Radar: Never Normal Heroes – Ethan Mollick on Co-Existence & the End of Apprenticeship

Thank you for having me. I'm excited to be here.

You know, it's interesting, right? Like, I mean, so there's two ways of saying how frustrating to write a book. This is my fourth book. It is the least frustrating book to write because AI helped me. I wrote very, very, I mean, very little AI writing in the book at all, except, you know, the like, I wrote basically every, you know, everything otherwise it sounds like Claude, right? And so, and I'm a writer of the old school, so I write all my own drafts, but like, honestly, as a researcher, it fact-checked every book. You know, I paid fact-checkers, I got better fact-checking, more research. It made it much easier than I've ever written a book before. On the other hand, it's still, writing still is terrible. But on the other hand, things are changing fast. I think the thing I've gotten better at is kind of seeing where things are going and focusing a little less on the technology and more on the trajectory, you know, but it's called The Next Phase of AI because I think it's valuable in the pre-AGI world as long as that lasts.

For everybody listening, October. This was before AGI happened.

I mean, so nobody is, right? I mean, I can tell you as somebody who talks to the AI labs, it texts the government officials, like CEOs, nobody's got a hand on this change happening right now, right? Like, all we're doing is discovering stuff. I mean, you're leaving alone that Opus 5, which I had some early access to, turns out to be even better than it first appeared once you start poking around at it. That the labs are, you know, several iterations behind. They've announced that they have better and better recursive self-improvement. I mean, we are in it right now, right? I mean, this is a kind of classic, you know, I think people, what happens next, we don't know outside of science fiction, but this is what a takeoff kind of situation feels like, right? News is happening more quickly, but also even if this was just like the steam engine and wasn't evolving itself, adoption is happening very quickly and adaptation has to happen. So there's news coming from everywhere and then there's waves of news, right? So people who are regular listeners or readers of your Substack probably have a better sense of what's going on, but like part of why there's a wave of panic is this hacking stuff that happened to Australia is sort of old news for those who've been following things since Hugging Face and the METR reports, but is obviously bigger news to a larger section of people. And so it's all kind of happening at once. The only way to keep up, I mean, unfortunately, so like online social media, especially X, for better or for worse, is where most of the interesting stuff is happening here, but it's getting very hard to stay up to date on all the stuff that's occurring. I think if you have an idea of where the sweep is going, it's better than if you don't, right? It's less about this tech news of each day and more about, you know, I think a lot of stuff we expected to happen is now happening, even if more quickly than we thought.

Yes. I mean, I think it's, I mean, it is very clear if you use AI a lot to do a lot of things, if you're not consciously or unconsciously compensating for the jaggedness of that system, and that could be anywhere from the jaggedness of human systems. I still need to go and talk to a person about this thing to the idea that, like, listen, the AI results are kind of similarly, and it might look like you're running a two-line prompt, but you actually know the AI really well. So you're guiding it towards the area where you know it's going to do really impressive and away from the area where it won't be impressive. There's a lot of like these jagged edges and writing is a key one. I mean, AI is, it still remains a terrible writer and plotter of long, like if you were doing your Substack entirely with AI, people would stop reading at this point, right? Like it just can't pull together the argument as well and it's fluffy. That's a jagged frontier, right, of these systems. So I think there's still jaggedness there that's real, you know, but there's less jaggedness than there was, right? I mean, math used to be the obvious weak spot of AI two years ago. And now it's, you know, conquered mathematics. That's a real thing.

I mean, I think it's going to be kind of like that for everybody to some degree at some point, right? And I mean, it's interesting that for all the focus going into math, I think coding is the area where this is happening fastest and most weirdly, right? True. And the thing is the jagged frontier is jagged enough that if as a coder you are freed from writing code for most coders and giving directions for even the best coders in the world, that feels exciting. But I mean, that is a field that is in complete transformation. What is the art in this now? You're being pre-, you know, I mean, it's very industrial revolution looking, right? So the same kind of thing, art and craft was like one of the things that people don't get about math, right? It's a very craft-heavy field. Like they're not really trying to prove those theorems. They're trying to do a lot of, I mean, they are, but that is a small part. Like in proving the theorems, you learn things, it's beautiful to do, you are guardian of this great intellectual tradition. So there's lots of other things that are threatened when the core thing that you publicly are famous for is threatened, right? And, you know, there's debates in math whether it's useful or not to have this solution to this problem if we didn't do it by humans because we didn't learn anything. Is it like the equation doesn't really matter very much? And so I think part of what we're going to see is it's sort of like the flip side of what we're seeing in organizations. Organizations were starting to learn, do we actually want that report? Like, was the report the goal? Was the PowerPoint the goal? Because we could now do 4,000 PowerPoints. So, obviously, the PowerPoint wasn't the goal. There was something else that was the goal. I think we're going to find that field after field where we're going to actually reexamine what the goal is. Like, there's a, you know, what is the game mathematicians are actually playing? And I think that kind of search for meaning is going to be a wave across everything, right? And it's compensated for in some extent by the fact that now you can do, you know, electrical engineering if you couldn't before, and now you could do coding and you couldn't before. But there's losses and gains that are coming for all of us.

