The Munk Debates Podcast - Munk Dialogue with Bruce Schneier: AI and democracy

Episode Date: June 14, 2023

Over the past few months we’ve heard many warnings about the dangers of Artificial Intelligence. But are there some positive aspects about this emerging technology that are being overlooked? On this episode, we’re joined by internationally renowned security technologist Bruce Schneier who argues that dangers associated with Artificial Intelligence are being overblown, and that chatbots like ChatGPT could actually strengthen democracy and restore trust in our governing institutions.   The host of the Munk Debates is Rudyard Griffiths - @rudyardg.   Tweet your comments about this episode to @munkdebate or comment on our Facebook page https://www.facebook.com/munkdebates/ To sign up for a weekly email reminder for this podcast, send an email to [e-mail].   To support civil and substantive debate on the big questions of the day, consider becoming a Munk Member at https://munkdebates.com/membership Members receive access to our 10+ year library of great debates in HD video, a free Munk Debates book, newsletter and ticketing privileges at our live events. This podcast is a project of the Munk Debates, a Canadian charitable organization dedicated to fostering civil and substantive public dialogue - https://munkdebates.com/ Senior Producer: Ricki Gurwitz Editor: Kieran LynchBecome a Munk Donor ($50 annually) to get 72-hour advanced access to the full length editions of Friday Focus and Munk Dialogues. Go to www.munkdebates.com to sign up. Hosted on Acast. See acast.com/privacy for more information.

When you're a journalist and people don't trust you, it's always your fault. These people need to be represented. They are Canadian. They deserve to have a voice and a seat at the table. It is time to go back to the office, and the time is now. Russia had reasons to be concerned. They had reasons to be fearful. We're at an absolute turning point in reproduction. This is the problem with realism. They just treat all countries the same. They don't distinguish between dictatorships and democracies. Hello, Monk listeners. Roger Griffiths here, your host and moderator.

Welcome to this, our continuing conversations called the Monk Dialogues. These are in-depth questions and answers with some of the world's sharpest minds and brightest thinkers on each monk dialogue. We go deep into the big issues that are transforming our world and shaping our future. Today we're talking about the big issue that is part of that future, that will affect our jobs, our social interactions, the democratic process. perhaps our very existence on this planet. Of course, I'm talking about artificial intelligence. In a lead up to our big public debate on AI, taking place in Toronto on June 22nd,

we're hosting a series of monk dialogues with experts in the field of AI to explain to us how this fast technology is evolving and how it could affect us in our day-to-day lives. My worst fears are that we cause significant. We, the field, the technology, the industry, cause significant harm to the world. I think that could happen in a lot of different ways. I think if this technology goes wrong, it can go quite wrong. Over the past few months, we've heard many warnings, like the one from Sam Altman,

the CEO of OpenAI, the company behind Chat, GPT, about the dangers of AI. But on this monk dialogue, we want to take a different direction. give you a different take on this emerging technology. There is a debate here, a controversy, a live conversation about just how transformative will this technology be, how negative could its effects be? Well, our guest today, Bruce Schneier, is an internationally renowned security technologist, a fellow at Harvard's prestigious Kennedy School. And he views, thinks that artificial intelligence could actually strengthen our democracy and restore Trump. in our governing institutions. Bruce, welcome to the Monk Dialogues.

Thanks for having me. Looking forward to this conversation, it's one of a series of discussions that we're having with renowned experts in the lead-up to our June 22nd Monk debate on AI. I wanted to focus with you on the kind of social, democratic, societal dimension of AI's impact.

And one of the reason we kind of sought you out is for your kind of balanced and considered take on the potential impacts of AI, you know, outside of the lab in society, where I think a lot of people are scratching their heads and trying to work through to some kind of understanding of what those effects could be. Let's begin, Bruce, by just having you set out a time frame for us. Is this something that people should be expecting to have impacts in the near term? Are we still talking potentially years away? What are you looking for to try to understand when AI is going to be pushing against a lot of our democratic institutions,

beliefs, assumptions about how and why our society should work. So this is, I think, the most important thing I can say here is that no one has any idea. If you interviewed people a year ago, they would not have thought this year would have happened the way it would happen. If you enter them six months ago, they would not have predicted today. and anything we say today might not be true in three months, six months, a year, two years, five years. No one has any idea. We can speculate, but we're really just guessing.

