Artificial intelligence is the topic on everyone’s lips. For our series of 2026 Nobel Prize Conversations, we have created a bonus episode focusing on AI collecting thoughts and perspectives from a mix of Nobel Prize laureates. The transcript for the episode can be read below.

Peter Howitt: Is AI going to enhance productivity or is it going to replace jobs? In a sense, it’s going to do both.

Omar Yaghi: If we want to transform science, scientists need to experiment with it.

Daron Acemoglu: There isn’t a universal law that good technologies will give us good outcomes.

Adam Smith: So this episode of Nobel Prize Conversations is about AI. And there you heard the voices of two 2025 laureates. First, Peter Howitt in economics, and then Omar Yaghi from chemistry, followed by the voice of Daron Acemoglu, who was one of the 2024 economics laureates.

Karin Svensson: What do these three people sort of exemplify in terms of the things that we’re talking and thinking about with AI?

Smith: Those short statements kind of encapsulate quite a lot of current thinking, don’t they? Peter Howitt expressing the deep uncertainty about which way things are gonna go. Omar Yaghi reminding us that the key is to experiment and to really have an evidence-based view of what all this will mean. And Darin Acemoglu reminding us that technology isn’t necessarily going to deliver good.

Svensson: Yeah, it’s both very encouraging and very scary, this area of though, isn’t it?

Smith: It’s exciting. It’s confusing. It’s hopeful and optimistic as well as being pessimistic and worrying. It’s everything.

Svensson: Well, great. An episode about everything. I look forward to that.

Smith: And of course, there’s the doomsday scenario. As expressed rather well by Geoffrey Hinton, another of the 2024 laureates.

Geoffrey Hinton: There’s very few cases of more intelligent things being controlled by less intelligent things. Once they’re a lot smarter than us, I don’t think they’ll put up with that.

Svensson: So we’ll focus on this year’s laureates, but since 2024 was such a significant year for AI-related prizes, we’ll throw in a few of those as well. This is Nobel Prize Conversations, and my name is Karin Svensson.

Smith: And mine is Adam Smith.

Svensson: And this podcast was produced in cooperation with Fundacion Ramon Areces. How often do you hear the phrase artificial intelligence these days?

Smith: I suppose a trite answer would be to say not often because everyone calls it AI. It comes up everywhere all the time. Sometimes I think to people’s dismay, we were doing a panel the other day in front of a group of mainly academics, and the question of AI’s impact was about to be discussed. And I asked a question of some panelists about the whole chestnut of whether AI will, to some extent, replace human creativity. I just heard this deep sigh from somebody in the front of the audience. I sort of sympathised with them. Yes, oh, here we go again. So it’s talked about a lot. Maybe it’s not talked about in the right ways.

Svensson: AI is transforming research in all fields. Let’s hear some examples of how this affects laureates in their respective fields now. First up is 2025 medicine laureates, Mary Brunkow.

Mary Brunkow: If you can measure every health-related aspect to a person’s person, multiomic profiling of all different kinds of molecules combined with imaging and medical health records and all that sort of thing, you need AI in order to make any sense of it. In order to be able to incorporate those kind of disparate data types and find meaningful patterns and trends over time, you have to have AI for that. So there’s a very strong awareness and excitement here for figuring out best ways of kind of harnessing the power of AI and making sure that it’s being used appropriately. You know, not losing control of it, but always knowing exactly how it’s working and how it’s being trained and that sort of thing.

Svensson: Do you have any other examples of sort of specialised or maybe unexpected uses for AI that you come across?

Smith: Well, so many. Everywhere you look. I was just talking to a film director in Korea doing amazing things. But from medical diagnosis through to trying to predict humanitarian disasters, I know a group in São Paulo who are using it to try and predict the conditions for genocide, everywhere you look. But it all depends on the quality of the data. Boring to say. But that is so important and in most cases, so difficult to obtain robust data.

Svensson: Well, AI can also speed up the pace of discoveries. We’ll listen to 2025 chemistry laureate Omar Yaghi talk about this, but first we need some context. He talks about COFs. What are they?

