work, right? And but like, for example, when I tweet about a technical paper now, I write the tweet on my own, but I will actually paste the paper into the AI and say, am I getting something wrong here? Like, am I missing? And usually it's overly cautious.

• 38:44Speaker 04

  Like, no, no, you're, you're using such sweeping language. I'm like, don't worry about it. Like, that I understand what to do about. But like, is this technically right? I have like, if I'm writing, I get stuck on a paragraph. Give me 40 versions of how to end this paragraph. And I might use taste to kind of pick among them to get me thinking again and writing. I absolutely have built like a whole bunch of systems. Like, you know, I have like, when I'm putting together a set of slides for stuff, I'll throw that into Claude and give me 20 variations of how I could explain this concept. And then usually I use none of them, but pick on something on my own to go forward. On the academic research side, this stuff can write a good paper at this point. Increasingly, at our lab, we are guiding the AI to do the first pass research, pre-register the experiment, do the research work, gather the data, give us the results, and then we read it through, just like I would with a grad student.

• 39:27Speaker 04

  So I think anytime I could use a team, I use the AI to do that work.

• 39:30Brian Keating?

  You get a referee request. You have to prepare slides. You have to give a public talk. You have to talk to donors. At what level are you handing the keys to the agent? And what level are you, you know, I'm just going to do this, don't matter how it takes? Or I'm going to turn down these opportunities because I don't have the bandwidth to do it in the bespoke, human touch fashion?

• 39:48Speaker 00

  The wheel was brilliant, but the road made it iconic. Electricity was powerful, but the grid brought it to every home. Innovation needs infrastructure. AI is no different, especially in finance, where AI can't run without context. But it can run on Plaid's infrastructure. Built over 13 years to understand why money moves. Plaid makes AI work in finance. Visit plaid.com/ai to learn more.

• 40:18Speaker 04

  I feel like it's really important to be spoken human touch. By the way, you see this bite people all the time, right? Like, you know, Pan Graham is a fairly good AI detector for right now. There were no good detectors, you know, a year ago. And everyone who used AI to write all their stuff, like, you know, if I'd used AI to write all my subset, you would know because also you'd recognize Claude speak right now. So like, if I wasn't being authentically me, that would show through.

• 40:39Brian Keating?

  And I'm sorry to interrupt, but you actually make the case in the book that it's actually the way it writes has affected the way you write in that co-intelligence had 154 M dashes and this current book has 54.

• 40:51Speaker 04

  Yes, I mean, that's the coexistence piece. I removed M dashes because it's a signal.

• 40:56Brian Keating?

  I won't, because I wrote with them and I love them. I'm never going to remove them.

• 40:58Speaker 04

  And I remove phrases. The things I say every time my kids say to me, you sound, you know, I think Claude was based on you because we have a similar kind of speech. Like, so like, that's a problem for me. But like, yes, that's part of my reaction is to show.

• 41:10Brian Keating?

  I view that, sorry, again, it's hard to draw. But I view that like saying to me, you know, Keating, you dye your hair, you know, stop dying your hair. But I view that as a compliment. Oh, you sound like Claude. Well, I wrote it myself. So Claude sounds like me or sounds like Ethan.

• 41:20Speaker 04

  It affects you. You get in your own. So what I was going to say in that big picture is all the things you talk about, I would not use AI first. Referee review, never going to put it through AI first. I would be, and you know, it depends on the journal policy, right? There's all these, like, some of the journals I write for, you're not allowed to use AI for everything. Sometimes you're encouraged to do it. I will always do AI as a second path, never as a first path. It's actually interesting. I'm, you know, I study innovation. And brainstorming was invented in 1955 and by 1957, we knew it was a terrible idea. Because the problem is as soon as you do any group process, your ideas fit the group. All right. So you always want to start working on your own before anything else. Adam's written about this as well. I feel the same way about writing. If the AI writes, it will write a seductive, wonderful section that I will not be able to get out of my head and will influence my writing. So I have to write my own replies first to things, right?