We can think about what might be possible. We're just guessing there, too, but time frame definitely. Don't believe anybody who gives you a time frame. What do your instincts tell you? I mean, again, what we've all been informed is that at least amongst scientists like Jeffrey Hinton and other kind of pioneers in the field of neural networks and large language learning models that things that they had expected to see a number of years from now, possibly decades, are happening in the short term. In some cases, the breakthroughs that they're seeing, at least in terms of the technology and how

it's behaving is now. Can we build off those assumptions to understand that where and when these effects are going to start to be felt? We might be able to, or they might be wrong. I mean, we actually don't know if we are at the knee of an exponential growth curve or it's a step function. I went to a lecture two days ago where the researcher talked about we've reached the limit of what LLMs can do, and that is a dead end. Other researchers just say different things. We truly don't know. And you can probably get people to make predictions, but they're really just guessing. And that's important to understand. Things that we think are easy end up being hard. Things we think are hard end up being easy. Remember, what was it, 2014 when AlphaGo beat a world master in Go? When I was in school,

I learned that would never happen. And a company did it in their spare time because kind of they were bored. LLMs are an emergent property. The things that it's doing, researchers are surprised at. Things it can't do, researchers are surprised at. Really? I mean, I get it. As a non-scientist, you want to know what the future is going to hold. But as a scientist, they actually can't tell you. Well, let's think a little bit about what these technologies then are and what they're not. I think a lot of people hear this phrase, machine intelligence, artificial intelligence. And because we are who we are, we naturally equate that with somehow with human intelligence.

I'd be curious as to your thoughts about how you're starting to think about ways that machine intelligence is different than human intelligence and how we might go about trying to understand a different form of intelligence and is intelligence really the right way to describe what these machines are doing, the products that they're producing, or are we anthropomorphizing this in some way that's confusing the whole? whole issue. I mean, you've got a lot of truth in what you're saying, and language is the problem here. When we think about AI or talk about AI, we get our cues from the movies from television. We think about Star Trek, we think about Star Wars, right? We think about the different movies we've seen, and that's what we think AI and robotics is. And when we think of artificial intelligence, we think of that general purpose intelligence, right? That's C3PO, that data from Star Trek. This is nothing like that.

And this is really just large language models, which is just a subset of what the research of AI is. We also misconstrue language for intelligence. The fact that LLMs can say stuff, can produce language, has us predisposed to believe that it is thinking, that it is somehow conscious. And none of that is true. So the language is bad. both the language we use and the language that is used to describe it that we hear and the language that is produced. So, yes, this is nothing like thinking.

This is not conscious. This is not intelligence. I've heard described as the world's most sophisticated parrot. It just produces words statistically. And we think it's intelligence. And that's large language models. There are other avenues of research of AI that are not. as sexy as public that might be better in the long run. And these are very specialized. They're not

general AI, which is what we think of on the movies and television robots, so of those androids. They're special for a task. And they can be very good at the task, whether it's reading chest x-rays or producing language or recognizing photographs, but that doesn't transfer on the domains. Got it. That's helpful. Let's try to just help our audience understand better what these large language models are doing. My kind of basic understanding is that it's really, in a sense, raw computational power.

It's really powerful microprocessors. It's a whole lot of them all at once, taking tons of data inputs, many of which have been cleaned up and kind of polished by human hands before. going into the system as data, and then it's making probabilistic predictions about what follows X and what comes after Y. Again, there's nothing intelligent in the human sense of thinking up an original theory or making a jump or leap of logic. This is based on, in a sense, mathematical calculations run at scale.

on a ton of data and producing outcomes that, I guess, probabilistically, are accurate to the extent to which these systems produce results that have utility to us. That's the indication of their success. So that's mostly right and shows you how fast this is moving. A month ago, I would have said that's all right. So let's go with the stuff that is correct first. What these models do is they ingest language, all sorts of it, whether on the internet, books, posts, dialogue, everything, and in all languages. And it uses that input to basically create a model of kind of how language works, of what comes next, right? To be or not to, you know, the answer, B. That's an easy one. Some of them are harder than that. But with you can get massive amounts of language,

you can produce a model that will know what comes next. And then there are front ends that we use. So GPT is the one a lot of us have played with. That's the model. Chat GPT is that model attached to a front end that allows us to converse with it. That takes our queries and turns it into input, the model understands, and takes the model's output

and turns that into answers that we understand. removing hate speech or whatever guardrails they put in. So what's different is the size. A month ago, I would have agreed with you. It takes a massive amount of computation and a massive amount of data. And only the biggest companies can do it. And they spend millions creating their models. Turns out that's not true.

