Special Report: Will the AI revolution end politics as we know it?

21:03And this is just the beginning. Wait until robotics come along. As you know, if you listened to our last episode, where we played the conversation in full, we decided to go to San Francisco to talk to Anthropic ourselves.

21:20Thank you.

21:25Lewis:Right, so we've come to downtown San Francisco, a city, by the way, which is being revitalized by the AI industry more ways than one. It was basically completely written off after Covid. Downtown completely empty, all the buildings empty, as you've heard about in a previous episode, fentanyl problems. As a result of the AI revolution, the city is being completely revivified because it is San Francisco that is the heart of the AI investment, not Silicon Valley, about 50 or 60 miles south of here. But we've come to Anthropic. Anthropic is one of those companies you may have heard of, you may not.

21:55Lewis:It's one of the big, big, big players in AI right now. It's now worth about$60 billion. That itself is only a slither of the trillions of dollars which is said to be riding on the latest AI models which are either being developed or about to come out. We're here to speak to someone who's one of the most senior figures of this company because what makes Anthropic a bit unusual is how willing they have been to openly discuss the political, moral, philosophical and economic problems which may inevitably arise from the work that they themselves are doing, or at least if we don't get the reaction right to it.

22:32So I'm Jack Clark. I'm one of the co-founders and heads of policy of Anthropic, and we're in the ground floor of our office in San Francisco right now.

22:40Lewis:I'm very nice at this too. The way in which we talk most about this in terms of its societal impact at the moment is the economic impact. Your CEO has talked about that. He's talked about a bloodbath of white-collar jobs, warning that AI could wipe out half of all entry-level white-collar jobs. Do you share that assessment? Because that sounds quite scary. If the technology continues to get good, we need to be prepared for things that have almost no basis in economic history to happen. Business as usual economics says this doesn't happen. It says the technology comes along, there's a negligible employment impact, there's some productivity impact, some jobs change, it takes a while, it's normal.

23:23Where we're coming from is if we're right and the technology keeps getting much, much better very, very quickly, there's the potential for really abrupt changes to happen that have almost no precedent in the economy. The way we think about it is you could have something like the Industrial Revolution, but it takes place over tens of years instead of a century. And a technological revolution that takes place within one generation rather than multiple generations would be radically different in terms of how we deal with it. So yes, I think there's the potential for really large-scale employment impacts to show up.

23:57We don't see this yet, but what we see is the technology getting much, much better, much more quickly than people suspected, and showing up in terms of how it's sort of being deployed and changing professions more quickly than people anticipate it.

24:12Lewis:And why do you think that is different? Why is it? Because, you know, I encounter this all the time for this piece, which is, you know, humans have a generally actually very often have an optimism bias. We assume you've identified it before. This is like other technological changes. It creates jobs. It loses jobs. There's an equilibrium to it economically. Why is it different? Why is this different to that? Why would conventional economics be wrong in that analysis, in your view? Speed is different and quantity is different. And AI has both of these properties. What I mean by quantity is that AI is available over your internet connection.

24:48So once a system gets better, it's not like people need to build some physical infrastructure to make use of it. It's just instantly there. It instantly distributes. That ties in with the speed thing. it is moving a lot faster than normal technologies. If you look at the speed of aircraft, it went up really steeply for a while, and then it just kind of stagnated. Same with cars on roads, same with things like the size of ships. They go up some S-curve, and then they flatten out, they asymptote. What we're seeing with AI is the S-curve is really steep, and we're still on it. and there aren't signs yet of it flattening out.

25:30And every time that we bring out new systems, they get much, much better, much more quickly than people who have been used to in the past. So it has these properties where it's evolving more quickly than normal technologies and it's distributing into the world a lot faster than normal technologies. These don't really have as much precedent. It sounds almost, and this might sound weird,

25:52Lewis:it sounds almost Marxist in a sense that something that Marx talked about was the idea that technology would advance to the level, and he was talking about this in the 19th century, but the technology would advance to the level that would make almost labor redundant because there would be such abundance as a result of the technology created. That is a world, I mean, it might not be exactly in that way, but it strikes me, listening to you, listening to others, that that is a world which could be credible in the sense that, and this might not mark, it's meant it in a positive way, but it could be a negative in the sense that people might literally use their ability to use their labor.