• 42:05Speaker 04

  And for reviews, I have to do my own review. Then the question becomes, and the bigger question of science, okay, the AI is going to fall. I mean, what is the, so it was funny. I fact checked coexistence with AI and it was great. There was no errors. But it also not only found errors in my work, but it found an error in the work I was citing. Like you would actually say, actually, the statistical analysis here is wrong. This cited the wrong city. So I made the mistake of putting co-intelligence into this too. And it's like, oh, like I made all the, like, and I had a fact check. I paid fact checkers, like I, I read it myself. I had readers. I, I'm a very careful writer about these things. I read every art. This is pre any AI that could do anything. I read everything myself. And there was just like, and I'm sure any book, right? Every academic article I've ever wrote is full of errors. So part of this is our decision too, about what does science and writing look like at this point?

• 42:48Brian Keating?

  Yeah. I mean, the dirty secret of publishing is that, you know, 90% of the time, at least, you know, previously, we're emailing Doc X files across the country to our, you know, agents and our editors. And I don't know, you know, I'm writing my fifth book now. And I don't even talk to my editor half the time. I just put into Claude Fable to turn it up and say, where I say it has guardrails. It can't write anything for me. It can tell me what I need to dictate or to type out or write. Actually, I want to bring that up. Typing, you know about the concept of lock-in, I presume, or maybe I'll explain it. But, but for my audience that may not know it, you and I are of a certain age that, that we remember using actual keyboard typewriters, right? I mean, you used a typewriter probably when you were a kid. I certainly did. I'm a little older than you, I think. And the keys were set on the QWERTY keyboard such that they wouldn't interfere with each other because the word frequency

• 43:33Brian Keating?

  was set such that it was less efficient to type with than Dvorak, which is the other competing. And I view LLMs plus GPUs in that same vein. In other words, they're so good at what they do that there, there's never going to be sort of a replay. There's no economic incentive to come up with some other state space modeling, or I forget the other things you talk about in the book. But even like GPUs are so dominant that, that even a company like Google with their TPUs, they're not really competing as, as well, as much as if two other frontier labs that use Nvidia. So GPUs married to LLMs lock us in, in the exact same way that QWERTY doomed us to have less efficient typing. And so I'm worried that, you know, with, with the fact that these LLMs need humans, and this is my pushback to Roman and to the doomers.

• 44:14Brian Keating?

  Because they need us, they need a training data. Like, if they have to wait for Fast and the Furious 16 to come out before I can have a theory of everything. It seems to me like human physicists are going to be okay. What about lock-in and other phenomena? Are these things a victim of their own success in some sense?

• 44:28Speaker 04

  LLMs are a evolving concept, right? So the modern LLM is a bunch of stuff glued together. It's the main LLM plus this reasoning loop that's kind of put on top that was invented a year and a half ago. That's why AI is good at math. And there's other systems. I think we will see LLMs as part of a series of systems that we talk to. They're remarkably good at human facing stuff, right? But you may end up having, you know, AlphaFold, which is a specialized model built for protein folding. You might have, you know, physics engines. Certainly part of the reason why LLMs are good at math right now is that Lean exists as a formalization language, right? So we're going to see LLMs pair with other things. And I do think that now that AI is a thing,

• 45:07Speaker 04

  you are seeing other angles. There was a bunch of vast decision-making system, one LLM, is system one thinking systems that just came out a couple of weeks ago, where the AI makes very fast judgment calls, right? And now OpenAI released their competitor Jav already, as did Cloud. Like, so I think that what you're seeing is systems being built, right? LLMs are not one thing. There are many things harnessed together. I do think there's a lock-in in direction. We're kind of lucky in some ways that the AI we ended up getting was one that talks to humans really well and one that anyone can access and use with that specialized coding language. Like, science fiction, all of this stuff tended to be this kind of crazy, like, the world's computer and you had to be a techno priest with the right punch card. And we ended up with something you could talk to. So that, that's a good outcome.