Facebook released their model to researchers. It got leaked to the public. And thousands of researchers are working on it, tweaking it, playing with it. And one of the things we learned from them is that small models, models that you can compute overnight on your laptop, which take maybe $100 of cloud processing power, are almost as good, much smaller, much faster,

easier to iterate, and you can put things on top of them. So that only the very powerful can do it has changed in the past month. but basically you have that right. It is a black box of statistical weights that produces the next word given the words in a sequence. So that's fascinating, Bruce. You're saying in a sense that this technology could be scalable, maybe in a way that we hadn't assumed, as you say, it's almost ridiculous to be having this conversation. But, you know, four or so six weeks ago, or just in the last. last few months. What do you think the implications of that if this stuff is scalable? Because now I want to

move on to your kind of insights about democracy and society because there's been a lot of hand-wringing of late about LLMs and about their impacts on society and democracy. The 2024 U.S. election cycle is coming up fast. What happens here if this stuff, instead of being kind of heavy machinery, like industrial strength, AI is in fact something that you and I, anyone potentially could access and start to use. What's the good and the bad and the ugly that would flow from a scenario like that? Let's talk about the good. And this is before we get to democracy. The good is that it's not in the hands of a for-profit corporation whose interests might not be the same as yours. If you think about the way we want to use LLMs and this near-term AI, we want to use them as assistance to help us

make reservations or figure out a movie to watch or do all sorts of things on our behalf. And people are thinking about this, right? Can an AI process my email and tell me what's important? Can it summarize meetings? Can it compose meaningful messages? Now, if it's going to do all that, it needs to know a lot about us, about us as individuals, about us as businesses. And it's going to be an interesting question is, do you want to let a for-profit monopoly eavesdrop on all that? I mean, we kind of already do that. Google is intimately knowledgeable in us because we don't lie to our search engine. Facebook is intimately knowledgeable about us because we don't lie to our friends and our lovers and our associates.

And those companies have inserted themselves into those very intimate human actions in order to sell us stuff, right? In order to eavesdrop on us, learn about us, and manipulate us. This is going to be much worse than that. This kind of power that we would like to give our digital assistance is incredibly intimate. You wanted to know your emotions so that it can respond to you in a way that mirrors your emotions. That's more effective. It wants, it's going to want to know your vulnerabilities, your strengths, your hopes, your dreams, your fears. And if these models become much more democratized, if these smaller models can run on your laptop, it can be yours.

It doesn't have to be at the benevolence of a for-profit corporation. Open AI let us use chat GPT because it wanted to steal all our. training. It didn't do it because it was benevolent. It made money from all of our actions. And that's going to be very dangerous. So I like these open democratic models because it means we could all run our own. And that will be much safer and better. So that's one of the good things. You know, people talk a lot about misinformation, right? The Republican National Committee released that AI-generated video of a post-Biden dystopia. You know, those are all concerns,

but honestly, they were concerns in 2016, in 2018, in 2020. The difference had to have an AI do it. I don't think it's going to be a difference in kind, might be a small difference in degree. You know, already everybody's able to post memes and fake stuff and lots of fake images. I don't know if that's really that much of a concern. I think that is more of a moral panic.

I am concerned about artificial people that for rulemaking, you know, if the U.S. government agency produces a rule, it goes up on a website and we can issue comments. Fake comments are a thing, and they have been, and this can supercharge fake comments. So fake comments to rulemaking, fake comments to a newspaper article, fake comments on Facebook, on Twitter. You know, we're already having trouble with bots. This makes them more effective. Again, I'm not sure how much more effective, but it's something to watch. So there's a good thing and a bad thing.

And there are more. We can go with good and bad all day. And a lot of it is we have to see. But I tend to be more optimistic and pessimistic, really. Next week, June 22nd, the Monk debate on AI is taking place in Toronto, Canada, in front of a live audience of 3,000 people. you can join as a virtual participant live in this debate via our online web stream of the entire debate proceedings. You'll have the opportunity to watch four of the world's leading experts on AI, debate the motion, be it resolved.

AI research and development poses an existential risk. To watch the debate online, live and archived, you can do so by becoming a monk don't. for as little as $25 a year. You can do that right now on our website, www.w monkdebates.com. Simply look for the donate button in the top right navigation. Become a donor and we'll send you a link next week to watch the debate live in the comfort of your home.

