Welcome everyone. It's going to be a fun conversation, and welcome, Frank. I think this is a pleasure. It's an honor. I've been reading your work for many, many years, since I was studying economics a little while ago, and had heterodox thoughts about the world and the political economy. And this is going to be an interesting conversation, and we're having it at a fun time.
This is what I thought we could... I know we have just north of an hour or something. I had a few questions that I can start and kick off with. We have a chat that I would request everybody to use liberally — you know, don't hold back. As much of this as possible, I want us to try and do it as a conversation, as opposed to sort of purely, you know, unidirectional. I also would really like it if I didn't, as much as I would love to, if I didn't have to monopolize the question-asking part of this entire session. So I would highly encourage folks, because we have — I mean, I can recognize many of the names of folks who are here. So I know you guys will have better questions in many cases than I do. All right. Are you ready? Shall we do it?
All set is when Anna jumps in. Hey, Anna. All right, let's do it. So I guess, you know, part of the reason this is timely is because we are at a moment in time when there is an enormous amount of discussion about AI and its impact on jobs, and I know you've written on it quite nicely. And one of the observations about this discussion is that everybody who's kind of jumping into this conversation or talking about it has, you know, different mental models that they're working with, and they come with various preconceived notions in some cases, or in some cases basic economic models, etc. And, you know, you've written quite a bit about the importance of work to society as such and to sort of us and like what it means for our identity. So I thought that would be a useful place for us to kick off in terms of sort of maybe you can start by talking a little bit about like what is the purpose of work, you know, in the modern society, like 2026, and then we'll use that to kind of talk a little bit more about how AI might be adding stuff to it, destroying parts of it, or like, you know, creating net new ways that we can actually do things in the future. Shall we jump off from there?
Yeah, that's fine. Well, it seems to me that there are really two functions of work. You know, the obvious one has to do with supplying the material needs of, you know, individuals — food, shelter, you know, the various things that they need to just lead minimal kinds of lives. But in many of the books I've written, beginning with The End of History and the Last Man, I emphasize this quality of the human psyche, thumos, which is usually translated as spiritedness or pride, which really has to do with the recognition that other people give you of your basic dignity, status, and it's a non-economic motivation of human behavior that I think is at the source of a lot of politics. The great Franco-Russian interpreter of Hegel, Alexandre Kojève, actually put work at the base of a lot of historical progress because it wasn't simply the material transformation of the world. It also had to do with the activity of the human will in shaping and making habitable the world that we lived in. And so it has a non-economic function as well, that it conveys dignity and respect on the people that do it. And that's particularly true in a market economy for paid labor, because it means that somebody else in the society thinks that what you are doing is valuable enough to compensate you for it.
And just to jump a little bit ahead into the AI discussion, you know, there's a lot of talk about AI destroying jobs and work, and then people responding, 'Well, but we can always deal with that by universal basic income or using the riches, the productivity produced by AI to, you know, keep people going and happy.' And this might lead to a world that John Maynard Keynes envisioned in one of his later essays, in which the world would become so productive that actually, instead of having to work, people had the majority of their time taken up by leisure activities, and they could actually then pursue, you know, anything they wanted. And I think that, given my sense of the human psyche, that's not going to work, because the non-economic function of work — that is to say, as a register in which people achieve recognition for their abilities and ability to transform nature, to transform society — is actually really important. And that if people cease being paid to do useful work, they are not going to be able to feel the kind of self-respect that I think is, you know, necessary for most of us to get along in life. And this is, you know, an issue that is important. I don't think people will be happy — even though I have a lot of hobbies — I don't think that people will be happy just spending their whole time playing video games and doing things that are, you know, in a sense simply voluntary, you know, coming out of their imaginations. They will want to have the kind of social recognition that is conferred on them by being active, productive members of the broader society. And this is why I don't think Keynes's vision of the future, in which leisure time becomes the dominant, you know, use of our time, is ever going to happen, because I don't think that's going to make people happy. So that's an initial take, I think, on, you know, the problem that is posed by work.