• 45:48Brian Keating?

  I have this concept called the academic hunger games. You know, it's like you first compete in high school, you know, to do well in grades. And then SATs, PSATs, I've got teenagers going through this now. And then it gets into a good college. You want to go to Wharton, you want to have Professor Molik. And then you want to do well in his class. He gives you a lot of recommendations. Maybe you do research in his lab. Then maybe he writes you a grad school. You go to get your MBA. You do whatever and then you keep going. Then it's a postdoc. Then it's a assistant professor, if you can do it. And each step, the probabilities are going down incredibly low. But if everybody has access to super intelligence, as your illustrious alumni now forces us to call it, then does anyone have an advantage, Ethan?

• 46:24Speaker 04

  Right now, the answer is yes, for the four, the things we talked about here, right? That some people are better using the systems than others. Expertise matters. Agency and willpower matters. There is a world of like, you know, in the current world, in the short term, there's definitely differences. And people's ability to use these systems is a return to expertise and intelligence and willpower and using these things. In the longer term, if they really are just better at everything and you're like, you know, so how's that string theory stuff going? And it's like, oh, here's the answer. You're like, you know, what do you do then? And the answer is no. I mean, you know, I don't know what your advantage is at that point.

• 46:53Brian Keating?

  You mentioned a word in the, in the book that I hadn't heard about in a long time, which is MOOC. And it made me think, you know, there, there are two phenomena that I thought would be, you know, at least, at least two, but maybe more, actually many more now. But instead I thought academia would be doomed. You know, first it was the MOOCs. I never really thought those were too engaging. I mean, they, they exist and there's things like Circle and, and other online communities, School with a K and all sorts of things where you can learn stuff. But then I thought, oh, AI will be great. And even before AI really got that popular in 2020, you and I remember teaching during COVID and on Zoom. And, and it was just horrific. And I talked with, with Adam Grant about this last week, but I want to get your impression.

• 47:35Brian Keating?

  Like, why is academia so resilient to these, you know, kind of just, just insanely disruptive events in the case of COVID and tools and technology like MOOCs and AI?

• 47:48Speaker 04

  There's the good reason and the bad reason, right? The bad reason is inertia and not just like the institutions themselves. I mean, you mentioned Penn was started by Ben Franklin. Like, it's not changing fast, right? Like, everything, it's a, all of this stuff has been flash in the pan stuff that didn't actually change education forever. And, you know, so, and then also education's hard, it turns out. Like, part of why I write about in the book is I was a, you know, I was involved in MOOCs. I, I've been doing AI, you know, prior to AI, I was doing teaching with games, interactive material. And it turns out it's just really hard to move the needle on education. So like, all these technological advances just weren't good enough. Like, they seemed the theory they should replace the university. They didn't do that. So the negative side is like, nothing has kind of worked. The positive side, the university side is universities aren't an entity, really.

• 48:32Speaker 04

  There are a bunch of professors making choices. Ed Wharton, there are people who will teach with, you know, overheads today, overhead projectors, if you remember those. And my classes are 100% modified with AI. I don't know who's right, but there's an evolutionary kind of game. You were talking about how the rarefied air of academia, part of what we do is there are people who are competitive. There's, you know, professors, really interesting reactions to AI, where you hear accelerated retirements from some people who are like, I just don't want to deal with this. I do not ask to be teaching in a world with AI systems. On one hand, on the other hand, you have cases where people are absolutely excited by the idea that they can solve hard problems before. So the fact that it's people and a bunch of really smart, motivated people who are all out to prove that they're the best is also our advantage, as well as can sometimes make faculty meetings challenging.

• 49:15Brian Keating?