So, I mean, there's a few things. One is I actually think the value of it, I mean, I'm contrarian about this and you have to recognize I'm a professor who also went to 26th grade myself before, you know, getting degrees. But I think that there is, I think education becomes more valuable because it's the only place where we can actually both force you to get wide knowledge and deep knowledge and certify that you've done that. And like it got a little disconnected, right? There was sort of an idea in business school, look, you go through, you do the work you're supposed to do, you get your degree and, you know, your socializing is part of it, learning is part of it. But I mean, the certification was sort of the weakest part, right? Like whether you got an A or an A minor. But I think that increasingly the mission that universities had of, we need to teach you to be broad in knowledge and understand many topics. We need to go deep in one area. We need to certify that you've learned about this and that you are good at this. That mission feels more real than ever. And the disruption to teaching styles feels more temporary compared than other fields. Like, yes, it's a complete mess right now, but education is always a mess, right? Like we pretend case studies are, there's no research that says case studies are a good way to teach. Like none, right? And so like we still do that and like we pretend that's the important way to teach. We don't know the answer to these questions yet. And so what we do know is active learning and tutoring are both good. So we'll end up having AI tutors outside of class, active learning inside of class. We'll figure it out. I'm not worried about us. I'm worried about us as a link in the chain to that city job in London. What I'm worried about is what happens is you're a generalist coming out of LBS and you become a specialist by working at a bank that teaches you stuff. And it was a great deal because we taught people the same way we've always taught them, apprenticeship. And apprenticeship just broke 100%. Right? Because the whole idea of apprenticeship was incentives were aligned where you had junior people who were desperate to show they were good and they would do everything they could to show they were good even though they weren't very good and middle managers who even if you yelled at everybody every day they were still learning and that all broke because everybody's using AI for everything. So I'm actually worried about the second phase. I almost wonder if, you know, in the book I proposed two solutions. One is that companies need to start being serious about learning, which they're not really. Like the function is not zero to one learning at a uni-, like it's about topping people off and certifying them, or they will, and they will need to become less efficient to do that because learning is not efficient. They'll have to have people work without AI to do stuff, or they're going to have to partner more deeply with universities. You may end up having a third year, you know, or second year at LBS where, you know, the, where what you learn during that year is going to be, you know, associated with, you know, Baillie Gifford or whatever the, like, sponsored by this organization that is paying for you to get formally trained there.

I mean, I see more and more experimentation. I'm very doubtful of that approach. One, like I said, deep knowledge is one thing and wide knowledge is another, right? Like increasingly, if you know a little bit of, you know, like you're this, you know, you're obviously mixing genres. You're talking to me from a cathedral full of Mac things, right? Like you have to understand a bunch of different stuff. What makes AI great is if you know a little design and know the language of design, how to appreciate design, you now can do design work in a way that defaulting to Claude won't let you do. If you learn a little bit of how coding works, like there is a broadness of this too that's intellectual that's important and I think will become more important. And I'm also extremely doubtful that we can target people for the new world of, like, I don't think Andrej Karpathy has been particularly great at predicting what the world's going to look like in the future. True.

And I think that, and I think there's a, there is a Silicon Valley, you know, crusade against education for a wide variety of reasons. Some of which is very well-founded, some of which is philosophical. And I, you know, and but I think that there is this sort of idea of like, and if they do, like, and it's motivated, right? If they could skim the top set of people off, look, it's just a natural progression of entrepreneurship, which is if I can get 16-year-olds or 17-year-olds to join my university and skip college and I fund them, I can fund them for cheap and they'll do labor for me and that's great. And so I think we have to be careful of motivations. I do think we're going to see a lot of changes in education.

So, it's interesting. There's two sides to this. One is, I mean, you know, to, as Claude would say, you're absolutely right. The Industrial, I mean, the Industrial Revolution had really weird effects when you get into the micro level, right? Like it did actually raise standard of living for huge amounts. Like wages might have been flat, but the standard of living increased by a huge amount, but it also decreased in some areas. I mean, it was the era of like Dickens and Marx and like every, it was chaotic to live through, right? The end results are undeniably good, right? Like you have to really reject public health along with modernity to say that they were, you know, that, you know, and starvation, like it is undeniably good, but living through it kind of sucked, right, for a lot of people. And part of how wages ended up growing in the end was there was a fight, a negotiation between capital and labor working through government at strikes and unions, and it was a process by which workers clawed back some of that returns. I think when I talk to economists, what's interesting is we have much better tools right now, like we have much better tools for adjusting in this kind of, you know, than we did in the 1800s to adjusting to boom-bust cycles, to adjusting to worker displacement. The debate is what happens, right? We don't know what the, like, we can't do policy till we actually know, is this going to lead to a boom or a bust for which kinds of workers? There is also no doubt at the individual level that we get to make choices about whether this is a reverse centaur process. I discuss Cory's argument in the book. I think he underestimates the power of the technology to some degree. Like I think that there's a bet that this is an automation technology rather than the, like, I can ask the, to do, and Epoch has a study showing you could with even just Claude Opus 4.7, you can get two to 18 weeks of engineering work out of it. That doesn't, that's not a, like, that is not you pull the lever kind of stuff anymore. I agree if we get to a world where all we do is verification, that's a disaster, but that's not where we're going either because the AI is going to be superhuman at verification if it isn't already in your field. So I think there's something more profound here than the Industrial Revolution. So there is obviously Industrial Revolution parallels, right, which include the idea of disruption of labor versus not. But also the people being attacked now are also the people who have the most money, the most levers of power. Like I often tell people, do you really like the mathematicians are pushing back against math being changed, which is weird, right? Like, but wait till they come for the doctors and lawyers. Like there, like Congress in the US is like a third of it is lawyers. Do you think the bar associations of the US, whatever the equivalent is in Belgium and, you know, the UK and everyone else are going to, like, we are going to be required to use human lawyers along with AI, no document can be signed without a human lawyer signing off, you know, doctors will have the same kind, like we're going to have adjustments that are, that slow this down, or it's going to be everything, everywhere, all at once, in which case we're in a post-industrial society and I don't know what happens, or post, you know, need society, or we're all doomed. I don't know, it's fictional pretty quickly.