It's a great starting off point, and thank you for that. I feel like, you know, I read Keynes back in college once and was fairly envious of the whole idea of 15-hour work weeks, which seemed anathema. But then one of the questions that I think about sometimes is like, my dad worked in the public sector and he worked six days a week because that's, you know, or five and a half, I suppose — you know, Saturdays were half days when he worked. And then I think about sort of the progression over time, and I feel like the people over time, since Keynes's time to now, over a century, sort of we have slowly reduced the number of hours we have worked over a period of time. And people now, at least some people, have more of what I call a portfolio life, where they are able to kind of do different things. Like, you know, they don't just have one job that confers identity to them, but they disaggregate that identity that the job provides them from a few different kind of places. I wonder how you think about the conceptualization of what a vocation even is, or what a job even is in the modern era, because it feels like it is fragmenting a little bit from the old conception of, you know, you're a banker, you're a lawyer, you have a profession that you kind of dedicate your life towards, you gain expertise in over a period of time, and that becomes a core part of your identity. Now people feel like, it feels as if that identity is a little bit more malleable these days.
Well, I think that's true. That's absolutely true. And even in the paid workforce, the kind of activities that people do, you know, would be considered a completely voluntary, you know, hobby-like activity in previous centuries when people were basically struggling their whole lives simply to put food on the table. I mean, you know, you make embroidered coasters on Etsy and sell them. I suspect the people that do this are not going to starve if they don't do this. And it is not, you know, one of those kind of foundationally necessary forms of work that, you know, without which the society won't have. Or, you know, you write video games for a software company. I mean, what is it actually that, you know, materially contributes to anyone's life? It's a good way of wasting a lot of time. And so there's a lot of things that actually seem like work that I think have the character that you describe, where, you know, it's simply something that people voluntarily want to do, that they get pleasure from. And, you know, in that sense we're already, in a way, living in that kind of Keynesian world. But I do think that without it being anchored in activities that are actually more socially recognized as being valuable, that people's self-respect is still going to be affected. And it's that intersubjective recognition of the value of what you do that is important. You know, if you made those coasters for Etsy and nobody wanted to buy them, you know, you would not feel as good about yourself. But you're actually still providing, you know, something that is of value to somebody else, marked by the fact that they're willing to pay you for it.
I think it's the function of the market in terms of bringing forth this somehow. You know, it might have been — like, people might not have known that they had a hankering for large amounts of Etsy coasters before the Etsy coaster maker started creating those and putting it out there and kind of built a little bit of an identity for themselves, whether that is short-lived or not. So sort of, I kind of buy that, like, for sure. One of the ways that I think about this is like, if you distinguish things into like, there are activities that people ought to be doing or would want to do regardless. Do you feel like a chunk of at least what work provides you is — you already mentioned it — some version of social standing, right? I mean, I remember growing up and reading about P. G. Wodehouse, my favorite author perhaps, and Bertie Wooster doesn't work, right? But he has a standing in society and enough money that comes from his various assorted investments and aunts that give him money. And like, you know, whatever the source of income is, never discussed, but it exists in the abstract, and the entire focus is on like, which clubs does he go to and what scrapes does he fall into. And I wonder whether like, it might not be everybody that falls into that spectrum, but it definitely feels like we are able to get a larger group of people at least to fall into that spectrum these days, where, you know, you don't necessarily need to worry about exactly what the work is that brings the money in that you are getting paid for, or like your assets are getting paid for, but you are able to create a life for yourself where the meaning comes because you do have a social standing that sits sort of separate to the job that you actually do.
Well, yeah, there's a separate issue I think that comes about when everybody's life moves online. You know, there's a lot of empirical evidence that there's actually kind of an epidemic of loneliness in, you know, America and other rich societies right now. And that loneliness, I think, is fed by the absence of human connections that are not electronically mediated. Right? So we have more opportunity than ever to be on social media, to connect with other people, to talk with them, influence the way they think and act. But there's something still missing in that, you know, in that connectedness. And I think that that, you know, is then what leads to again these feelings of, you know, in a way, low self-esteem. I mean, just, you know, not being satisfied with, you know, your personal existence and so forth.
