How AI makes initiative beat intelligence — Tyler Cowen

[00:00:00] Speaker 1: I'm here today to bring you a mix of optimism and realism. The optimism is that AI will do and is already doing remarkable things. So just yesterday, it proved a striking mathematical theorem that some people called the greatest achievement of AI so far. It will extend our lifespans, give us a wealthier society, more entertainment, more instruction. But the realism is no less important. And the realism is, first of all, a lot of people won't like what is going to happen. And second, a lot of the benefits will come more slowly than advocates would like you to think. So, it will feel bad for a while. I would stress the following point. No one in this room is old enough to have lived through a radical technological revolution. I believe my grandmother did. She was born 1905. She saw the birth of consumer society, automobiles, airplanes, radio, television, so much in the first 50 years of her life. But today we all have grown up, the internet was a change, computers are more important. Our lives have not really been disrupted. But what is coming is that virtually all of us will, in a radical way, have jobs that are very different from the jobs we expected. They might be less routine, they might be more fun, they might be higher paying. But they will be different, people do not always like change and uncertainty, and furthermore, when your kids, your grandkids, come to you and say, "Grandpa, what should I do for a living? What skills do I need to learn?" You won't actually have that great an answer. So there was an older world we grew up in, we all grew up in, where if you were very smart, went to a good school, worked hard and followed the rules, the chance that you could end up, say, living in Manhattan with a partner in a consulting firm, law firm, earn $2 million a year, have a great life, feel like you're master of the universe. Those are actually the people who might lose. The people who will win are the people who are best at taking initiative, figuring out how AI works, figuring out how agents work, doing something different. In academia, Ethan Moloch, you heard from him before, he took initiative. The founders of SANA, they took initiative. So it will radically remix status in society, again, in a way none of us have lived through. And this will hurt, because when some people go up in status and some go down in status, I will tell you, those who lose suffer more psychologically than those who gain. So you may have heard Eric Schmidt gave this commencement speech, I think two days ago. And when he mentioned AI, the students started booing him. AI doesn't pull that well. It pulls least well with young people. So this will be difficult. One thing that is not true, AI will not bring mass unemployment. But it will change most jobs. Key feature of AI, there will just be many, many more projects. More than we can imagine. You know, they gave these Swedish slides here. This shows more projects. Sweden's a project. It's a great project. But there will be two new, not entirely new, but two categories of jobs that will just go crazy for how many new jobs there will be. The first is what I call running experiments. So AI will have many, many new ideas, as it does already, most of all in the area of biomedicine. We will need to test those ideas. In the world today, we're like, only PhDs run experiments. That's on the way out. We don't have enough PhDs or they want to do other things. So, so many of us will be testing ideas of AI. Some ideas of AI will figure it out through simulation. So a lot of coding, it's a kind of internal system. You just, the AI figures out how good the AI is doing. But, most things we want to test in the real world. So huge new job category, running experiments. Another huge new job category is just gathering data so the AIs can do other things. So, Mercor puts out an ad. They want people trained in basketball analytics to teach the AI how to watch a basketball game on TV or live and do basketball analytics. So that is the act of gathering data. Most of the world is not currently expressed in the form of data. Insofar as we turn the world into data, AI can do much more for us. That process will not end anytime soon. Huge new job category. Energy and compute. We need to build those out. Is it going to be nuclear fusion? Will it be something else? Will it be installing a lot, lot, lot more solar panels on the roofs of homes or in the desert? Whatever you think it will be, a lot more jobs. So it's a wealthier society. There is still scarcity. There will be more entertainment, more instruction, more artists working with AI. So, believe me, mass unemployment will not be a problem. What might be a problem is you may not like having to adjust to the world of AI. I'm in academia. I speak to my colleagues all the time. On average, they do not like it. They have mastered a skill publishing research papers. And now they're supposed to learn some other skill working with the AI to create something research-like that embodies knowledge. We'll see how much they enjoy that. But I think in general, you know, tech and healthcare will be a huge growth sector. Healthcare forever. People can never live long enough to be happy. I think, "Anthropic is still hiring." I asked Sam Altman in my podcast with him, "Sam, when will the finance department of open AI like be 80% agents?" He said, "Not anytime soon." So tech will continue to grow, though many, many jobs will be taken away, as already been presented here. But just healthcare, my quite subjective opinion, but I think over the next 30 to 40 years, most of the things that kill you, we will fix. You might die in accidents. You'll probably still die of old age. I'm not sure what we can do with your brain past