Smith: Covalent organic frameworks. They are the next iteration of metal organic frameworks, which is what Yaghi and his co-laureates were ordered the Nobel Prize for, which are these basically nets. It’s a thing called reticular chemistry where you design net-like molecules that were a great surprise to everybody that they could be stable and crystallised. And they turn out to be immensely useful in all sorts of potential applications, and they can be used to absorb all sorts of molecules from the atmosphere or from solvents, from water, and store things, and then they can be used to release them. So they have innumerable uses, and there are thousands and thousands and thousands of people working on the various generations of these chemical nets.

Svensson: So let’s listen to what Omar Yaghi has said about COFs and AI.

Yaghi: My decision to pursue AI more strongly is based on facts. And the facts are that with AI, we can double the rate of discovery. Students that are using machine learning algorithms that have been adapted for discovery of new classes of materials make double the discoveries compared to students that don’t. So my logical mind or the logic in parts of my mind says this is a great direction. Imagine making a new material like covalent organic frameworks. It’s easy to make coughs, but it’s not so easy to crystallise them because you have to go through a lot of trial and error and finding the condition under which they would crystallise. And sometimes it takes a researchers two, three years, maybe even more, to crystallise a cough. But by just using an LLM, a ChatGPT, to look at all what has been done across chemistry in the open literature and ask it to give you conditions under which such cough could be crystallised. And we devised a sequence of a cycle where each cycle of communicating with ChatGPT gives you better and better solutions and better and better conditions under which you can crystallise your COF. So a COF that would take you years to crystallise because of the trial and error is now taking us only weeks, even less. And this is a new contribution that we made.

Smith: That’s extraordinary. So it can really look through at what everybody else has tried and come up with sensible suggestions that are better than…

Yaghi: In our case, it gives you a whole bunch of conditions. And we have a robotic system that checks these conditions and classifies their crystallinity. And you put that data back into ChatGPT. The reactions that gave you crystalline material and the reactions that didn’t give you, this positive and negative data is put back into ChatGPT, and then it gets better and better at giving you good answers. The thing that amazes me is that it also can correlate things that you normally wouldn’t correlate. So it might come up with 90 different conditions that you and I would look at or a trained chemist would look at and say, “Okay, well, that’s expected.” But those 10% that it might come up with or even less, 1%, 2%, make the big difference. We’re not looking at the middle of the bell curve. We’re looking at the edges because the edges is where the novelty comes from. And so even though you have few conditions that may be fertile for what you’re trying to do, they’re still very significant and they make all the difference between having a mundane discovery versus a major discovery. So it’s a no-brainer for me that this is an area that we need to invest more in. If we wanna transform science, scientists need to experiment with it.

Svensson: Well, he sounds very enthusiastic.

Smith: Yeah. Well, what’s not to like about that? Of course, if it can get you to the solution to your problem faster in original ways, it’s wonderful. I suppose, if one’s looking for a downside, the downside might be that in so many cases in science, we’re hearing of AI taking what are, you might describe as mid-level problems and solving them. So these are things that were traditionally the kind of groundwork on which you built your career. You learned to do these things, and because you learned to do them, you became a better scientist and you prepared to be a great scientist. And now if you just can press buttons and get AI to solve your problems, where are you learning? What are you learning? It’s a whole question around skillsets and what makes one a thinker? What makes one creative? So it’s really interesting and challenging.

Svensson: But is this positive attitude towards AI typical for the scientific community, you think, that Omar Yaghi displays?

Smith: I think a positive attitude is pretty typical generally for scientific community. The optimism about potential. We were having a fun conversation in Japan recently. We’re realising that a good t-shirt that scientists might wear could say optimistic, but not stupid. And basically that kind of sums it up, you know? Of course, you’re optimistic about the potential for new technologies to transform things, but you don’t go into it blindly. You need to think through things. And that’s generally the scientific approach.

Svensson: Well, for the sake of balance, let’s bring in a more critical voice. And ironically, it’s one of the scientists who made AI possible, 2024 physics laureate Geoffrey Hinton.

Geoffrey Hinton: So the real question is, not will they get more intelligent than us, but if they’re more intelligent than us, will we have a way of making sure they don’t want to take over? And we just don’t know. We don’t know whether that’s possible. But given that you’re about to make things more intelligent than you, it would seem wise to put a lot of resources into figuring out if you’re gonna be able to keep control. We’re not gonna stop AI. I think saying we should stop now, that might be the rational policy, but that’s not gonna happen. There’s too many profits to be made.