  Okay, last segment, and I want to talk about parenting. You're obviously a very devoted father. So I want to talk about raising children. I, I did a podcast with a friend of mine, alumni from UCSD, Vivian Ming, about her book called Robot Proof, which kind of is, you know, encouraging a lot of these different, you know, principles, first principles for raising children and, and surviving, you know, and thriving in age of AI and robots, as she calls them. So you mentioned before, and we discussed it, my resistance to letting my writing be changed by the predilections of these models, because I feel like this is who I am. Maybe people will suspect it, just like they suspect I dye my hair with just for men. I'm not, I'm not talking, but in reality, these things are changing us. They're prompting us.

• 49:55Brian Keating?

  And in a certain sense, as you, as you have, you know, mentioned, you, you change your writing style. I mean, you didn't have to, but you did. And I'm wondering, what is that going to do to children? You know, if your kids were young, when would you give them access to these very powerful super intelligences?

• 50:10Speaker 04

  We can talk about the narrow case of education, but let's look at the broad case, which is, I mean, you should be monitoring what your kids are doing with technology, right? Like, I mean, I think it's very reasonable for you to say, no using AI through middle school unless I'm sitting in the room with you. You should do fun projects with them. Let's, you know, use it to create some images. Let's create a story together. Let's code together. Like, I think it, like, because there's a creative aspect of this technology that kids love, like, let's build a, you know, a bluey guitar and see what happens. Like, you can guide, like, you and somebody have this capability to guide really exciting, interesting, interactive experiences that are not watching a screen, but are not, that are not passive. Screens are less the problem than passivity, right? So do the same thing you do with any screen. Monitor what's being going. Don't give your kid access to a computer without doing this. Check if there, if there's AI applications on their phone, you want to make sure that what they're doing.

• 50:56Speaker 04

  Now, let's talk specifically about education. So for younger kids, I would not let them use AI to do for learning themselves because we know the problem is they'll shortcut learning on their own. Again, observe, I've seen kids build little simulations and games, like, they're obsessed with sharks. Tell them about my favorite sharks. Let's have a conversation about sharks. If you're watching it, that's great. But for actual homework. Younger kids, the secret is you take a picture with your, your favorite AI app of their homework and you ask the AI, how would you tell a kid to how to solve this problem? And so you get to be the genius parent who has the interaction explains with them. When they get into high school, you might want them to use AI for learning. But the high school systems are pretty good now for Anthropic and OpenAI and for Google. They do lock down some stuff. They have some accountability, right? You can't do any, get yourself in too much trouble.

• 51:40Speaker 04

  But on the learning side, you need to lock them down into using study mode for these things, which makes the AI tutor rather than giving you the answers. And it turns out AI tutors are quite effective and they're very useful learning tools. So again, observe, work with them. And you're hopefully also teaching them values that they should write on their own. They should like, we're going to have to do this like other technologies. I think there's a lot more positives than there was in social media for kids using this, right? Like, there's a lot of upside. The models are pretty gentle to, you know, increasingly better at mental health issues. People get help talking through them, but you want to observe this. This isn't something where you're like, turning kids loose with any technology is a mistake, right? And, and you'll know your kids. Some kids are like, oh, they will like, they'll never use this for any bad. Great. You're the parent. You get to make decisions about that.

• 52:22Speaker 04

  But I think part of what you want to do is be thoughtful and I don't think it's that hard. It just takes some time to do.

• 52:27Speaker 00

  The wheel had roads. Electricity had the grid. Innovation needs infrastructure. Plaid makes AI work in finance. Visit plaid.com/ai to learn more.

• 52:42Brian Keating?

  Last week, Adam was on, your colleague, he mentioned the importance of friction and the counterintuitive benefits of having people not agree in, in organizations. As a psychologist, you know, you think, oh, the group think and the brainstorming, as you mentioned a few minutes ago, you know, all it's brainstorming, you know, but it turns out that it just converges. And as my past guest Seth Godin once said, you know, if you just let things evolve by committee and brainstorming, everything on the internet becomes pornography. But what level of friction is important? And I guess concomitantly, you know, is boredom good? I was bored a lot as a kid and, and it made me, you know, kind of daydream and come up with ideas, some of which were useless and some of which were important for my intellectual create the development. How much boredom and how much friction, you know, should we let kids fail?