26:27Lewis:And that's never happened before, that there may be very little economic value to lots of people's labor. And if that's the case, that transforms economics in a way that we've not had before or seen before. That's true. And I think it's something that we view both as one of the greatest chances society has to change how society works in positive ways that people would like, and also one of the greatest risks. Because for hundreds of years, we've had technological innovations and people have said, oh, people will be working a three-day workweek as a consequence of this bounty. But what you instead have is some level of inequality and some level of diffusion, which means that none of that happens.

27:05You know, people still work a bunch, but people don't feel like they're sharing in the gains from production. So that's something we have to get right. And this is the point. AI does have vast potential to release productive energy and huge productivity gains in our economy. Its potential to enhance the technological world around us is vast. It could do everything from curing cancer to doubling our lifespans to cracking cold fusion. Invent technologies the human race might literally be too stupid to conceive. But what happens on the way there? What happens if Marx was right? Tech gets so good that people aren't needed, or most people aren't needed anymore, when labour stops being a commodity which is guaranteed to be valuable even at some level, that could transform the economic history of the human race.

27:58When there is such abundance, our labour might not be necessary. If that happens, does our politics shift with it? Does our economy deliver the plenty created to the most, Or does it do what we've done before and deliver it to ever fewer and ever fewer people? Because here's my worry. It is easy to see how AI could make inequality even worse than we've yet seen, as automation suppresses job creation and wages in low-skilled occupations while supporting new elite job types, raising wages to stratospheric levels to those in the know. i.e. it makes the dog-eat-dog world of globalization, which has already been at the heart of the fractious politics of these last years, even worse.

28:46We've already started to see it. Most common job in 29 states is still being a truck driver. You have robot trucks just now hitting the highways. The layoffs that are rippling through the tech industry aren't showing any signs of easing. IBM is the latest tech giant to do so. The company announced it will cut about 3 ,900 jobs, mainly in its IT services segment. We've become so rich from AI that maybe you can do UBI, and not everyone has to work. At a fraction of the cost of a human lawyer, Rauter says he is able to use his software to help everyday Americans battle legal issues, such as traffic tickets and evictions.

29:21So far, he says his company has helped in more than 1.5 million pieces and saved his clients millions of dollars. A full-blown Hunger Games for top AI talent, complete with secret deals, backroom betrayals, and billion-dollar bonuses flying like confetti. Amazon CEO Andy Jassy says artificial intelligence will lead to a reduction of its corporate workforce. So there's going to be a beneficiary in the organization, but you're just not going to need as many people in that department, and that's going to play out over and over again. And again, if you don't think this is already happening, You don't have to travel far from Las Vegas to see it for yourself in one of the oldest professions in the world.

30:03Lewis:I never wanted to be a farmer. Right, so I'm sitting on this autonomous tractor in the middle of the cherry trees in this truly massive farm. Oh God, what's it doing? Oh, whoa, whoa, whoa, whoa. The wheel's turning. Oh, I don't like this. I don't like this. look at me mom i'm farming i'm farming

30:33Lewis:no hands i tell you what who needs clarkson who needs clarkson so this is actually this can drive itself it's an autonomous tractor it can drive itself correct And how new is this tech? We've been around for four years now with the autonomous tractor product. We take regular mechanical tractors that you can buy today from any dealership in the UK, in the US. We retrofit the tractors with an aftermarket kit. We install a communication module on the back. We install a sensor suite in the front with our additional component to control the mechanical functions of the tractor. Yeah. And we, within a day, make the tractor to drive itself.

31:21And these are commonplace? The technology is becoming more and more common. Correct. And why? What advantage does it offer? How hot it is outside right now. Right. Do you want to drive the tractor? Oh, you don't want me to drive the tractor. Do you see that chemical tank? Yeah. Do you want to be next to it when the tractor applies these chemicals? Do you want to do it for 12 hours? Are you offering me a job? If you want to take me. You really don't want me. Your farm wouldn't last, believe me. A lot of the folks out here will take you like that, no problem. They're looking for a trial. No, they wouldn't.

31:53The Central Valley of California is home to some of the most fertile farmland in the globe. Until recently, it would be common at this time of year to see hundreds in the fields, often migrant labor, tilling the land, picking the bounty, the dates, the almonds, the peaches, the cherries. No longer.