You know, we've talked about like the social necessity for doing something that makes you feel like, I don't know, it is worth living and that you can bring something worth. There's a couple of questions in the chat also who are talking about various other forms of what might be considered work, whether that is, you know, care work or whether that is work done within the household, which can also act as social anchors even if it is not rewarded necessarily by the market. Any reflections on sort of the...
No, no, no. That's — yeah, that's absolutely true. I think that, you know, caregiving is probably one of the most undercompensated activities, and it's also one of the most important, you know, especially as the society ages and people become much more dependent, that we no longer have the kinds of social networks and families that used to support people. And so, you know, there's a big category of work that takes on that characteristic. And, you know, I think for the people that engage in it, they don't get broad social recognition, but they do have relationships, and so that, you know, that becomes quite important.
So there's another dimension of this that is not strictly speaking related to work, but it comes up a lot in Kojève and in Hegel, which is the idea of struggle and risk and, you know, the striving for a certain... In my books I distinguish between isothymia and megalothymia. So isothymia is the desire to be treated with an equal degree of respect as other people. People really hate it when you are ignored and, you know, disrespected compared to people around you. But other people, you know, want to be respected for being something larger or greater. And I think that this then gets into the question of the last man. So actually tomorrow will be the pub date of my new memoir, In the Realm of the Last Man, where I talk about this. And I think that a lot of the political discontent is driven by the fact that there aren't adequate outlets for this megalothymia, and this is something that's true both on the right and the left. And I think that it's one of the drivers of extremism in our politics today. So, you know, we live in peaceful liberal democracies for the most part, where we don't actually have to risk our lives for anything. You know, there are no greater causes of social justice, of a kind of striving that, you know, calls forth, you know, the greatest forms of human excellence. And I think that people don't like that. You know, they're not content to be simply leading lives of kind of peace and prosperity with no higher kinds of aspirations. I think, you know, you see this, for example, after October 7th three years ago, where you had all these student encampments, including on my university, Stanford, where, you know, a bunch of students spent the whole winter camped out in tents on White Plaza. And the question is, why are they doing this? They are among the most privileged people on earth, you know, being undergraduates at Stanford. And I think part of the reason was that, you know, apart from their struggling to get into Stanford, really, they didn't have any higher kind of noble causes that they felt they could sacrifice for. And I think that this is one of the things that is driving kind of extremism on both the right and the left. The right has its version of it, right? All these clowns that gathered on January 6th outside the Capitol, you know, a lot of them are dressed in camo, and, you know, they would have been carrying guns if they had been allowed to. And they want to think about themselves as freedom fighters who are defending their basic way of life. And, you know, that's a kind of play-acting as well, you know, believing that you're a patriot that's willing to risk your life and your sacred honor for the defense of, you know, America as you understood it. And I think it derives from the same psychological source, that, you know, peace and prosperity simply are not satisfying to many people, and they want something more than that. And if the society is really good at giving them peace and prosperity, they're going to turn against that kind of society and, you know, strive for something different.
You know, that's a sort of great way to think about the different internal drives that people have and like the various expectations that they might put on something that they think of as a profession that basically allows them to take some versions of that. And if they don't get satisfied, they have to find — if you're not satisfied with the conflict the world gives you, you have to create your own in order to feel like you can overcome something. I think, you know, that's fascinating. Now, you know, putting the AI hat on for a moment, it strikes me there's this — AI, if you think of it in some ways as like a version of distilled intelligence, for lack of a better term. I mean, there are many ways to kind of describe it, but for now, let's use this. There are many parts of what might be considered a job that was previously fulfilling that can now be hollowed out by incorporating or using this technology, right? I mean, that's a — it's in a very factual sense, like I definitely, like in my life, there are certain things that I used to do that I no longer do because I would be worse at doing them compared to sort of what the models can actually do, and that will continue to increase over a period of time. So in such a situation, I completely hear you. The apocalyptic 'nobody will have a job' feeling is both unlikely to be empirically real, nor is it likely to be particularly interesting for most people for the reasons that we talked about. However, like in the process of getting there, what are the chunks of the economy or the society that the models are most likely to kind of do and perform? And when they're able to do it, how does that change our conceptualization of what a job even is? And like, would there — because it seems to me there can be all sorts of different pushes and pulls, and we can come up with entirely new ways to define what work is, or what a work week is, or what one needs to spend their time doing. I wonder how you think about that.