• 53:26Brian Keating?

  You know, when my son loses a baseball game, I say, good, I think it's good. You learn more from losing than winning.

• 53:31Speaker 04

  So there's three things there. There's friction, which is absolutely, I mean, friction turns out to matter for anything you want to make gains in, right? So we already know if you want to, if you want to lose weight, you have to diet and that's painful to work out. You have to lift weights that, you know, like that's not, like that the repeated friction matters and in learning is desirable difficulties. If you're not being challenged and pushed and you're not uncomfortable, you're probably not learning enough. Now, we can make things too uncomfortable and then you lose interest, but we have to kind of keep you on that right line and teach, it's hard to do that with a class of 80 people to have everybody at the right level. But that's why people cheat, right? There's incentives to do it, but also like homework sucks. It makes you better at the test. It turns out testing works like it's really annoying. All the things that you're being taught to do in school actually matter. Now, it doesn't mean people are assigning homework thoughtfully or minimal amount of homework for maximum effort.

• 54:16Speaker 04

  We don't even know how to do that, but there is this kind of piece here that people cheat. So friction does matter, right? Friction between people, friction between other, other sets of stuff. And it's similar idea diversity. The AI produces fairly similar ideas. You want to have different ideas coming into play here. And then on top of, of the kind of friction piece, you know, we have other aspects of kind of working with AI that matter too, right? Making sure you're not using it in the wrong kinds of cases, like you are keeping your own skills up to date. You're being strategic about what you're de-skilling yourself on. Like, I'm very happy to de-skill myself forever on my school's HR portal. Like, I don't ever want to learn how to use that. And if the AI does it for me, great, right? Like, that's a friction. My grandfather built the fire suppression system for the launch pads at Cape Canaveral and doing it with a slide rule in Matrix mouth.

• 55:00Speaker 04

  I have no idea how to use a slide rule and I can't do Matrix mouth in my head and I am fine with that. So we have to make choices about what we want to de-skill and what we want to de-prioritize and what we don't. In terms of boredom, it's really interesting. Boredom does create creativity and there's this great set of papers on it, but only when people feel agency and positive and that they could do something with it. Otherwise boredom leads to disaster. So people who are bored are much more sadomasochistic. They'll do anything to stay engaged. Like, we don't like to be bored. So if there's something interesting you can do when you're bored, and actually this is why kind of I like AI as opposed to passive recording. Like, I build weird games. I work on it with poetry. Like, it is a really interesting boredom killing tool that has a creative aspect to it too.

• 55:38Brian Keating?

  Second to last question, this one's solely out of left field. Do you believe that extraterrestrials exist?

• 55:46Speaker 04

  Statistically, they kind of have to, right? Like, it would be kind of weird if they didn't.

• 55:50Brian Keating?

  Okay. So now you're, you just launched me on my favorite hobby horse, which is

• 55:53Speaker 04

  Is the Drake equation out of date now? I don't know how, I mean, I have a

• 55:57Brian Keating?

  Well, the Drake equation is like P doom. It's just, it's a useless, like, no actual scientist. Let me, let me look at P doom and then use that to for you. I'm like, nobody uses that and nobody uses the Drake. Oh, the Drake, you know, it says, because the uncertainties, you know, as a good scientist, you know, yourself, you know how important it is to account for systematics and statistical uncertainty. And the Drake equation hasn't either. It's just a number. If my students gave me a number, you know, here's the

• 56:22Speaker 04

  There's five aliens. I'm glad you asked. Five aliens.

• 56:24Brian Keating?