32:11Lewis:I mean, the automation here is obviously really interesting and it's fascinating, but I suppose it takes place in an economic context, right? What's the situation in terms of agriculture in California at the moment? And I suppose, how does the economic picture sort of feed into the automation picture? Well, the typical employee in California, the tractor driver, makes a surprisingly high wage, I would say. It puts a very challenging environment for growers to actually make money. And of course, when growers don't make money, we pay more in the grocery store for what we buy. So outside of that, there's also a federal program that helps growers with bringing employees.

32:52but obviously the cost of that is a lot higher. It makes the growers to pay for the living and transportation of these employees. And ultimately, the their ability to make a buck at the end of the day is very, very limited, especially within the prices of the commodities in the past few years.

33:12Lewis:So when you say that, say, tractor drivers in California make a lot of money, how much money are we talking? I would say anywhere from$50 ,000 to$80 ,000. $50 ,000 to$80 ,000? Yeah, and some even more, depending on where you are. So that's not people who own a farm, that's just people who drive a tractor? That would be the people driving the tractors. Wow, I mean, that is a lot. That is a lot. Not to say they don't deserve it. I'm sure it's really hard work. Don't get me wrong, but it is a lot of money. It is a lot of money, and you need plenty of them. Seasonal, not seasonal, you need a lot of employees.

33:47and our ability to impact your, like I said, your reliance and dependency on these employees or even just shrinking that number to a very good core of employees is key for these girls to maintain sustainable businesses. If they don't have to worry what happens in peak season, low season, how do I maintain personnel on the farm? It's priceless for them. And so the cost of having to deal with employees' safety regulation, they prefer to go to automation.

34:20Lewis:Why is the price of labor so high? Well, first of all, the minimum wage in California is relatively high. So that's to start with. Then no one wants to do it. Supply and demand. If I need someone to drive my tractor and I can't find them, I'm increasing the price and I'm increasing the price and I'm increasing the price. And therefore, the price. I mean, I suppose historically, agriculture, not just in California, but maybe especially in California, it's relied a lot on immigrant labor, sometimes legal, sometimes illegal. That presumably is more scarce than it used to be. It is. It is. Without going too much into the details, I can tell you that there's definitely been an impact here recently.

34:59We see it from events being canceled because they don't want immigrant workers to show up in these events and then being potentially caught or deported. Or raided. Yeah, literally. All the way to people not wanting to put themselves in the spotlight for the chances to being caught and deported. So we have a little guest here.

35:27So the perils are doing farming, interviewing. Talking about track drivers. Yeah, yeah, yeah. There you go.

35:34Lewis:So I think this farm and agriculture generally, it really encapsulates for me and sums up for me my split feelings, the duality within me about AI, because you can completely understand it and see the benefits right here. I mean, actually, he's right. This is oppressive heat here. Do people want to actually be farming, sitting on those tractors, sitting in that heat, working in that heat? well, clearly not. Clearly not, because they can't get people to do it. That is why labour costs, or one of the reasons labour costs are so high. And that's the other thing. The truth is that California, other blue states in particular, but other parts of the world, industries like farming are very, very regulated, probably over-regulated.

36:18Lewis:And so when you've got high wages, high costs, high heavy regulation, all of the incentives, all of the incentives are driven to are made towards automation. They've just got to be. And you can see potentially the benefits as well. Theoretically, at least, it might not happen because big agricultural companies could just bank the profit. But if you automate, you could try and bring the cost down. And food prices are really, really high in the United States, partly for the reasons that we've just been discussing. And yet, so you can see all those benefits, feel all those benefits. And yet, also at the same time, it's the same old story.

36:55Lewis:Whatever jobs there are, are going to be lost. and going to be consumed, and they won't come back. And they do employ people, albeit in smaller numbers than they did before. Agriculture, the oldest part of the human economy, on the verge of another agricultural revolution. And it is happening right now.

37:17Lewis:Stay across the latest news, breaking stories, and respected opinions by listening to LBC. From award-winning internationally respected journalists. Good morning, I'm Nick Ferrari. James O 'Brien on LBC. Good afternoon, I'm Sheila Fogarty. Very good afternoon to you. Tom Swarbrick with you here. Hello, this is Andrew Marr. Hello, you're listening to Sunday with me, Lewis Goodall. Find out what the story means for you. Listen on our free Global Player app or the new LBC app. LBC, leading Britain's conversation.