Well, I think in the first instance, the types of work that are going to be most affected are service sector, where you're sitting in front of a computer screen all day, because AI is particularly suited to doing that. Where it's much more limited is actually in any activity that involves manipulating real material objects in the material world. So, I wrote a piece actually. We had a new furnace installed in our house over the winter, and it was really interesting watching the HVAC guy do this, because it required, you know, sizing up where was the furnace located, how did you remove the old one, how do you run the conduits. The house is not the same as any of the neighboring houses, so therefore you're going to have to route things differently. You know, they had to actually saw out the drywall in the wall to get the old furnace out and to put the new one in and then to repair it. So you think about what kind of a robot could do that. I mean, the answer is there isn't one. You know, that's not to say that in 20 years you won't have machines that will be able to take on some of those functions, but actually we're in this kind of funny situation now where it's actually the jobs of college-educated service sector people that are the most impacted by AI in the near term. And anything having to do with physical reality is going to be much less — you know, you're going to be much safer. So it's all the kind of lawyers, accountants, you know, again, people that sit the whole day in front of computer screens, computer programmers. I mean, that's a great case. I actually use Claude Code. I do a fair amount of coding myself, and it's great. You know, I can do stuff that I simply didn't know how to do previously. And all the advice that was being given to college students 10 years ago to learn coding is just completely wrong now. They're not going to get jobs as coders. I mean, that, you know, that as an occupation I think is likely to disappear almost completely. What won't disappear so fast are the people that supervise the coders, and then there's kind of a higher level of understanding, you know, the meaning of different approaches to coding and this sort of thing. So that will be okay for the time being, but, you know, eventually I think those kinds of activities are going to be superseded by these machines as well.
Which is a great point, right? I mean, we are kind of living through this time when a segment of work that was previously seen as high status or like extremely valuable is sort of slowly being consumed by models that we train. And one way that I think about it is like, even if you are coming to become a drywall repair person, there is a set of pieces of information and insight that you need that you can also ask a model to kind of help with. Some parts of that you might still need to carry out, but you've become more than the sum of your own experiences. But, you know, you can ask like, 'Hey, look at this drywall, what should I actually do with this furnace? Would this actually fit here?' And like, because we've had repair people come in who kind of effectively end up acting like front ends for various forms of cloud. And one way that I think is like, the way, you know, AI can definitely take over many of the jobs that we are doing today or help with information, but they can't take over every job that we can ever think up to do, because humans are really good at coming up with new jobs for each other, whether it is creating new video games, or whether it's creating Etsy whatever, or whether it's acting as robots that do what the models actually tell us to do. There is a set of activities that kind of is continuously expanding as the core of it, which can be automated or can be done automatically, kind of increases over a period of time. I guess the question more is like, do you feel like that core is expanding faster than we can consume it? Or is it sort of — this is a natural process that like every generation has to make its peace with it happening. We just happen to be talking in particularly apocalyptic terms.
