The future of Human-AI coexistence, according to Kevin Kelly (co-founder of Wired, futurist, author)

0:00Kevin, you just formed this compelling theory about our future. the economic handoff to bots. It all starts from human population decline. Every year you have a smaller audience, a smaller marketplace, a smaller workforce. That's got to be a different kind of capitalism than we've had. I think it's kind of no coincidence at the very moment when the number of species of humans on the planet plunges that we actually are creating artificial species and AI. We're going to want to have an autonomous economy populated by millions, billions, trillions of bots. We will soon be outnumbered. So there might be a way in which the AIs are actually consumers as well as producers.

0:43You make a robot, it needs a room. So we're expanding the possibility space to include other kinds of entities. And then the question is, what will motivate them? What are the incentive signals? It's not hard to program in values and judgment and stuff because it's almost like software, it's like code. The difficulty is for us to come to some consensus about what it is that we want to give them. We're trying to make them better than us. A lot of our experiences are about struggle. What does a bot-to-bot economy feel like for us? We don't struggle about surviving and starving. We have other kinds of struggles about who we are and what we're here for.

1:20And I think AIs can help us with that because they're going to answer, what are we here for? Why are we doing this? And help us elevate to a greater purpose. Hey everyone, it's Aziz. You've just heard highlights from my discussion with Kevin Kelly. Kevin is a co-founder of Wired Magazine and a renowned author and futurist. Decades ago, Kevin predicted much of today's technological and cultural landscape. Now, he presents his new bold vision for what's coming next. We originally recorded this for Friday with Azeem Azhar. It's a new live video format we're running every Friday at 9am Pacific Time.

1:56If you're on the East Coast, it's at noon and 5pm if you're in the UK. I'd love your feedback, so feel free to comment or email my team at live at exponentialview.co. We read every single response. Thanks for being here, and now enjoy this deeply thought-provoking glimpse into the coming decades. You point out that there are going to be fewer and fewer humans alive to contribute labour and to consume what is being made. What do you think is that fundamental driver of this baby black hole? You know, we don't really know. It's very, very correlated with modernity, with high education, increased living standards, increased higher child survival and birth control.

2:45there's a thousand different factors, which has reduced the desire for having a lot of children. And the issue is, is that whatever the multiple causes of it are, we haven't seen anywhere in the world that's been able to reverse it once the fertility level lowers beyond replacement level. Yeah. governments are trying different things. And so we don't really know. There are a lot of theories about it. And there are some natalist movements that are changing that. And they have different dimensions. In fact, just the other day, there was something weird happening in China, which now China apparently now has the three-child policy versus the one-child policy.

3:42and they were trying to somewhere a company was trying to mandate children. So I think we're at the beginning of that. And I don't really have a theory about why. My analysis is only saying that I think this is where we're headed to and it would change the economy. And that's my argument is that for all of the recent human history and progress has always been associated with increasing population. We have no experience as a species of increasing living standards and decreasing population. I think it is a common pattern of modernity because we see it happening across cultures in rich and developed countries, and you see it within strata of emerging economies as they get richer and richer.

4:37One of the things that surprised me about the population question was that it's such an important one. I mean, population is literally the temperature of humanity. And it hasn't really been something that I think has been part of the mainstream debate for a while. In the last year or so, places like the Financial Times have started to talk about this a little bit. One of their journalists, John Byrne Murdoch, had done some good work on this. Stabsacker called Alif Evans, who has asked the question, you know, why are we seeing this happening? And she breaks it down a little bit. She says that, you know, it's economically and socially easier to be a single now than it ever has been.

5:21You know, you get services from DoorDash rather than your partner, and you can have sort of economic security through your work. It's expensive to have children, especially as countries get wealthier and wealthier, the relative cost of having a child goes up. And there are other forms of entertainment now. There's Netflix and there's online gaming and swiping on different pieces. And these seem to me to be quite fundamental. And you rightly point out that pronatal policies have not succeeded anywhere at all. But I'm quite curious about why it's taken some time for the mainstream to wake up to this.