Svensson: I always feel like my pulse starts to race when Geoffrey Hinton talks about the risks of AI and the scary future there might be upon us soon.

Smith: Yeah. He speaks about it very compellingly. And who better, as certainly one of the inventors of it all, to worry us. And he’s not alone. Of course, there are many voices saying similar things and they seem to know what they’re talking about. So I guess we need to worry alongside them. I don’t know. Something that occurs to me is that every time that there are scare stories about the power of AI, it all bolsters the kind of hype around the power of the technology and how transformative it’s going to be. Of course it’s transformative, but somehow when something like this is going on, the challenge for humanity in general is to sort out the reality from the hype. And to stay cool and levelheaded in the face of it all. Not easy.

Svensson: 2024 chemistry laureate, John Jumper, who figured out how to use AI in sort of cracking the problem of protein folding. He thinks it will change how we view experience.

John Jumper: As fast as this field and world is changing, you really need adaptable people. We should downweigh experience because after all, the field we’re in AI keeps changing so fast that nothing from four years ago is all that relevant to today. And so, like, 10 years of AI experience is still only really three years of AI experience in that way.

Svensson: So flexibility before experience as being important in science. How do you feel about that assessment? I

Smith: I think it’s exciting, really. It speaks to the possibility that, um, there’s much more flexibility in who contributes and how they contribute. And such an arresting thought, because it’s contrary to what the world has been built around for a very long time. You know, it’s like the question of why’d you go to university? What do you go to university for? Some people might say you go there to get knowledge to be educated, but I think the main reason to go to university is to work out what you don’t know. I suppose experience does, to a certain extent, allow you to have that perspective on what you don’t know. That might be important too. I mean, the history of science is that people turn up and say, “Let’s try this.” And then people say, “This, that won’t work.” But it does work, and that transforms things. So in general, that’s great. But sometimes you, we’re all aware of people turning up and saying, “Let’s try this,” and it’s just bonkers. It’s based on the fact that they really don’t have the experience. So there’s a balance between knowing nothing and knowing enough.

MUSIC

Smith: I was having a conversation with a great AI advocate, and she was dismayed by my lack of desire to use it. She really told me, “You’ve gotta go and experiment more with it.” But one of the experiments I did was a small little thing, but I’d just been to the Prado. And I’d spent quite a lot of time looking at Goya’s black paintings, which are arresting and disturbing and confusing. And I asked ChatGPT to describe a picture that’s very well known called Saturn Eating His Son. It’s very dramatic and very widely reproduced. Anyway, very quickly, of course, immediately produced three paragraphs, which I began to read, and it seemed to be describing the picture very nicely. And then it removed all those paragraphs. And there was a message saying, “This content contravenes ChatGPT’s usage guidelines,” or something like that. Presumably, it had realised that it was talking about a father eating his child which was not content to be to be discussed. But I thought it was interesting that it could write its own content and then read it and self-censor. So that was revealing but not very useful. But then I asked it to describe a painting which is very alignmatic, a huge canvas, which I think some people sometimes call the drowning dog. It’s basically a dog at the bottom of a very gold rectangular picture.And it described it in beautiful detail with much scholarship attached to the description. But reading it, I can only phrase it this way. It struck me very strongly that it had never seen the picture. It’s a funny thing to say, but it came across in the writing that it was extremely scholarly and accurate, but what it was describing wasn’t quite what you see when you look at the picture, you stand in front of it. That, I suppose, is, of course, getting better second by second. It somehow encapsulates the problem that it, it hasn’t really seen. It’s synthesising everybody else’s vision, but that doesn’t mean itself can see. And see, you can extend to be all sorts of things.

Svensson: I mean, this is why this also is an area for philosophers and psychologists, is that there are so many aspects of how it interacts with us and how we perceive it. And I remember asking for travel advice from ChatGPT, and it was saying, “Well, this is my favourite coffee shop in Vienna.” So I was asking, “What do you mean when you say it’s your favourite?” Because obviously you haven’t been. And it’s like, “Well, that’s my way of explaining it so that you will understand.” It’s sort of adapting to the way that we see the world. That really did a number on my head, I think, because it’s trying to please us at the same time as it’s sort of manipulating us.