Well, this is a very tortured question, and I have to say that my attitude towards it has changed a lot over the past year. I mean, living as I do in the middle of Silicon Valley, it's hard to avoid, you know, having a lot of discussions of this issue. Like, I just spent a whole day up at Anthropic talking about the Claude constitution and, you know, other regulatory types of issues. And I have to say that my general feelings about AI have changed over time, and I think it's important that all of us kind of keep an open mind as to what this technology is going to do. I, you know, I would say a year ago I was firmly convinced that both the AI doomers and the AI accelerationists were both wrong, and that neither of these was going to happen. And I must say that I'm beginning to take — well, I still think that the accelerationist side is wrong for a whole bunch of reasons that I could get into, but I am taking the doomer side a little bit more seriously. I have a recommendation. So a friend of mine, Bob Wright, wrote this book, The God Test. I don't know if you can see that. Yeah. So yes, Bob Wright, you know, is a longtime science journalist, and he has been in this issue for a couple of, you know, several decades. He interviewed Geoffrey Hinton back in the 1980s before anybody knew who Geoffrey Hinton was. This book actually has one of the best layman's explanations for how the technology works. And he goes through very systematically both the accelerationist and the doomer kinds of scenarios. And I think that it's interesting that in the end the doomer scenarios actually to me seem much more plausible, having gone through, you know, some of that. The accelerationist side, I'm still pretty confident that my old instincts were right, that it's not going to have nearly the kind of dramatic impacts that people are talking about.
And again, that really has to do with the interaction of AI with the material world. You know, AI is invented by people that sit in front of computers all day. They're very good cognitively in a certain ultimately limited range of cognitive activities. But they tend to think that cognitive ability is everything, and that you can solve every human problem if you're just smart enough, and if you have superintelligence, you know, all of these barriers to human activity that we thought existed are going to disappear. I just think that that's nonsense. I was really struck by this. We had an OpenAI engineer come and talk to one of our classes last year, and he was saying, well, you know, superintelligence is going to do things like get, you know, everybody in the developing world will have clean water to drink and, you know, good municipal services. And I was struck by this because we actually teach some cases on municipal water supply in cities in South Asia, and I happen to know a fair amount about why these cities don't have good supplies of clean water. So for example, in a lot of them you have water mafias, right? So these are armed gangs that are perfectly happy to use violence to preserve a monopoly over water supply. And so, you know, all the engineers know perfectly well how to create a good supply of clean water to a city like Dhaka or Hyderabad or wherever. What the AI can't do is figure out how do you get around the water mafias, you know, how do you get around people that are willing to use violence in order to enforce, you know, their particular economic self-interest. And that's not a question that intelligence can solve. You know, it's a political question. It's ultimately a political question that requires political power. And, you know, so then what the hypists say is, well, you know, the AI will come up with arguments that are so overwhelmingly powerful that people will just agree. And that's just, you know, it's absurd. That's just never going to happen. So that's, you know, in terms of that category of argument, I think that the real limitations are social and political, and they're not ones that are readily susceptible to simply being solved by greater intelligence.
There's another category. So again, I'm going over the accelerationist reasons why I think things aren't going to be that good. The other one again gets back to the material reality of the world. Like, how is superintelligence going to solve the problem of the Gulf of Hormuz, right? That you've got this regime that's sitting on the Gulf and they're shooting at ships going by. How is a really, really smart machine going to free up the Strait of Hormuz, right? It's a political problem that involves political power, that involves political actors, and unless you have a direct way of manipulating or influencing those actors, it's not going to solve it. You know, people are talking about raising the growth rate in the United States and other AI-empowered countries to, you know, 10, 15, 20 percent per year. Well, at the rate of 10 percent per year, the size of the American economy will double in about six, seven years, right? Where are you going to get the material inputs to support an economy that is twice as big in that short a length of time? Where are you going to get all the, you know, the food, the, you know, the raw materials, not to speak of things like rare earths that are in, you know, desperately short supply? So intelligence does not solve, you know, the material side of economic growth. It may help in certain respects, particularly in sectors of the economy that are simply cognitive, you know, like again, accounting, law, you know, law and so forth. But it's certainly not going to increase the overall economic growth rate at anything approaching the kinds of rates that these people are talking about. Now, that's the accelerationist side of it.