some point. But I'm here to tell you, buckle your seatbelt. Don't go bungee jumping, whatever. This is tough for me. You know, like, what's really the time horizon here? I'm 64, so I'm exercising more. When I hear, "Oh, it all might be great within 30 to 40 years," I'm like, "I'm 64." Like, you know, my daughter, that's wonderful for her. The grandkids, their age is three and four, incredible for them. Like, "Can I hang on to 84?" Well, I think so. When I hang on to 94, well, maybe along the way, AI will help me again. I'll get to 94, then I'll get to 99, and then finally, that will be it. But that will be, in essence, the way AI will be sold to the public, because it won't be that popular. It will help us against China and in foreign conflict, and we will live much longer and fix many things that go wrong with our bodies and maybe our minds, and along the way feel disoriented and sometimes demoralized. But we will live through one of the most fantastic chapters of human history ever. So be ready. I would just say, don't be too discouraged. The benefits are truly large. But here's one of my views. People I know in Silicon Valley, they make fun of me for this view. They think I'm crazy. And you know what? I think they're crazy. So here's the view. My view is that AI in the medium term, certainly in the short term, it will boost the economic growth rate from an average of about 2% to about 2.5%. People are like, well, if AI is so powerful, if it's so smart, if it does these amazing things, proves these math theorems, how is it that it only gets us to 2.5%? And the problem is what I call the human bottlenecks. The smarter the AI is, the harder it is for you to work with it, the harder it is to get organizations to adopt it, the harder it is to figure out how do you integrate AI into an institution. Right now, we have all this freelancing, people on their own. They may not even speak up. Oh, I had to write this memo. It used to take me five hours. Now it took me five minutes working with Claude, right? Well, that's great. It gets converted into leisure time. I'm all for leisure time. But it's not really converted into higher productivity for the organization, because the organizations still are not set up to take those gains and have them with people making each other more productive, because a whole set of jobs is changed, and those people do other things, and a bunch of things happen much more quickly, and then we get to the next thing much more quickly, and the whole company is more productive. That's already happening in coding. It's happening in a lot of startups, in so much of the American economy. That is not happening. So when I say we go from two to two and a half percent, I'm not bearish on the AI. I wouldn't say I'm bearish on the humans, but we're holding it up a bit. I actually think that's reasonably stabilizing to go from two to two and a half percent. And some of my Silicon Valley friends are like, "What? You don't think we're going to grow 20%, 40%, 200% in a year?" And I'm like, "No, I don't think we're going to do that. On average, it takes the FDA 10 years to approve a new drug. How frequently in this country are new nuclear power reactors approved? Until recently, there hadn't been, I believe, one approved for decades." So that is the stickiness of the world. It is in many ways unfortunate, in some ways stabilizing. And we are combining this radical new technology with that stickiness. The nice thing about getting from two percent to two and a half percent, if you do budget calculations, you all know about the U.S. debt, right? You feel we're screwed, my kids are screwed, grandkids are screwed. Right now, I think debt, it's over 38 trillion. It might have hit 39 trillion. What does that number even mean, right? Can we conceive it? But if our economy can grow two and a half percent instead of two percent, that debt, rather than exploding and making us the next Greece, that debt actually converges to a manageable level. So the way to get out of this hole, like if you work in AI, you are our savior. You're our plan A, that we're going to grow a bit faster and we're not going to see a huge tax increase. We're not going to see our Medicare, Medicaid benefits taken away, our Social Security ravaged. We will see America's fiscal position converge to something sustainable. You know, it's the other piece of news. You are our plan A. There is no plan B. So if you're sitting in this audience as our plan A, I applaud you. So this is truly important. There'll be a lot about AI people don't like, but what they don't see is that if we don't get AI, there is no other obvious productivity savior on the horizon and we're all kind of screwed. So I'm pretty sure it's going to boost the growth rate. Now, there's still some people like, oh, it's all hallucinations or it only does a few things that fewer and fewer people believe that. But here's a way to make the half a percent boost in the rate of growth plausible. In the years in this country 1995 to 1998, we got some very basic software and we got software to do inventory management for companies like Walmart. In those years, our growth rate, it basically went up half her percentage point. So just ask yourself, do you think AI can do for us what we did in the mid 90s? I'm pretty sure the answer is yes, even if we're stingy with approving data centers, even if we make some other mistakes. The mid 90s were very nice. There's some nice advances. Real wages went up, but it wasn't unprecedented. It wasn't utopian. So all we've got to do is a repeat performance, but make it last for more than three years, because the smarts behind AI, they're not going to go away in three years. So I'm not convinced, but pretty optimistic on the positive side, and I think we're going to get there. Household