Smith: Absolutely. So to which you say, “So you’re lying to me.”

Svensson: I didn’t wanna annoy it, so I didn’t say that.

Smith: You didn’t wanna offend it. Well, it’s all so peculiar.

MUSIC

Svensson: So the 2025 prize in economic sciences, it’s deal with a lot of issues connected to AI. Why are economists so interested in AI?

Smith: That’s a huge question. I mean, I suppose you just have to look at the news. It’s having such profound effects on the economy from top to bottom. And again, separating hype from reality in that, and how transformative it is on the job market or the productivity of firms or any of it. It’s very much front and center in people’s thinking.

Svensson: We’ll hear Peter Howitt and his co-laureate Philippe Aghion’s prediction for the future. But first, an AI surprise from Howitt’s Nobel lecture.

Smith: You yourself have embraced the technology in as much as you allowed ChatGPT to write a paragraph of your Nobel lecture.

Speaker in clip: Come up to the stage, Peter Howitt.

Howitt: Generative AI has the potential to displace a significant number of jobs because it automates not only the routine repetitive tasks, but also complex cognitive work. There was once thought to require uniquely human judgment. As models become capable of producing high quality text images, software code and analytical outputs, entire segments of knowledge-based occupations from customer support to administrative roles to journalism, marketing, and parts of the legal and financial system face the substantial risk of restructuring or outright redundancy. That was written by ChatGPT, not by me. I didn’t change a word. I asked it to write a paragraph on the job destroying potential of generative artificial intelligence. I think it’s pretty good for a first draft written in about five seconds by a three-year-old who’s growing up very rapidly into something that well, we don’t know.

Howitt: Well, I thought it was a catchy way to make the point of how powerful the technology is. And also its potential for creating what we would call superstar markets. You know, writing a speech has become a lot easier. And many other tasks have become a lot easier because you can farm out a lot of somewhat routine tasks, some less routine tasks. It has to be checked, of course. And still makes lots of mistakes. But it’s definitely going to make a lot of people much more productive. The area of speech writers was the one that occurred to me that I could imagine somebody writing speeches for politicians or business leaders, whatever, could produce many more speeches in a day than they could have before, just with the help of this drawing up a first draft or coming up with ways of expressing things that have to be checked, have to be modified. What that’s gonna mean for a lot of markets like this is the very best people are gonna become so productive that there isn’t gonna be room for a lot of other people in that field. And, you know, sometimes people, they ask the question, is AI going to enhance productivity or is it going to replace jobs? In a sense, it’s going to do both. Because it enhances the productivity of the most productive, it’s going to leave fewer jobs for other people in a lot of areas.

Philippe Aghion: It will replace a lot of tasks that were not being replaced by… Now, a job is a collection of tasks. So for most of the jobs, it will maybe replace the most routine tasks so that you can concentrate more on more creative task. But for some jobs, it will make the whole job redundant. So it will lead to some job disruption, but I believe that it will also create new jobs for at least two reasons. First, firms that adopt AI become more productive, more competitive, and therefore the world market demand for their products increases, and that make them create new jobs. You see, they hire more to face higher demand. And the second reason is that AI, you can have many more new ideas with AI. New ideas are often recombination of old ideas. With AI, you can recombine much more. And new ideas are new jobs. The whole challenge is to smooth out the transition from old jobs to new jobs.

Svensson: I think the word challenge is the key, don’t you?

Smith: Yeah. It’s a very challenging scenario. I don’t think that it’s fair to define a job as a collection of tasks. Because pretty much everybody I know who does a job is more invested in it than that. They don’t just see themselves as a cog. The firm might see them as a cog and capitalism might see them as a cog. But people don’t see themselves that way.

Svensson: It’s sort of that nightmare scenario of having to sit down and write down the tasks that make out your job.