  But that's not what I'm getting to. My kind of counter example for it is that we don't see, you know, like if we, if we were sending out probes and whatnot, you know, sending meat bags out over the solar system and the universe, then that would be kind of inefficient versus sending AI, you know, send some artificial intelligence to it. And the fact that we don't observe it to me gives some, you know, inkling that maybe we aren't, uh, maybe we are alone. Maybe, you know, space is just too big or for all intents and purposes, effectively alone. We don't, we don't, we will not actually make contact, even if there are things that are fundamentally unobservable. The way I want to conclude with, with this wonderful conversation is, is the next phase, the next, next phase. I mean, this book is called Coexistence, first one, Co-Intelligence.

• 57:05Brian Keating?

  What's the next, next phase going to be when it comes to our creativity, you know, entrepreneurship, productivity, thoughtfulness, scientific discovery, parental, you know, oversight, education? What's the next, next phase, Ethan?

• 57:19Speaker 04

  I've thought of two titles for the next book, right? Depending on how things work out, and I think one of them is co-evolution, right? Like, we work with these systems, these systems shape us. And the other is co, oh no, which I think I'd rather not write. It's hard to know what the next phase is. Like, even this book, right? Like, I think the agentic era will last a bit. And part of the back half of the book is that our systems actually don't absorb this change as quickly as, you know, we do, right? Not much has changed in the outside world here, but I think that we are going to evolve with these tools. I mean, you know, one of the most fascinating sets of stuff is there's a set of research suggesting that humans co-evolve with tools. Like, for example, if you hit a, if you're holding a bat and you speak of baseball and you, or a metal rod and you hit it on some length, you can estimate where on that metal rod you, a hit happened by the vibration in your hand.

• 58:04Speaker 04

  It's almost like a spider can tell where something is on its net. That is an evolved reaction probably to our own, you know, to, to working with tools for, for enough time. So we, we co-evolve with our tools and I think we'll co-evolve in it, obviously not in a biological way, but in an intellectual way quite quickly with what AI should fill in, what don't, what it won't. We're running a big experiment for better or for worse on that right now.

• 58:26Brian Keating?

  Looking forward to seeing what Bryce Harper does next. Uh, Ethan, this has been a lot of fun. I encourage people, again, get all three versions, the collector set. It will be hard to convey how fun it is to listen to really these real time experiments that you're running in the book and on your Substack and all the great stuff you do online. Ethan, Malik, thank you so much for joining us today. Good luck with the book.

• 58:48Speaker 04

  You're having, this is a great conversation. Thank you.

• 58:51Brian Keating?

  Ethan just told us that if AI froze today, it would still be enough to remake almost all intellectual work within a decade. If that changes how you think about your own job, subscribe, turn on notifications. Tell me in the comments which of the four edges you're betting your career on. And for the idea that Ethan borrowed, that language is a model of the world, watch my conversation with Stephen Wolf from, linked right here. Thanks for watching. See you next time on Into the Impulse.

• 59:14Speaker 00

  The wheel was brilliant, but the road made it iconic. Electricity was powerful, but the grid brought it to every home. Innovation needs infrastructure. AI is no different, especially in finance where AI can't run without context, but it can run on Plaid's infrastructure. Built over 13 years to understand why money moves. Plaid makes AI work in finance. Visit plaid.com/ai to learn more.

• 59:45Speaker 01

  Stonefire naan is made with real buttermilk and ghee. It's the perfect flatbread for any occasion.

• 59:50Speaker 02

  I put out Stonefire naan with dips at every dinner party.

• 59:53Speaker 01

  I make naan grilled cheese when my daughter has friends over.

• 59:56Speaker 03

  I use Stonefire naan to reward the humans who do my bidding.

• 1:00:00Speaker 02

  Right. It's perfect for camping trips.

• 1:00:03Speaker 03

  Dance for it, feeble mortal.

• 1:00:05Speaker 02

  Sorry, is he talking to me or?

• 1:00:07Speaker 03

  Dance.

• 1:00:10Speaker 01

  From our tradition to yours, Stonefire.

• 1:00:13Speaker 02

  Together, we naan.