Thanks. Now, back to our program. Let me just try to, for a moment, just stick a little bit more on the glass half empty because I think that is in some ways where the, the finger is on the debate when it comes to the scale of public opinion. And again, let's hope that that can shift and change as we become more knowledgeable and insightful about this technology. But one of the things that I think alarms people about these large language models is, as you say, they're parrot-like abilities, their ability to mimic.

Right now, they're doing that in the context largely of text, but as we're already seeing pretty interesting experimentations with audio, converting people's voices into a digital, automated digital conversation and response. You know, video is starting to be manufactured by some of these bots out of whole cloth, not sourcing separate images around the internet, but literally assembling pixels into videos that are queued by a user. At the end of the day, Bruce, isn't there a fundamental challenge that we only have our senses, sight, hearing. We don't use taste too much when it comes to our democratic engagement in the public square.

Maybe we can start doing that more. But I'm just thinking of our basic senses that in some ways all we have is what we have. And are you concerned, Bruce, at the extent to which we can be fooled, that these technologies, not because they're intelligent in any human sense of the understanding of that word, but that they're such powerful mimics and the ability for bad actors to get this technology, you say maybe on their laptop,

manufacturing this stuff at scale, you know, at speeds approaching that of a digital processor, I don't know, Bruce, that makes me worried. So digital manipulation is as old as digital photographs. I mean, this isn't new. We've had fake videos and elections in the past. I mean, these things, are worrisome, but they're not new. And I think that should give you solace. Yes, you only have your

senses. And that's been true for thousands of years. Certainly it's been true for the hundred years we've had these technologies. Remember the, right, Orson Wells and the War of the Worlds, right? We all learn this. People turn on the radio and think there's an alien invasion. They're fooled. Remember the Nancy Pelosi video from what, eight years ago? And this wasn't even a deep fake. This was kind of a cheap fake where someone edited a bunch of videos to make her look drunk. You know, these things are already happening. And yes, this will give that ability in more hands. But I think that's probably better

because we're going to all learn not to trust a random video we get on the internet. Because two thirds of them will be faked. And it's only when it's hard that when you see a fake video, you're fooled. Right? We are, we're going to become, discerning because we have to be but this is not new we've had fake audio we've had fake video we've had fake

newspapers on the web have we ever have we ever have we ever had a presidential candidate phone us up and have a half an hour conversation with us and a million other people simultaneously at the same time in a population where statistics would say half of voters have an IQ under 100? The point is, Bruce, that is, that is new, isn't it? I guess I mean, this feels like a moral panic. It's not that new. What is new is a candidate being able to engage with voters at a very meaningful level in parallel. So here's a positive. I mean, right now, if you're a candidate and you want to reach a voter, you have to make a video they all watch. You just send them an email. You need to engage in both. You need to engage in both.

These large language models are going to revolutionize learning, that instead of reading a book or watching a video, you can engage interactively with an AI that will teach you something. And it can teach you about politics. It can teach you an issue or it could teach you about a candidate. And the ability for a candidate, and let's assume good intentions here to start, then we can look at the dystopia, can go into everybody's living room and talk to them, right, with a face, with a voice, with their opinions, with their issues, and engage, I think it's really powerful. And at the back end, we're going to have the all the summarization. We'll talk about that later. It's another good thing. Right. Now, this requires, you know, politicians to care about voters and for a lot of things to change, I think, about the way the United States

politics works. But in theory, this is incredibly powerful and a force for good. So, yeah, I mean, people are going to write fake videos and fake audio just like they did last time, the time before, the time before that. And yes, they'll be better. And we'll get better at not trusting them. I just don't see that as the – that's not the thing I'm worried about. It really isn't. We can come up with all sorts of stories, but it's not what I'm worried about. But, Bruce, what if you, let's say, take a Trump chatbot that, as I say, is phoning up tens of thousands of people, Simon.

to have conversations with them. And in the back end of that chat bot, you install Facebook's 25 to 30,000 unique data points on each and every one of us. And that chatbot says to those tens of thousands of voters that it's speaking to simultaneously, exactly the things that those people want to hear based on the deep data that, as you say, you pointed out accurately, we've shared ubiquitously, unconsciously, almost, with Google, Facebook and all these large platforms. I mean, aren't you really looking at a new age of persuasion that is far more powerful than anything that we've seen in our democratic process or in the public square to date? I mean, this is a step change. Yeah, not really. I mean, I think you're buying

Facebook's marketing bullshit there. I know they have lots of data, but it turns out most of that is not relevant. And they want you to think it is because that way they charge a lot of money, but, you know, that's their issue. One of the reason you don't see a lot of data and how effective Facebook ads are because they don't want to show them to you. I mean, yes, right, you can have that. You can have targeted marketing. And the targeted marketing will be better than the targeted marketing today, which is still pretty good.