6:05You know, the United Nations still has this view that 2086 is the peak for human populations. My view is aligned much closer to yours that it's more like in the next three decades. But where do you think that mismatch comes from? Well, partly it's because historically, when birth rates have lowered, they've bounced back. And so there was some sense that that was a natural oscillation. So the idea that we're kind of on a one-way plunge is very radical. And then secondly is I think our own estimates of the population in the world by country is also plus or minus at least 10 % or more just because there's a lot of political incentive for having those numbers, particularly in developing countries.

6:54And then I think also the pace at which it's plunging right now was sort of so unexpected that there's a resistance to imagining that it keeps going. And people, you know, when I talk about it, people say, well, if that can go to zero, we don't know that. Again, we have done nothing to reverse it, other than we do know that in religious communities like the Mormons, the Amish, Hasidic, they do have high birth rates, And that's one possible scenario is that the Amish take over. Right. You know, that's right. Right. So, but, but, but to do that, that's not going to happen very quickly because there are so few of them that it would take many, many, many generations for them to become a majority of the population.

7:39So that's not going to happen by the end of the century either. And so, um, again, I'm agnostic about what actually is causing it. I'm looking very closely at the trials, and I suspect that we may come to a very high number. So it may be a matter of money in this way, that there might be countries that do a calculation that says, well, the value of an additional citizen to our country over their lifetime is, say, worth$8 million. And so therefore, maybe it's worth a million dollars to buy one, to get an additional one. And so then you can imagine that families could be given a million dollars over the span of, say, 21 years to have children.

8:25And so if you have more than one, you might at that point actually have families that are their business, their career is actually having a family, which would be kind of a weird trad thing. But also economically, it might be feasible. And so that might make a difference where you actually are basically subsidizing families through the government because you want to have those additional people in your economy. But we don't know that. We haven't tried that. South Korea offered$75 ,000. It was not enough. I think that number has to be much higher to even make any dent. But even that, we don't know if it is.

9:03But I'm assuming for the moment that we don't change that. So I'm not I don't have any policy about advocating what would change it. I'm saying let's assume that that's not going to happen. And what would that have an impact on the economy? Well, let's get into that point, right? I mean, I think it is an incredibly seismic idea that human population doesn't rise. Essentially, it's been rising consistently since there were a handful of us. There have been a few moments every century where cooling or a plague has reduced population. And, you know, I suppose most recently around the time of the Black Death across Europe, but they've bounced back globally.

9:47globally. And what we're talking about now is really, for the first time, a secular decline in our population. And a few months ago, I wrote an essay called Why Humanity Needs AI. And it was along the lines of your handoff to bots. I was saying, look, to be human, certainly to be a modern human has been to live in a society that gets more complex with more people in it, more people producing things, consuming things, exchanging ideas, building complex social structures, and also undertaking scientific discovery, which has always been functionally related to the number of people we have and the number we can afford to allow to be scientists.

10:29And that at some point in the next 30 or 40 years, we're going to see a peak of population, And that will be almost a species level change, transformation that will require us to make use of AI for different reasons. And you've got there and you've made a very strong kind of economic argument. Just reprise what that argument is. So the idea is, again, if you map out the curve of living standards and population, they're very highly correlated because we've had growth. And that growth has come mostly from increased population. and if we had declining population every year, which we'll get to in a minute, that means that every year you have a smaller audience, a smaller marketplace, a smaller workforce, and that's got to be a different kind of capitalism than we've had.

11:26Right. So we'd have to kind of shift a little bit away from our ideas of what growth entailed and having growth more like maturity rather than gaining weight, you know, just getting bigger, you're going to actually drink, like not drink more wine, but drink better wine. I like that idea. Yeah. Right. So that's one thing. But I think it's kind of no coincidence in the cosmic view at the very moment when the number of species of humans on the planet plunges that we actually are creating artificial species and AI. And so there might be a way in which the new economy is based around the fact that the AIs are actually consumers as well as producers, that they are in some ways an integral part of the economy and the way humans are, that not only do we generate stuff, but we actually consume it.