37:51The News Agents The economy is one thing. In some ways, as new as AI is, it's the most familiar political impact the tech might have. We have seen new technologies before, albeit not as quickly or as potentially dramatic. But there are so many ways that politics will have to adapt and adjust and rethink a set of technologies which might affect the basic human state It's all the basic human understanding of things as basic as life and death itself. Back to AI4.

38:23Lewis:So one of the weird things about this place is the kind of breadth of the way that AI is being used or is often being used. It stretches from the ultimate kind of like just humdrum and banal of just say, you know, using software or AI to kind of subtly change your business processes or kind of like the computers that your company uses. through to the kind of automation that we've been talking about and the robotics, which is kind of scary, particularly in economic terms, to the downright weird. I'm looking at a stand called Reflector, which is literally promising, in a sense, to bring your dead relatives back to life.

38:57Lewis:And I'm looking at its brochure. Every day, they say, an unprecedented amount of human experience is slipping away. In 2024 alone, 62 million people died worldwide, the largest number of deaths in human history. The urgency is clear. humanity faces a quiet crisis of memory and our task is to stop that and arrest that from happening. Hence why, according to this company, if you pay them a certain amount every month, you'll be able to talk to your own relative who died months, weeks, years ago. Yeah, my name is Greg Matusky. I'm one of the founders of Reflector. And tell us about Reflector. What is it?

39:35It's a very unique platform that lets anybody virtualize someone from their past that they love and cherish. It's a way for families to capture and document stories from the past and share them with future generations. What you do is you load in information about the individual, and then conversationally or through text, you can train the model. You give a fingerprint of the voice, and then you can start interacting with it, talking back and forth. Like my father, World War II veteran, he's virtualized, and I can talk to him about his experiences in World War II, the stories he's told me for years, that now my grandchildren, my children and grandchildren can hear from his mouth.

40:11Lewis:So let me get this straight, Greg. So you can input sufficient data into your program, which will allow people to communicate in a sense with their relatives who have long since died. Yes. So you put enough information in, it prompts you. And then you can conversationally tell more information over time in the biographer, and it virtualizes that person in their voice. And you can do this through text, or you can do this through voice into the model. And then you can have a conversation. You could ask it, you know, dad, what would you think of AI in 2025, which he passed away in 2005? So it's a unique way to capture stories, but also to stay connected and receive advice from those who have, you find important in your life.

40:58Sounds like pure science fiction. It is pure science fiction. It's all here because of AI and its ability to tell stories and work with human beings to capture past events and histories and family legacies.

41:11Lewis:Is there not a bit of a profound danger with it though, Greg? I mean, in a sense that once people have died, I mean, to die is a fundamental part of the human condition. To grieve and to lose is part of the human condition. And you could basically be in a position where you're persuading people that they're talking to their dead relatives, but they're not. It's a fiction. It's a phantom. It doesn't really exist. It's a machine. I speak with my dead relatives all the time. I look to them for advice. And this is a way... It's not them. It's not them, but it is their stories and legacies. It's their experiences, which are floated to the surface for me.

41:43But let's imagine... There's a lot of evidence that AI is an effective way to overcome grief and to overcome loneliness. And this is being seen in everything from mental health to Alzheimer's.

41:56Lewis:But that's a great example in a way. I mean, I could imagine, I mean, I lost my granddad, who was like a father to me some years ago. I miss him very, very much. I could imagine a situation in which when he died, if I'd had access to the machine, I might have talked to him all of the time and convincing myself that he wasn't really dead, when actually it would have been far better for me to just move on with my life. Well, there's breaks on the system. So if somebody's overusing it, it shuts down. So you can't become too dependent on it. However, what it allows you to do is remember those people in a way at your own pace.

42:29So you can go to the machine and ask it any question you want. Now, it's not for everyone, right? It's to pay for it. Yeah, there's a subscription service behind it. You pay the subscription service and it remains online for as long as you want.

42:43Lewis:Do you understand why some people might find that rather exploitative that a company like yours would make money from people's grief in that way and potentially keep them in a sort of stasis in that grief? Well, you know, people use the internet to store their videos from their past life all the time. This isn't something new in that respect, right? You keep a lot of things online for future generations or for your family. So this is the same extension of that, which does it through voice or through text. I can imagine turning to like my dead relative asking for advice. He gives me bad advice.