You know, here's where we need good laws to make sure. these things don't happen. I mean, the reason that could occur is because the transparency laws are lousy. If we have to see a database of what political ads look like, candidates can't get away with that. So, yes, I mean, that is a thing and it's a little worry. But again, it's not on my top 10 worry list. It's, I mean, it's easy to write a scare story. It's hard to know how much more effective it's going to be than what we have today. Another thought here. If you have a scenario where these types of technologies are out there,

let's say hopefully in your model, which I like, one of kind of democratization and you've got access to them outside of some set of technological fiefdoms in Silicon Valley. And people are using these for, you know, all kinds of different, you know, purposes and objectives, what would be some of the positive outcomes that you could see people engineering when it comes to democracy in our society? So you've talked a little bit about education, like I kind of knew Socratic return to the Socratic dialogue, I guess. But what are some of the practical ways that you think this technology could be used to strengthen community,

engage people, inform them, like in the hands of good actors, because good actors will get a hold of this stuff. What can they do? So one of the things that I think will be very powerful is these models' ability to summarize and make things visible. So right now, if you write a letter to a public official or you write a comment in rulemaking,

they'll get thousands, tens of thousands of those, nobody reads them all. I mean, really? They're put in categories for and against. They're made smaller, right? You're bucketed. This is something where a large language model can take all of that input and produce a much more nuanced representation

of what people are thinking and saying. And that's something we don't talk about a lot, but that's something these models are very good at. And I think we'll have very beneficial effects. You know, right now, we tend to, you know, be in so few pigeonholes because that's all that can process us. And the ability of these models to process more fully is going to be very powerful. And maybe just give us an example. I mean, would this be in the context of, you know, elected officials seeking, I don't know, advice on a new road or a piece of public works? How,

how would that work practically? It's two things. It's elected officials, as you say, and people write to elected officials all the time about issues that matter to them. But it's also public rulemaking. All rulemaking bodies, when they come up with a draft rule, there's a comment period. Now, it's usually corporations and lobbyists that take advantage of that, but people do as well. Net neutrality was famously something that an FCC rulemaking. which got tens of thousands of people submitting comments

because it's something that those of us use the internet cared about. And there you can use these models to really make visible what people think in a way that human summarization can't. And that is extraordinarily powerful. And I think it'll be a great benefit. Could there also be Bruce a sense that, again, maybe it's not, hopefully it's not Google or Facebook that we're giving this data to, but let's say we have, in the short order, these things on our laptops

that we have some control over. Again, I really like that idea. We can upload all of our emails and train these models up that we own. Wouldn't those models then start to become very interesting, almost authentic representations of us? And the whole idea of having a digital self that was somehow imbuted with authenticity would suggest maybe that that digital self would then have some kind of standing in the public square. I mean, could you envision a future where we start to, I mean, outsource, it applies a negative connotation, but put in substitution or in service of our own civic objectives, these digital selves that will be harder to ignore because they're not just an email. They're a reflection of potentially hundreds of thousands of data points, of hours of

conversations, of emails, of books and podcasts that you and I have done together, this will be me in a way that no other digital product to date has ever been me. I mean, it's good and bad there, right? I would love to have an AI avatar that argues with my cable company for the refund I deserve. I mean, that would be awesome, right, for me not to have to spend half an hour on the phone and have an AI do it. So having an AI be my representation, I think is very very good.

valuable in some cases. In democracy, I think you're right. It's something we would have to worry about. I mean, the real work of democracy isn't doing it. If we outsource our beliefs, we outsource our civic responsibility to an AI, I think that makes our democracy worse, even worse than it is right now. And now, you know, I mean, we're being very theoretical here because, you know, the United States just, you know, democracy doesn't work this way anymore. But it's, you know, it's, you know, supposed to be, you know, citizens thinking about issues and coming up with their opinions about them. And you don't want to outsource that. So now it's going to be a matter of what it is we do outsource and what it is we don't. You know, would you like an avatar to read your email,

respond to the dumb stuff and tell you what's important? Or maybe you tell your email avatar. Yes, I want to, you know, respond this way, accept this invitation, tell them I'll be there at and it crafts the email. You know, I think we're going to see that kind of world where these avatars are going to do things in our stead in ways that make our life easier. It'll be like everybody having an assistant. Now, and again, this isn't a change, right?