12:17And that's a large part of this whole big engine where people make things and other people use those and buy those. And so it may be that the handoff is to AI agents, bots, robots, AIs, that are in some ways doing what a lot of the humans were doing, meaning that they were consuming a lot of the stuff that we produced. And again, you make a robot, it needs a room. You have a car, it needs a garage. It has a garage, it has a car hospital. You have an AI robot hospital. There may be ways in which we're not imagining that they're handing off the economy to billions of agents and robots instead of billions of humans.

13:03And they take more part in this new kind of economy that we're going to construct. And that's the thesis. Yeah, I mean, it's a fascinating idea. The idea that essentially we continue to have sufficient number of economic entities in the economy to keep things turning over. I mean, in a way, we have these artificial entities in the economy already. We call them corporations. And we're talking about many, many more of them in order to maintain enough vibrancy. And there's a great line from a legal scholar, markets and capitalism is the comedy of the commons because of the intersections and interactions and the emergent properties of that mixing.

13:47And there is this notion, perhaps, that that's the vibrancy that you need. One of the questions that I have is, what does that path to that future look like? You once described humans as the reproductive organs of technology. But where do the market forces fit in with that mental model? Yeah, so again, I think we're in the process of kind of revising the standard model of capitalism, which was based on growth. And it's clear that we have to have a different kind of growth. And again, I call it more like maturing, where you have an evolution, not in terms of making it bigger, but you have evolution in terms of making things better.

14:37And so that's a big step that would take some adjustment in terms of our models, taxation, regulation, all these other things. including, again, all the issues we have with AI is who's responsible for them. If they do something that is harmful, who's accountable? And so those are kind of the basic infrastructure changes that I think we'll make as basically our population declines. And the aha moment for me was to realize that this moment is actually much closer than we think. the official UN figures was that it was going to happen, you know, the peak was going to happen around, you know, 2070 or something.

15:24When I look at the figures and also the rate at which the UN keeps changing, it looks like it's going to happen in 2040. Right. So in 2040, which is 15 years from now, the total number of deaths on the planet will outweigh or outnumber the number of births on the entire planet. So this is fairly soon. That's only 15 years away when we're actually going to reach this peak. So I think this issue of thinking about what that looks like, where we're headed into in 15 years, within the lifetime of the people who are listening, no matter your age, we need to think about what does that look like? And I don't have a policy.

16:07I don't have a program. I'm just saying, yeah, this is where we're headed. Well, we can play about with that in this conversation for a few minutes. One of the things I think that is also interesting is that even if population doesn't decline until 2060 or 2070 or 2080 as a UN, you used to think it would. The thesis is actually is still true because we can manufacture AI systems extremely cheaply. cheaply. The first version of GPT-4 cost a lot, and then you get the Nth version in DeepSeek V3, and it's much, much cheaper. And we will soon be outnumbered. We all choose to be supported by many more synths, AIs, whatever we call them, than there are people.

16:59I mean, if you just look at how you might work today or how I work today, I have five or six LLMs. I've got a whole bunch of other tools that keep me company and help me with my work and through my day. So we're already heading in a sense in that direction. And I suppose the thing to play around with is why is that beneficial? I think it's beneficial because the world gets more and more complex. And I actually, the human, really like the fact that when I turn the tap on the faucet in my kitchen, I know clean drinking water will come out of it. Behind that is a whole load of really complicated stuff that involves machinery and pipes and hygiene systems.

17:41I mean, I don't even know the words, but I can actually just trust that to work. And as our world gets more complicated, we actually, I think, do want to abstract away from that complexity. So even independent, frankly, of population decline, we're going to want to have an autonomous economy populated by millions, billions, trillions of bots. And then the question is, what will motivate them? What are the incentive signals that they get? And of course, how do we govern that sort of sea of entities for our own benefit? And I think that that is a really interesting infrastructure question and it's a kind of a governance question.

