The Munk Debates Podcast - Friday Focus: Debating Artificial Intelligence

Episode Date: May 12, 2023

Friday Focus provides listeners with a focused, half-hour masterclass on the big issues, events and trends driving the news and current events. The show features Janice Gross Stein, the founding director of the Munk School of Global Affairs and bestselling author, in conversation with Rudyard Griffiths, Chair and moderator of the Munk Debates.  The following is a sample of the Munk Debates’ weekly current affairs podcast, Friday Focus. On this week’s edition of the Friday Focus podcast, Janice and Rudyard spend the show talking about the upcoming June 22 Munk Debate on artificial intelligence at Roy Thomson Hall in downtown Toronto. What are the major fault lines that are likely to drive the debate between MIT’s Max Tegmark and Yann LeCun, Chief AI scientist for Meta? How could the rise of thinking machines constitute an existential threat to humanity? What are the policies or practices that industry and government could adopt to ensure AI safety? Janice and Rudyard dissect the big issues and ideas that are making the rise of AI one of the most important issues of our time. To access full-length editions of the Friday Focus podcast, consider becoming a donor to the Munk Debates for as little as $25 annually, or $.50 per episode. Canadian donors receive a charitable tax receipt. This podcast is a project of the Munk Debates, a Canadian charitable organization dedicated to fostering civil and substantive public dialogue. More information at www.munkdebates.com.Become 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.

The following is a complimentary excerpt of this week's edition of the Friday Focus podcast by The Monk Debates. To access full-length editions of each and every episode, along with all kinds of great additional benefits and perks, become a donor to the Monk debates. You can do that for as little as $25 a year, and you'll receive each and every year 50 Friday Focus episodes at full length. It's all available right now on our website. in just a few simple clicks. Triple W. The Monk Debates.com. Look for the Friday Focus option in our navigation bar, the top right of the website. Make your donation, and we will send you each and every Friday a link to listen to the

full-length edition of this program. Thanks in advance for your generous contribution. Rudyard Griffiths here, the executive director of the Monk Debates. Welcome to this, our regular Friday Focus podcast, as I am each and every one. week, joined by Janice Gross Stein, the founding director of the Monk School of Global Affairs, an internationally renowned scholar and author. Janice, great to be in discussion with you the 12th of May. Great to be with you, Roger, and boy, what an interesting and scary and exciting topic we're going to talk about today.

Absolutely. So today, Friday the 12th, is when tickets for the monk debate on AI are now on sale to our general membership before the public at large. So this is your last opportunity today and into the weekend to get tickets before this debate sells out. I think, Janice, you know, this could be one of the more important monk debates. AI, I think, has caused, as you said, a lot of, certainly on my part, really unsettled me. I'm confused. I'm worried. I want to wrap my head around this. And what an opportunity on June 22nd with the Spring Monk debate. We have Jan Lecun, who's the chief AI scientist for Meta, formerly Facebook, and consider one of the world's leading thinkers on AI. He's less concerned, Janice, about the existential

threats. And on the other side, Max Tegmark from MIT, who co-authored that big letter earlier this spring, over a thousand scientists, he and Elon Musk's, Max Tagmar and Elon Musk, saying we need a moratorium on the development of this technology. It's such a risk. We have to take a pause. So I'd love to know if you were invited on to the Monk debate stage and you had our resolution for June 22nd in front of you, which is be it resolved, AI research and development is an existential thread. Where would you come down, pro or con? Wow. That's why I have so much trouble debating, Rudyard, because it is an existential risk,

and it is one of the greatest opportunities we've had to make many, many things better. It's both at the same time. That's what makes this such a wicked problem. So I'd have to come down, yes, on the resolution, but I'd be busy saying, yeah, but what a fantastic opportunity. And the pause, isn't, wouldn't do anything. We really have to figure out, and this is so hard. Now we manage those risks because they're enormous, but how we get the incredible opportunities that AI is already created.