The wealthy have already had this. We think about GPT as writing people's essays and opinions and speeches. The wealthy have always had speechwriters and ghost writers and editors. Right, that'll make that more available to everybody. So a lot of these things we're looking at are not new. They're just been out of reach of most people. Yeah. Lobbyists, lawyers, accountants. If you're wealthy, you have someone arguing,

you know, on your behalf in a court all the time. Just as we moved to the conclusion of a program, we had the U.S. Surgeon General come out in the last month or so and come out with a statement about the deleterious effects of social media on teens, especially teen girls and women, seemed like it took us maybe more than a decade to figure that out. What if any lessons are there from our rollout and adoption and maybe the lack of thought around social media and how we approach that versus AI and what we could or should do? Is there a parallel here, Bruce, between, frankly, the failure maybe to address and understand some of the social and democratic political harms of social media.

And I don't know, are we going to do that all again with AI? Are we going to learn from our mistakes? We're probably going to do it all again. But there is a lesson. And the lesson is that the market is ill-suited for these decisions, right? That building an internet, building social media for the near-term financial benefit of a bunch of tech billionaires is a dumb way to organize society. and we would do much better if we were more deliberate about the technologies we developed and deployed and the way we use them.

But that's not the way we think. Surely that's not the way the U.S. thinks. So likely we're going to make the same mistakes all again. Let's say we didn't want to make those mistakes. What should we do? I mean, is this a case where it's our institutions that we have to look to, to solve this? As you say, they didn't do a great job last time. They didn't do a great job because we don't let them do a good job.

There's this belief that government can't do anything and therefore should do nothing. And yes, there's a role for government here. There's a role for us as citizens instead of us as consumers. Now, the place to look is Europe. I mean, you know, still not perfect, but the EU is much better at regulating these technologies. And they're working on an AI regulation right now. It's not the best, but it's certainly better than anything else we've seen. and more deliberate thought there, I think would be beneficial.

I mean, in general, we regulate technologies when they can kill us, right? Airplanes, pharmaceuticals. AI is going to reach that point pretty soon, right? It'll be driving our cars. It'll be controlling our medical devices. You get it wrong and it can kill us. So given that, you just don't want unfurricular. fettered innovation without some kind of review and limitation.

We don't allow that in aircraft design. We don't allow that in pharmaceuticals. Because if you get it wrong, people die. We don't allow that when you get on an elevator. I mean, it's pretty, it's pretty ubiquitous. You're right, Bruce. Final comment from you, you're optimistic in a way, I sense, Bruce. Through all this, you think that we're going to muddle, we're going to make mistakes.

but your feeling here is that there is not an existential risk that we're courting with this technology. Is that correct? You know, I think there is. I think you're right and we will muddle because that's what we do as society. We don't get it right the first time you muddle through. The danger, and it's going to come up in the debate, is that the muddling process becomes risky when getting it wrong could be fatal. And that's the worry, that we just don't have the time and space to do our normal

sort of human muddling. So that's the risk. I tend to be optimistic, but I think there are risks. Okay. Well, Bruce, thank you for coming on the monk dialogues today. I love some of your contrarian opinions here. It really helps move the discussion and the debate forward. Let's do this again, Sue. Yeah, thank you. And let's do it. Well, that wraps up today's monk dialogue. I want to thank our guest, Bruce Schneier. You've certainly given us a lot to think about if you have any feedback or reflections on what you've just heard or any of our other podcast, please send us an email to podcast at monkdebates.com. Also a reminder that our big public debate on AI is approaching fast.

June 22nd will be joined by four of the world's top AI experts to debate the motion, be it resolved, AI research and development poses an existential threat. Tickets to this event are sold out, but we will have a live stream of the event available to monk donors. Details can be found at Triple W. monk debates.com. Thank you for lending your time and attention to our efforts to bring back the art of civil and substantive conversation, one dialogue at a time. I'm your host and moderator, Rudyard Griffiths. The Monk Debates are a project of the Aure and Peter and Melanie Monk

charitable foundations. Rudyard Griffiths and Ricky Gerwitz are the producers. Be sure to download and subscribe wherever you get your podcasts. And if you like us, feel free to give us a five-star rating. Thank you again for listening.