18:23What does a bot-to-bot economy need to look like? Yeah, exactly. I did a calculation a couple of years ago about the amount of energy used on the planet. And I think it was three quarters of the energy that we were now producing or consuming was being used not for the benefits directly of us, not to move us, not to heat us. It was being already devoted to the technium. It was moving other physical things. It was heating other physical things. So in a certain sense, already a large portion of our economy is machine to machine to begin with. And we're just kind of making it even further where your agents will interact with other agents.

19:09They will consume the information generated by those agents, and then they will send it to other agents. And so there could be an economy built on that all behind the scenes. As you said, it's kind of infrastructural-like. And by the way, you know, 95 % of the AI that's going to be ever generated, you will never see. You will never be aware of it. It's going to be in the back offices and it'll be this kind of agent to agent contact. And so the question is sort of, you know, everybody asks us, what's the role of humans? I'm very, very optimistic about this. And I think the reason is, is because what I think technology gives us, you mentioned complexity.

19:47I have another term. I use the word options, possibilities, and opportunities. That's what we get with technology. It's not just complexity. It's the fact that we have more options and more choices about what we want to do in life. And by the way, if we want to be an Amish farmer, farming the old way, that's still a possibility. And that is there. If you want to be a mathematician, if you want to be a ballerina, if you want to be a mortgage broker or a web designer, you now have that possibility. And we have even new ones in the future. And so that's what technology gives us is expanding possibilities and options that you don't have growing up in a little village with organic food and strong families.

20:34You have only one choice when you're going to be a farmer. Now we have, you come to the city, which is a possibility factory, and you come to modernity, which is increasing the options. And that's what these agents are going to do, is they're going to give us even more possibilities of things that we can do, ways to be creative, how we can spend our time. If you still choose to farm organically by hand and go to the opera, those are still choices. And so what we're doing is we're making a new economy that's going to have even more possibilities and choices. And a lot of those possibilities are for the AIs, not just for us.

21:16So we're expanding the possibility space to include other kinds of entities. I would definitely want to touch on that and explore that idea because there's a lot that you've raised around choices and options. But before we get there, I would love your, you've been part of the, I think, the sort of deep, deep creation and narration of Silicon Valley culture for 40 plus years. This technology right now is largely being built by a certain class of companies. Of course, there is the sort of open source movement alongside. How comfortable are you with the ways in which those choices that will define this world that you see coming in 10 to 15 years will be made in a way that you would feel comfortable with?

22:09And I'm thinking, you know, OpenAI and Meta and Microsoft and the rest. First of all, one of the principles that happens in this kind of technology is that there is a tendency to win or take all. There is a tendency of network effects to kick in because of the communication technology. And we see these monopolies happen. And that makes a lot of people nervous. And a lot of the concern about the AI where we're going is that there are a few companies that seem to be running big. And there are several things to say about that. One is that I think these are natural monopolies, but they're always very, very temporary.

22:49You'll notice that despite the strong efforts of the incumbents, Amazon, Google, Meta, they were not the ones to dominate the conversation with what's happening. It's startups. Because that is the general trend. You know, you had IBM. Who could compete against IBM with making computers? Nobody could. Well, it was a software company, Microsoft, that came along that displaced them. And then who could compete against Microsoft? Nobody. Then it was a search engine. It was always outside of the current regime. And so these monopolies are very, very temporary. And I think they're good while they're there, and then they're going to be displaced by something else.

23:32And so right now, it's the startups. It's DeepSeek. It's all the outsiders where the real disruption comes from. So I'm not that concerned about the fact that there's this dynamic of the increasing returns. That's just part of the landscape. The second question is, what are the values? What's being embedded in it? And I find that the people who are working on this right now, which are my friends and my family, very, very consciously trying to think about what it is that they're doing. What is the values? How do we align it with humans? I find them really trying very hard, but it's a very difficult prospect.