budget. This is another thing that will worry people. So economic growth will be higher, but the gains will go to people who take initiative. And you might wonder what percentage of the population is that? I think people who are quite poor and especially in poorer countries around the world, they are the biggest gainers. They already have access to free diagnostic, legal, business consulting advice. It hasn't really spread yet. In the last six months, I was twice in rural Africa and Ghana, South Africa. I went around, I asked people, do you have chat GPT? They said, what's chat GPT? I showed them, I downloaded it for them. But people in those positions, not getting good educations, not having access to whatever, they are the biggest gainers. The other biggest gainers, a lot will be immigrants. They're people who are willing to take initiative, did not necessarily go to the best schools, but think creatively about how to use AI. The biggest losers are those masters of the universe who become partners, live in Manhattan, or maybe on the other coast, earn $2 million a year and played the game by the rules. I think in this future, they will be sent to Houston. They'll be an energy company executive earning $350K a year, not a humanitarian tragedy. It may feel that way to those of you who are New Yorkers, but I tell you, Houston has better taquerias and actually somewhat worse weather. But that group, they're the biggest losers. I worry greatly about the politics of that change. On the other hand, you will probably live to 98 or something like that. But when you're 37 and being sent to Houston, again, the living to 98 thing sounds like a fantasy. It will be worth it. I desperately hope our society makes this choice. But again, not everyone will feel good about it. Speed of change. They made this Swedish video for me. It's blurry on purpose. Like how much of our society is really efficient? I mentioned before I work in higher education. Other than Ethan Mollick and a few other people, how well are we adapting to the AI world? We sit around, we bitch about cheating. Oh, the students are cheating. We got to stop them from cheating. UC Berkeley Law School. They just issued this code. You can't even use the AI to ask about a new idea. It's crazy. If AI can help you cheat on a task, take that seriously. But it means you're teaching people a task that humans may not be doing in two or three years. So the cheating is a signal that the whole system is screwed up. And don't blame the cheaters. Blame the system for teaching the wrong things. To get higher ed, tenured professors, to do the adjustment will take a long time. You ask what percent of the American economy is static or stagnant like that? Well, there's government. There's higher education. There's some, but not all parts of our healthcare system. A lot of the nonprofit world. You can go on, make your own list, add up all those sectors. I think it's at least 40% of GDP. So 40%, 50% of our GDP will be very slow to adjust. And that's another reason why I don't think we're ever going to grow 20% in a year. But with some luck and skill, we'll grow another half a percentage point and avoid some kind of fiscal catastrophe. You might wonder, how does all this affect me? I'm talking to you all. I'm not a GPT. My name's not Claude, right? I've reallocated, I'd say, two thirds of my own time. I do much more mentoring. The AI can't do it in the same way that I do. I give many more public talks. I work more with human beings. I travel more. I do less writing. Huge reallocation of my time. I'm old enough. I could just sit on my behind and run out the clock. But I don't want to do that. This is like people talking, trying to inspire others. Charisma will matter more. How you look will matter more. The egghead class, I think, will take a tumble in status. Again, it will be a very weird, strange world. Being in the world and being human and being physical will be more the thing. It will be more what our lives are about. If you're at this talk, it may not be exactly the world you grew up in. I think for the vast majority of Americans, it will actually feel more normal, maybe happier. Oh, there's too much screen time. Just have your AI agent manage it for you and tell you once every two days when there's something in your WhatsApp chat that you actually care about. It'll be great. So it will be a return to a kind of realness and physicality. Geopolitics. This is the geopolitics slide. It shows Sweden. I was happy to see that. An important country you should all visit. They have amazing pizza. And get this, they serve cabbage with it. And the cabbage is wonderful. And then there's lingonberries, great ice cream. So go to Stockholm. But countries that do not have AI are at risk of losing their sovereignty. There might be good enough open source models that they can adapt. But for the European Union in particular, but much of the world, other than America and China, there's a real chance they will be dependent on our system, the Chinese system. The world will be divided increasingly into two camps, US and China. Countries will have to choose. This is not bad for the US, but very dicey. I would just say start thinking about that world. Our last slide. World of leisure. Yes, but here's the other final kicker. Right now, you have to work harder. There's new terms of competition. You have to adjust. You have to learn. Just keeping up with AI. I used to be so excited when a new model would come out. Now I'm like, oh, my calendar can't handle it. Help. Please postpone it a week. So more leisure in the long term, a lot more work and some stress in the short term. And I hope you all follow and watch this world and accompany me moving into it. Thank you all for coming and listening. [00:19:00] Speaker ?: Thank you.