Smith: That’s interesting, actually. Yes, you’re right, because bosses often ask you to do that. Could you tell me what you do? It’s pretty hard, actually, isn’t it? Yeah. We’ve all had to do it. Personally, I don’t really agree with that speech writing thing because I don’t actually want to listen to a speech that’s being generated as an amalgam of what other people have said. Which is currently what ChatGPT would produce. It may make perfect sense and be delivered as the consensus fact. But it’s not actually what I want to listen to. I’d rather listen to one person’s view and then another person’s view. But it all depends on what your goal is. You have to ask what the goal is, what are you trying to achieve? It’s obviously an enormously powerful thing. But just because you’ve got a hammer, you don’t go around hitting everything. You need to decide where you can be most effective and where you need to ease off.

MUSIC

Smith: There seems to be a bit of a separation between everybody’s individual worries about what the power of AI is and the headlong rush into its incorporation. And it’s funny in a way that individually, we all have our sense of doubt. But that sense of doubt doesn’t seem to translate into the way that the world is behaving towards it.

Svensson: That’s interesting. That leads us into another concern for economists. Well, for all of us about who is driving the development of AI. Why are so many economists and people in general worried about the role of big tech companies in this revolution?

Smith: I suppose it’s their unprecedented control of the technology. They seem to be writing the playbook for this and the basic development and everything that goes into it through to its rollout in society. They’re having such a part to play. And that surely is worrying. Because whatever they say, and however much they talk about striving for the good of humanity, at the same time, everything they say is marketing. And everything they do is marketing. So, because that’s also what they’re about. So, yeah, we need more voices in there.

Svensson: Well, Philippe Aghion argues that this started during the IT revolution. Uh,

Aghion: Google, Microsoft, Amazon, Walmart, initially, they boosted the growth process, okay? They knew how to harness IT better than other firms. But eventually, they became so tantacular, so egemonic, that they ended up discouraging entry of new innovating firms. And that explains the growth decline in the US as of the 2005 or so. These firms initially boosted growth, but eventually, they became an obstacle to growth. So, already with the IT option. So, there is the same danger with AI. In fact, if you look at the cloud, the cloud is dominated by Amazon, Google, and Microsoft. Those same firms are there. So, AI has a big growth potential because AI automates task in the production of good and services and in a production of ideas. But the danger is that those same firms that ended up discouraging entry of new firm would discourage entry.

Howitt: Once you get established and you get a big industrial complex built up, you can continue to be very innovative as the big companies in the US still are very innovative, but they don’t tend to be the really big, disruptive, major innovations that are gonna really push technology forward. There’s a danger that things will get dominated by the status quo that may continue to innovate, but they’ll also spend a lot of their resources trying to prevent others from displacing them. And they’ll, in effect, end up suppressing a lot of radical innovations that could take place. And eventually, you know, if you keep following the trajectory that any of these firms are on, you’re gonna run into diminishing returns, and you’re not gonna be able to sustain growth.

Svensson: Geoffrey Hinton, who is dubbed the godfather of AI, believes there is a safety risk in leaving the power of development of AI in the hands of a few large companies.

Hinton: My belief is, government’s the only people who are powerful enough to deal with these large companies, and even they may not be. So, my belief is the government ought to mandate that they spend a certain fraction of their computing resources on safety research. Now, it would be great if that happened. And the Biden administration was moving very timidly towards a little bit of regulation. In Europe, the Europeans would like to have some regulation of AI, although they explicitly say, “We’re not gonna regulate military uses of AI,” because so many European companies wanna use it for weapons. The UK and the US have said they’re not gonna sign on to the European’s Declaration about AI safety. Basically, they explicitly say, “We’d rather have the profits than the safety.” They say that by saying, “Too much regulation will interfere with innovation.” But you can rephrase that as, “When it’s profits versus safety, profits win.”

Svensson: There’s also the concern of how to make AI beneficial for all and not just for a few people or companies.

Smith: Very much so. And too infrequently, the conversation extends to involve the great majority of the world who are not so touched by the things that are being talked about when people are talking about the replacement of jobs. A lot of the time, it’s talking about it in the most developed economies.

Svensson: There is this sort of science fiction dream of a society where everyone will have everything, and because everything is automatic, but that requires a lot of political will in that direction.

Smith: I was told that there are philosophers working for AI companies. On the question of how we’ll deal with the post-scarcity world. Could that really be true?