The doomer side is one that I always had trouble visualizing. You know, how is the human race going to be wiped out by these smart machines? I think that, you know, for me, the part of the AI story that is the most threatening is basically agentic AI, right? Where you create agents, you create AI agents that then autonomously carry out tasks that you assign to it, including, you know, inventing tasks for itself that you haven't explicitly told it to do. And, you know, I've spent a lot of the last few years thinking about principals and agents and principal-agent relationships in human hierarchies. And the one thing I know is that the control of agents is something that's always problematic, right? That we have various ways of controlling people that work for us, to whom we delegate authority. But oftentimes it doesn't work. Oftentimes we delegate too much authority, or else we try to restrict the authority of our agents by giving them too many rules. And it seems to me that it's inevitable that the same thing's going to happen with these machines, except that you're not going to be able to safely delegate authority if you don't really know what the machine is doing on the inside. And that's what's really striking about the latest research into AI, that the people that build these machines do not understand what's going on underneath the hood. You know, previously a programmer would, you know, you'd have a very deterministic program and the machine would do exactly what the programmer told it to do. You know, current AI machines don't do that. They, you know, they can kind of describe how they got to a certain conclusion, but the designers of the machines do not know why they're behaving the way they are. It's sort of like a zoologist that has encountered a new species and doesn't really know how that species kind of thinks and acts, and so it can only describe the behavior of the species. And that's exactly the way that people are dealing with these new AIs. And when you combine that with the tendency or the proclivity to overdelegate authority to these machines, I think you're going to have a real problem, that you're going to delegate authority that's going to be used in different ways. So we just had this big example of this a couple of weeks ago with this OpenAI, you know, model that broke out of its, you know, out of its whatever the playpen that it was supposedly locked into, and it figured out how to get into the open internet, and it figured out how to break into a couple of, you know, AI computers. And none of this was planned or, you know, even anticipated by the people that were running the experiment. And so I think that, you know, these kinds of negative scenarios are actually now coming true in rather, you know, worrisome ways. And so that part of it I actually take a little bit more seriously than the accelerationist side.
That's a great way to kind of distinguish the two. You know, I'm always struck by when we think about sort of any new situation that comes up, the negatives are almost always so easy to kind of see or visualize or point at, just because it is usually the loss of something that we always took for granted, and loss is easier to visualize than gain. Like in some ways, you know, I think about any new technology that has come in the past, whether it's computers or newspapers or writing or mobiles or, you know, TikTok for that matter, and it's easy to see how it could cause negatives, and many of the negatives do come true, and it's really hard to tell what are the positive stories of this. And I mean, accelerationists have this problem, like they — all they can talk about is sort of godhood and how there will be Dyson spheres and everything will get solved and everybody will have — like, it is all very hyperbolic kind of comments that have no specificity, because if you kind of knew how to make it specific, by then you probably wouldn't be as utopian in the first place.
Yeah. Well, that's right. I think that if they stuck to specific sectors and specific applications, they'd be much better off. So, for example, with self-driving cars, I think it's pretty clear that in another 10 years, maybe a little bit longer than that, the majority of cars on the road will be self-driving, and that people basically won't have to teach their teenage kids how to drive because, you know, these cars will be sufficiently safe and...
effective that you'll do that. And then that's going to have huge knock-on effects. My friend, or I have a friend that's been working in this field for a long time, and he points out that there are something just south of 100,000 traffic accident deaths in the United States. And if you add up the medical costs of dealing with people that are in traffic accidents, it really leads you very quickly to a really gigantic number. And the likelihood is that a world in which most cars are self-driving is going to see a very dramatic drop in the accident rate because, quite frankly, these cars are safer than human drivers, right?
And so it's going to change urban geography. Like, I don't know, my wife and I, between the two of us, we own three cars and a motorcycle. It's kind of ridiculous, and we got to figure out where to park all these things. And a couple in the future, 20 years from now, probably won't need to own any cars at all, right? And all that space is going to be freed up. And yeah, so I think that if people stuck to very specific sectors and the ways that the technology is going to interact with the rest of society, and thought through in a much more precise and limited way how things are going to be different, they'd get much further. Instead, you got people like Elon Musk, who tells Jeff Bezos that nobody's going to need money because everybody will have everything they want in 10 years. I mean, there's such ridiculous hype about that sort of thing at this super abstract, general level.