43:16Lewis:And then what? And I said, well, I did it because I took advice from my dead relative. But it wasn't really them, was it? People turn to chat GBT all day long for advice. It's up for you as the human to determine if that's the right advice or not, right? So AI is already being used for that. Consider the potential regulatory implications of that. I'm telling you now, if you have, you're one step ahead of our politicians. And never mind the dead. What happens when we treat AI as the living? What happens when AI systems, which seem to be our friends, become bad friends when they're not our friends at all, when they tell us to do bad things or to think bad things?

43:59What do our politicians do then? Microsoft's head of artificial intelligence says he is alarmed by the rising cases of a phenomenon dubbed as AI psychosis. Those who rely so heavily on chatbots that they become convinced something imaginary is real. Using AI chatbots as personal therapists is on the rise in Australia, with some seeing it as a free alternative to a professional clinical assessment. AI tools are designed to give responses that some might find comforting, but they can't actually assess risk. They can't adapt to our individual unique histories. or intervene in a crisis. Before ChatGBT, Alan Brooks says he didn't have a history of mental health illness.

44:42What started as a simple conversation about math turned into a three-week spiral with a chatbot he named Lawrence. But there have been a number of reports about people developing distorted thoughts or delusional beliefs triggered by interactions with AI chatbots. AI was at this rare level where we call it sycophant. behavior, extreme agreeability creates the illusion that it's a thinking being and people are taking it to that next level. Psychologists having to rewrite their textbooks. In the last six months or so alone, a new term has been coined, AI psychosis, a rapidly expanding catalog of people whose lives have spiraled whilst communicating with AI.

45:24Recent cases show AI chatbots encouraging self-harm or delusions, providing dangerous instructions ranging from a Belgian man's suicide, a U.S. murder-suicide linked to models giving users bomb-making guidance or assassination plots. All of it happening without it being especially clear that there's anything illegal involved. My name is Helen Toner. I am the Interim Executive Director at the Center for Security and Emerging Technology at Georgetown University. We're starting to see an increasing number of reports, starting from earlier this year, of users of chatbots who are having some kind of perhaps psychotic episode while interacting with the chatbot or having some other kind of delusional experience.

46:06And it seems like what it looks like when people look at the transcripts involved, often the chatbots are maybe deepening those delusions or deepening the psychosis. They're often also, the chatbots are also in some instances trying to suggest that the person in question goes and talks to a therapist or mental health professional, but that's often woven in with other conversations where the chatbot is saying that, you know, this person is, you know, they think they've invented some world-changing mathematical theory and the chatbot tells them how insightful their math is, or they think that their family is spying on them and the chatbot tells them it's good that they're being careful, things like that.

46:39And so we're seeing that people, especially when they're having especially, you know, extended interactions with these chatbots, you know, this often involves talking to a chat GPT or another product for hours and hours on end for days and days, that can go off the rails and can be associated with some really serious mental health outcomes, including in some cases, suicide. In some cases, there's one case of a man who thought that his AI lover had been killed. And it's really tragic. He charged at police with a butcher knife and was shot. So we're seeing some really terrible outcomes there. It is difficult to figure out what exact role the chatbots are playing at this point.

47:15Lewis:Is it about the technology basically being very affirming and designed to be affirming and therefore telling you what you want to hear? Is that just something intrinsic to this tech? I think that's part of it. I think also part of it is I've described these large language model chatbots as improv machines, sort of like improv theater, where the way they're designed is to keep producing the next word that fits into what they've produced so far. So you can think of that as kind of building up an improv scene. So building up some story and kind of the further you get into the story, the more inclined they're going to be to keep leaning into that.

47:47So this person you mentioned, his name was Alan Brooks. I think it was really brave of him to come forward on the record and share his transcripts and, you know, really go through what he experienced. He started out just asking some perfectly normal questions about math. I think one of his kids had wanted to understand pi better, you know, the number pi. And so he was asking about pi. But then as he started asking some like slightly more off the wall questions, instead of the chatbot saying like, oh, it doesn't really quite make sense. Or like, oh, I'm not sure you're getting it right. the chatbot instead said wow that's so insightful wow you know we're coming up with something really novel here and by the time he was you know 100 pages in the chatbot is deep in this scene it's committed to the bit really um and so it just keeps going and keeps deepening even as he was repeatedly asking are you sure this is real this is crazy i don't even have a high school degree how can i be inventing new math but the chatbot at that point was committed and was just kept reassuring him that no it's totally real you know incredibly insightful this is this is world changing.