24:18The thing about the values is that we can put them in. That's not hard. It's not hard to program in values and judgment and stuff because it's almost like software. It's like code. The difficulty is for us to come to some consensus about what it is that we want to give them. And so that's the difficult part. It's almost very philosophical. It's like, yes, we can put values into these machines, but our own morality system is very shallow and very inconsistent. We're horrible with this. So here's the thing. We're trying to make them better than us. And that is really challenging and beautiful. We are actually trying to make the morality of the AI self-driving cars better than we drive, better than us.

25:06And so that making it better is a beautiful moment, I think. Well, I think we've got a rare opportunity here where we have two optimists on a call like this. So let's have this kind of virtuous circle that's going to build out. So I have a thesis that I have, maybe it's intuitions rather than a thesis, which is that the LLMs of today are trained on everything that's out there. and some of it they've got in ways that might make a copyright lawyer uncomfortable. And yet they all behave reasonably similarly despite their fine-tuning. They are pretty enlightenment, J.S. Mill, a small liberal. I don't mean in an American political sense.

25:50I mean in a political theory sense. Systems that embed a sense of fairness and kind of morality, and you see it eat across grot. You see it across running DeepSeek if you run it locally. You see it in Claude and OpenAI. They all express it in slightly different ways. But what they've got in them is a sense of the rationality that started to emerge after the scientific revolution of the 18th and 19th and 20th century. It's really hard to get your LLM to persuade you or persuade itself that the Earth is flat. it's really easy to have DeepSeek run locally and talk about things that happened in Channeman Square in June 1989 and to evaluate them from some kind of ethical or moral standpoint.

26:41So the optimist in me has this intuition that of course they're trained on how to make bombs and so on. But in reality, for them to be useful, they have to reflect some underlying rational truths about the world. And embedded in there are some moral truths, perhaps. Yeah, exactly. And by the way, you know, you said the right thing. They're trained on all the documents, let's just say, to the first approximation, including the best and the worst of the writing and the ideas. But in general, the way that their system works is they tend to put out to sort of the average, like the average human view, the average human consensus, the average human morality.

27:19However, in their latent space, they are capable of actually producing the best. But you have to nudge them in that direction. That's one of the skills of working with these LMs is that you're constantly trying to nudge them into the latent space of the professional, of the highest level, of the highest caliber, thinking of the highest morality, which they don't. They tend to want to come to the average. And so part of the skills of working with these agents, these aliens, is nudging them to their best. And so that's basically the prompting engineer's task and the skill that we're all going to learn is how do we get these to be on their best behavior, to give you the best answer, to take that extra time to do not just the average.

28:07They're kind of lazy in that way. You want to go, no, no, I know that you can do better. Behave the best you possibly can. All right, I'll do that. Yeah, well, I mean, confession, that's how I use them as well. You know, if I've got a difficult email to write, it's normally about, you know, getting my parking ticket overturned. I really nudge them to say, like, write the most persuasive, humbling, but effective email you possibly can. And you kind of run up with this idea, which I really love. You build off some of the ideas from Robin Hanson in his book, The Age of M. This notion that you'll have these kind of more empathetic systems that are out there that will be working for us.

28:48and this idea that, you know what, there's going to be lots of content out there. There are going to be lots of things to do. And our sea of agents can be consuming all of it for us, figuring out what we need better than we might otherwise need. And then I guess the challenge that that raises is that a lot of our experiences are about struggle. A lot of the things that grow and develop us, the things that we remember are about challenges, being able to make choices, recognizing that things don't always have to be perfect. And one could argue that some aspect of the Western condition and any sense of the malaise perhaps may come from an excess of comfort, right, culturally and economically than we've had.

29:35So the appeal of being sort of super served by a network of thought also sits counter to something that has also been inherently human, which is we have to struggle and we often do struggle and struggle forms a part of the creation of our moral virtues. So how do you bring those together? So I don't believe in utopias. I think they're a bad idea, even if you can make one. I believe in protopia, which is this incremental move to slight betterment, even if it's only 1 % improvement over time. And that you accumulate goodness over time. And that's what civilization is, If we can create 1 % more than we destroy each year, we can accumulate a civilization.