Svensson: Well, haven’t you seen Star Trek?

Smith: Who can possibly imagine that we are headed for any such world as you watch countless thousands of people trying desperately to get from places of utter scarcity to places of slightly less scarcity?

Svensson: Well, if you look at technological development in the last couple of hundred years, then if we keep on doing that, we’ll probably have the technological abilities to do this. It’s just that we. No one’s driving that because they think it’s too utopian.

Smith: Yes. And you’re right, that’s the goal that one should drive towards. It’s good to have science fiction, optimistic science fiction dreams. It doesn’t come up much dreaming in conversation about what the technology will do. Maybe it would be good for people to be a bit more crazy in their thinking and just to allow themselves to dream of what they want the world to be like and where they’re really trying to get to. So I think you’re raising a very important point. In all the AI discussions I’ve been involved in recently, I haven’t heard anybody do that. Actually say, “You know what? This is what we actually want in the future.” How do we get there?

Svensson: More utopian visionaries. That’s what we need.

Smith: Maybe. I don’t know. It sort of fits in with a Nobel Prize laureate theme, because sometimes when you’re talking to Nobel Prize laureates, you can sort of see that great scientists, they dream of possibilities, and they’re far out there just pondering the impossible. Maybe that’s good practice.

Svensson: We have one example of a little bit of utopian thinking in Daron Acemoglu, who was one of the 2024 economic sciences laureates. And he also has strong opinions about corporations, but proposes an alternative future for AI.

Daron Acemoglu: Humanity has never seen corporations as powerful as the tech companies. Now those companies are choosing the direction of another extremely powerful technology. And whatever your view on other things may be, I think most people have an instinctive agreement with Lord Acton when he says absolute power corrupts absolutely. So we are in such a situation. But one thing is very clear that there isn’t a universal law that good technologies will give us good outcomes. There are many instances in which better communication technologies have led to worse information. There are many cases in which technologies that have increased productivity, have deepened poverty, reduced wages, increased inequality. So really the details of how technology is developed, who is empowered, what regulatory and other countervailing forces there are in society matter greatly. Most people don’t even realise that there is a socially beneficial and technically feasible different direction of AI that would give such better outcomes. So that’s the collective imagination. But imagining isn’t enough. Different people have different imaginations. Different people have different incentives. In the current environment, it’s not your imagination or mine that matter, but it’s what the tech bosses desire. So that’s the sense in which if we can bring more people to the table, that might even the scales when it comes to whose interests are gonna be favoured with these decisions.

Smith: It’s difficult for people because they’re struggling to keep up. They’re struggling to get ahead with a technology that’s changing so very fast. That’s the scenario that these mega companies are thriving on because the pace just stops anybody from having the time to reflect and say, “Hang on a second.” And of course, lots of people are calling for what Daron Acemoglu is so eloquently talking of, but, you need people with determination to bring this about. I think everybody actually, the tech companies themselves, I think everybody believes that this is the right way to go, but it’s difficult to do that on top of an overwhelming day job.

Svensson: Well, at least it shows that we really need to make time and space for dialogue.

Smith: Yeah, it does. And again, I come back to the young people we talk to all the time who voice their real concerns about so much of this, that they least of all have time because they’re so busy just trying to work out how to navigate the path ahead. And yet, we must prevent everybody being caught up in an unstoppable direction that we didn’t want. It’s a really important thing to be talking about. So conversation is exactly what we need.

Svensson: Well, AI will, for better or worse, shape our future. So let’s look a little bit at how the laureates are addressing this future. For Omar Yaghi, it’s important that scientists have a seat at the table when AI is regulated.

Yaghi: We need to be participating in these activities so that we can make sure that the things we need for science are protected and not over-regulated. I’m all for it. Obviously, we need to be thinking ahead. What if my models can do everything that a graduate student does? What happens to graduate school? How do we train the next generation of thinkers? All of these things are not gonna happen unless we experiment with this new way of doing things. Otherwise, we’re just gonna be swept aside. We’re gonna become irrelevant. And we’re not gonna be able to recruit the best people into our science. I mean, really, scientists, it’s all about experimenting for discovery, and discovery is our currency.

Svensson: So what is the role of scientists and thinkers in this discussion?