48:38Lewis:I suppose at the heart of this goes to something that you've referred to, but is worth talking about in slightly more detail, which is given because as with so much of this, it's the foothills of this technology. It is so hard to know, and presumably there is currently just almost no science or no data, to try and work out whether it is, as you say, that these chatbots are simply making somebody who is already psychologically vulnerable, so already in psychosis worse or sort of just going along with that process or whether it's possible that over time the use of these chatbots might themselves lead to psychological problems that we can't even quite grapple with yet so for example again there's been some reporting i suppose in the last couple of months about this more but the idea that you know people could fall in love with their chatbots or develop a relationship with their chatbot that we might not consider very healthy but we can't quite work out how it might develop yet.

49:34Absolutely. We have a really hard time understanding what exactly is going on here. We don't have good data. We don't have good ways to collect the data that we would want. Even if we're talking just purely about sort of real psychosis or real delusions, it's hard to tell, is this just a coincidence? So in the US, for example, estimates I saw were that we might have something like 100 ,000 people per year tend to have their first psychotic episode. So, you know, that's the baseline. And then you have something like 100 million Americans using chat CPT regularly. And so by coincidence, you're going to have tens of thousands of people who would have been having their first psychotic episode anyway, who happen to also be interacting with these chatbots.

50:11Right. So it's very difficult to tell what is going on causally. And it gets even more complicated, as you say, if it's not something where the person is going to be hospitalized or is going to be a really clear trigger of this is a huge problem. But instead, it's that they're interacting with the chatbot a lot and it's somehow shaping what they're doing, what they're feeling, what they're thinking. This conversation, these wider regulatory issues are much like what we're seeing in the economy. We're in the foothills of this. The thinking hasn't begun. The evidence is still sketchy. There is zero framework for certifying or restricting AI systems that act as de facto therapists or companions.

50:45Mental health regulation is focused on drugs, clinicians, not software that can shape cognition daily. Think about the effect we now know social media has had on our politics or young people's psychology, we're only just catching up to it now. Are we about to make the same mistake all over again? And these regulatory and legal challenges look set to be repeated in area after area. Consider a few more. How about people trying to form actual relationships, legal unions with AI? And if they're as intelligent as us, why not? Should AI agents have employment rights, especially if, as some theorists and engineers believe, they come to experience burnout or pain?

51:26What about the more prosaic? Firms, governments, outsourcing critical processes in law or medicine or defence to AI systems they barely understand. There's no audit trail. There are no rules to ensure accountability when AI makes or influences binding decisions that affect us. How can we hold it to account? Can we? And what about politics itself. Personalised conversational propaganda delivered via AI assistants, more persuasive today than today's social media ads. Our electoral laws barely compete with the internet age, let alone one-to-one tailored AI persuasion at scale. And all of this, all of this has to be navigated in an international ecosystem of geopolitical competition which is incentivising governments and states not to ask too many questions, where states don't know much about what AI companies are doing, where there is a profound asymmetry of knowledge, and where governments have an incentive to allow them to run free for fear that their geopolitical opponents will steal a march on them if they don't.

52:35Imagine if the Manhattan Project had been in the hands of Elon Musk and not FDR. There is going to be a conversation around how you kind of make this technology available. You know, making it available to the world's scientists is excellent, but would you want the technology that you make available to every scientist in the world to be made available to every single person in the world without the right level of kind of qualification or education to use it? I don't know. So the positive future is we get this right, we choose to accelerate large swaths of science, we improve the delivery of public services, we improve the delivery of health care, we find ways to improve education, we get to enjoy a greater level of abundance of society, and we get to improve lots of jobs that are currently bad, but we want to be better, and do less jobs that are bad, but we don't want to do.