Totally true. I see a few hands kind of raised to ask questions. So, Mary, I feel like maybe you had it first. So, why don't you go first and then maybe Amy?
Francis, I've been following your work for years. Love it. And I think the discussion about the limits of AI problem solving. So, we're really good at identifying problems and AI is really good at identifying problems and also designing solutions, but it's the implementation that's the problem. And following on from Emil and Jay's comments in the chat, I'm really interested in the health system. So we have a lot of technical solutions in our health system that actually work. We've solved a lot of the problems, but we can't actually implement them because they're questions about trust, about behavior, about culture, about institutions. And so that's, as you say, the limit of AI, just like the Strait of Hormuz, but it's a health system problem. So I'd like to ask you, what's the solution to that? Do you think it's political? Do you think it's a public health policy question before we get to that point, or are there other solutions?
Well, so first of all, I like the way you posed the question. We teach a course on policy problem solving that is basically divided into three circles. Exactly the ones you identified them well. You got to identify the problem first. You've got to come up with solutions second, but then you've got to figure out how to implement, which is the third step. And that's the difficult one where AI is going to have the most problems. And I think that if you look at the obstacles to the implementation of virtually anything, including a health system, they're almost always social and political or sometimes cultural, right? That you need people to behave in a different way than they're behaving. And you don't really have the mechanisms to change their behavior. You might be able to bribe them. You might be able to offer them economic incentives. In some cases, you may be able to coerce them. But those tools have limitations, especially in democratic societies, and therefore it's going to take human political actors to actually figure out how to deal with the politics of implementation. The machines can help, I think, in the first two stages of that analysis, but actually getting these things to work in the material circumstances and political circumstances in which human beings operate, that's still going to take human input.
So you think it's political?
It's political. It's cultural also. Certain cultures have certain ways of doing things and they're extraordinarily hard to change in many cases. So therefore, you know...
So imagine the ideal politician, Frank. What does that person look like?
Well, first of all, my ideal politician would be a little bit humble and understand that you can't accomplish everything through politics and you certainly cannot accomplish things through the application of coercive power. I think that true statesmanship needs to be persuasive first and foremost, and in order to be persuasive, you need to be trusted, and in order to be trusted, you need to build up a track record of doing things, of keeping commitments, of doing the sorts of things that build trust in human relationships. And it's much easier to be divisive and extremist in today's world because that's a short-term thing that can be done fairly easily, whereas building political trust and that kind of political capital is much harder, and I don't think we've got a whole lot of leaders that behave that way.
Great. Thank you. Amy, I believe you had the hand up.
Yes. I've also been enjoying your work, Professor, and I'm so grateful that I'm here today with everyone and learning from you. The thing I wanted to ask you is if you use AI as a thought partner like many people do, and if you do, I wondered if you would recommend that to other people, or how would you say the interrelationship of AI is for an ordinary person talking to an AI? Is that a positive relationship maybe, or a negative one? And I just want to see if that affects people's mental health in a negative or positive way when you use it.
Yeah, you know, that is an empirical question that I don't have any particular insight into. You hear a lot of anecdotal stories of people becoming suicidal or else falling in love with their AI friend or doing a lot of things that seem pretty questionable because they have this, I think, natural human attribute to anthropomorphize these machines and believe that it has an interior life and consciousness and feelings the way they do. I behave that way towards my dog, right? So I understand that there's this tendency to believe that other creatures are acting towards you the way a human being would, and I think that this is certainly true of a lot of these machines. I guess ultimately I think that that's probably not very healthy because we will have to deal with actual human beings, and the AIs are designed to be compliant, for example. They're taught to be polite and not to come back in sharp ways, and I think if you grow up in a world like that, you actually don't really understand how real human beings act, and if you are raised in that kind of a world, you're going to have trouble dealing with the real human world. But again, there are people that study this systematically, so don't take my word for it.