Smith: I mean, that’s two different things. The role of thinkers is that everybody should be involved. Everybody’s a thinker. And it needs to be something that is, as Acemoglu says, in a way, it’s a democratic right to be involved in something that’s so influential and changing everybody. So thinkers around the world should unite to talk about this. The role of scientists, well, they’re good at evaluation. They’re good at being skeptical. They’re good at testing out ideas and not being too committed to a way of thinking in the face of evidence that says that that is wrong. That is one role that they play to bring evidence-based thinking to the development and utilisation of AI. I mean, that’s starkly illustrated when you bring scientists together with politicians. Because often you kind of encounter a scenario where the politician says, “Yes, okay, I’ve heard your evidence-based argument. I understand what you’re talking about. And now here’s the political reality.” And somehow, you know, these two things need to come together. It won’t wash if I simply go out there with, with evidence.

Svensson: Geoffrey Hinton has made it his mission to warn us about what could happen if this field isn’t heavily regulated. But when you spoke to him, he also delivered his best case scenario, referencing his former postdoc student, Yann LeCun.

Smith: How do you think we should think of them in the future? As friends, aliens?

Hinton: Okay, so Yann, who thinks we’re gonna be safe, thinks we should think of them as servants. Good old-fashioned servants who do what you tell them to. If they don’t, you fire them. I’m just worried by the fact that there’s very few cases of more intelligent things being controlled by less intelligent things. Once they’re a lot smarter than us, I don’t think they’ll put up with that. Well, that’s what worries me at least. Now, there’s one line of argument that’s more promising, which is a lot of the nasty characteristics that people have come from evolution. We evolved with small warring bands of chimpanzees, or our common ancestor with chimpanzees, and that led to this intense loyalty to your own group and intense competition with other groups, being willing to kill members of other groups. That’s sort of shows up in our politics. Quite a lot right now. These things didn’t evolve. And so it may be we can avoid a lot of that nastiness in things that didn’t evolve.

Smith: It’s a nice thought that we could learn how to behave from them.

Hinton: Yes. In fact, AI mediators are now quite good at getting people with opposing views to come to see each other’s view. So there’s a lot of good can be done with AI. If we can keep it safe, it’s gonna be a wonderful thing.

MUSIC

Smith: Science and reporting tend to be delivered in a very good or very bad way. It’s either that science is going to cure this disease or science has led to this chemical spillage or something like that. It’s a disaster. There’s not much middle ground. There’s not much science reporting of actual debate and though. In a way, we’re reaping the rewards of that mistake that it’s a very polarised discussion. It’s, of course, far more nuanced than most of the discussion implies. Technology needs to be discussed in the same way that we discuss other things that we understand better. Therefore are more willing to get involved in a nuanced way.

Svensson: Shimon Sakaguchi gives some very sage advice on how to stay steady when technological development is moving so fast.

Shimon Sakaguchi: Always I’m saying to my student that, uh, when you drive your car at a very high speed, you must look far away. Just looking in front of you, your car is so dangerous. You cannot drive. Nowadays, science is a very rapidly progressing. New technologies, analysis, and then bioinformatics, and so on and so forth. What is important is always think about not to be nearsighted, but then just looking at far away and then what is important. We are going to that direction.

Smith: You have to keep your eyes on both. What’s right in front of you and what’s far away. That’s what we’ve been saying really throughout this, that the trouble is that you need time and encouragement to do more than just deal with what’s in front of you on your experimental bench.

Svensson: But what do you see when you imagine the future with AI?

Smith: Well, okay. Honestly, I suppose I imagine it as problematic. I would hope for a world in which the wider concerns of things like conflict and deprivation become more important than the simple desire for power and money. Because undoubtedly, along with many other things, this technology has the potential to make an enormous difference for good.

Svensson: So do you think we’ve sort of penetrated the problems of AI and the benefits?

Smith: I don’t think we have. I think the laureates have in their statements. They’ve illustrated some of the things to be thinking about.

Svensson: There are lots more to be said in future episodes.

Smith: It won’t stop. It’s just gonna get bigger and bigger. People are gonna talk about it more and more. And hopefully in more and more sensible ways as we go along.

MUSIC