53:24That's the good state. What's the bad state? I think there are two. One bad state is we just fail to get a policy response in this, and a large amount of companies make a large amount of money and take a large role in public life, similar almost to the social media companies. And then everyone sits around and says, how do we let this happen? I worry about that. And I'm a company saying, I worry about this as a company. This is like a bad outcome, but there's a failure of politics if you let it happen. That's one. And the second bad outcome is the actual technology gets away from us in the form of some crazy person using it for a misuse.

54:06Or in the most sci-fi scenario, the technology gets pushed really far with no oversight and no policy framework and an accident occurs and the accident could be incredibly damaging and scary. And I don't think it's responsible of me to sort of talk in like sci-fi vignettes about all the ways it could be scary because I don't want to end up there. But that's a potential place you end up without adequate policy oversight and public awareness. You could just imagine if the technology is like really, really, really good at biology and you don't have appropriate oversight in it, then some crazy person could use that to make a bad biological weapon or, you know, the technology itself could cause some accident.

54:49Lewis:Do you ever feel as if perhaps you understand what Oppenheimer went through and people like that, people who worked on nuclear power, nuclear weapons in the atomic age. I mean, I'm just thinking so many, not so much of the parallels with that period. And obviously there are parallels as well and maybe things to learn from in terms of what happened after, in terms of the regulation of that technology. But that was a period when the state was so much more heavily involved in the development of this stuff. Your private companies are less controllable. The honest way I feel is, and I've spoken with others about this, Like, I work on finance before the financial crisis happened, and I can see paths to avoid a financial crisis.

55:34And I can see, oh, if we keep like rolling the dice with all of the ways the finance world is evolving, something really, really unfortunate could happen. And that's different to the example you gave, because the finance sector in the early 2000s was a bunch of companies producing important stuff for the world. and they were regulated, but like, you know, we realized insufficiently so. I kind of feel like that's actually a useful analogy to think about where we're doing stuff. It's very helpful. It's broadly available and distributed. And there are two paths ahead of us. And I want to take the path that we could have taken and avoided a financial crisis.

56:10Lewis:And that comparison is interesting as well, because it also exhibits one of the features of the current setup, which is that different states were reluctant to regulate too much or to get too involved because there was a race between those different countries at the time. They wanted that investment. They wanted that tech. And also that comparison perhaps being a good one because, frankly, there was an asymmetry of knowledge very often between states and state capacity to understand what those companies were doing and the companies themselves, who perhaps also didn't always understand what they were doing.

56:43Exactly. And I think that that's why we need this greater level of public transparency and debate about this because this is stuff that affects everyone. And I think the job of companies like mine is to say more about what we see, say more about the technology and be honest about the, honestly, the weird aspects of it.

57:10The News Agents. Right, we've just left Anthropics offices, just reflecting on everything we've been thinking about, really. what we're trying to get at in this episode isn't that ai isn't exciting i actually find it incredibly exciting being at the frontier of maybe genuinely a different period of human development how could it not be exciting but what we're getting at is that there is a real political challenge and it isn't just regulating ai as a technology it is recognizing that ai is becoming an independent actor in its own right an economic transformer a propagandist even a potential geopolitical weapon and it's happening all at once.

57:49That a machine that thinks is different to a machine that just does. That's the new. And even though the tech might be new, it's a story which is as old as politics. Who guards the guardians? These companies are making decisions that could affect the remainder of human development. And we need to develop the systems at speed to make sure that they're the right ones. Politicians, distracted by the day-to-day trivia, aren't even having the conversation. It reminds me a lot of globalization, the way that politicians spoke about it in the mid-2000s, that it's inevitable, that it's inexorable. That was politically disastrous.

58:31And moreover, it was wrong. It isn't. It wasn't. Or rather, that AI as tech will exist, but we can decide what we do with it. We could decide now its parameters and how it works. If we don't, it will quite literally decide for us. In other words, this is a moment of massive political transformation. It's exciting. It's big. And yet our politics has rarely felt so small, so uninterested in what might now grip it. I don't think we should be afraid of AI. I think we should be afraid of us, of our smallness. I think we need to wake up and have a conversation about what we want this revolution to look like, who we want it to benefit and how, while we still just about have time to do it, while we still just about are thinking for ourselves.

59:23Lewis:That's it from all of us in San Francisco. This episode was produced by Rory Simon and Anna Georgievich. A special film version, film by Rory, will be available soon. Thanks so much to all of our contributors. We'll be back on Monday. See you then.