So, Janice, one of the reasons I always love these conversations with you is your larger perspective. So let's go back and just for a moment, think about that other big technology, which in your lifetime and my lifetime, in a sense, has what political scientists called dual use. You know, it can do all these fantastic things, but it's also potentially really bad. And that was the Manhattan Project. It was the creation of atomic weapons that could both light up cities with power and destroy them with thermonuclear bombs. is that seems to be the kind of way the debate around AI is shaping up. Do you feel that that analogy holds is dual risk a useful way for monk members to try to

understand how to think about these types of challenges? Yeah, I think it's a great analogy, Roger, and a University of Toronto academic, Jeff Hinton, who is regarded as the father of the version of AI that is created, these large language models chat GPT4, thus causing all the controversy, said the other day, I'm having a Robert Oppenheimer moment. Robert Oppenheimer was one of the fathers of nuclear technology,

and after the bomb was exploded, Hiroshima Nagasaki spent the rest of his life, warning everybody about the dangers of nuclear weapons. But here's one big difference, which makes today scarier. When nuclear technology, when it first became possible to create a nuclear bomb, you had to be a mega wealthy state to do it. That's not true about where we are now with AI. The code that has been released by Open AI and others is so accessible.

that a smart grade 13 student can build code on top of that with the help of the AI and use this technology. So this technology can easily diffuse right down to the individual level throughout our society. And that makes it both much scarier than nuclear weapons, but also what it can do. let me give you one example of what it can do, all right? It can change the whole field of medical diagnostics, likely well, in the next year. Because it can process such large amounts, trillions of data. It can tell your radiologist what, that little thing that you're seeing on your neck, is that a problem or is it not? And it will tell your radiologist that. So instead of the

radiologist being able to look at it and say, well, of the thousand other cases I've reviewed in my life, how does this fit? That radiologist is now going to have trillions of cases and going to get the diagnosis literally in 30 seconds. That's in one tiny field. It's going to change the whole field of of health diagnostics, of health care delivery and how it's delivered, it is going to change. Science, we've already had this huge benefit in protein folding, which was unimaginable. I could go on and on. But the big risk, as we just try to rein in a technology that is going to create a deep fake of you, Roger, with your voice, going to look like you, and is going to run.

a video in which you say, God-awful things to monk members, that they can do it right now, if anybody's motivated. That, yeah, the large language models can do both of those things simultaneously. It is a wickedly hard problem to figure out how to rein this in. It might be our most important monk debate ever. On June 22nd at Roy Thompson Hall will be convening two of the world's leading experts on artificial intelligence to debate the motion, be it resolved. AI research and development poses an existential risk.

Be sure to attend this debate and listen and learn from Max Tagmark and Yan Lecoon, again, top experts who know AI like few other people. They're going to share their thoughts on how AI will transform our workplace, our society, our democracy. We'll cover it all the risks, the rewards, AI at this critical monk debate. For more information, go to our website, triple W. Monk Debates.com. We do have tickets set aside for Monk members, so grab yours now before this event sells out. We'll see you in Toronto on June 22nd for the Monk debate on artificial

intelligence. So on our debate on June 22nd on AI with Yan McLuhan and Max Tagmark, I think one of the things that I hope we explore is, idea of what is an emergent technology and why is that word emergent so important when it comes to AI. One of the things that I think Jeffrey Hinton has mentioned in the last few weeks as he resigned, as you said, from Google to start speaking more freely about his belief in the dangers of AI is the extent to which these programs are beginning to self- iterate. They are beginning to do things in terms of their own coding that we don't necessarily understand what they're doing, why they're doing it. And sometimes even when we go in and look at the code,

we don't understand how it does what it does. Janice, that's the part where I start kind of going a little bit squarely and reaching for like a strong genitone because I don't like the sound of that. Well, your right, that is what has let to scientists calling for moratorium. And Jeff Hinton, who did all this foundational work. And even two years ago said we were a decade away from what you just described, has done 180 degree now and is saying, oh my God, this is going much, much faster. Emergent means, in this case, as you just said, Roger, emerging out of in ways that even the

scientists who are building these models do not understand. They're talking about a black box, right? So what are what are AI, large AI models really doing here? They are doing super sophisticated computing, right? That's really what they're doing. They're doing really sophisticated probability analysis. And they'll tell you which word is likely to follow from the previous word. That we can understand. But something new has gone on just in the last few months, where it's going beyond that

and it is writing code without being told to do so. and it is unclear to the designers of the models. How, and I don't want to use the, I'm going to use the word it because it's not human. The biggest mistake we can make is think these are human, but how these programs are learning to, and going up the ladder of complexity without any human engagement. that's the point that we're at

that people thought was probably 20 years off and we would have a lot of time to figure out how to regulate. So Jeff Hennon says, and let's put this one out there for among members to think about these large language models are not human, but they are smarter, it's already smarter in some ways than almost every human being alive.

because they can do these complex tasks on their own and get better and better and better at it without any human intervention. Yeah, I mean, one way that I've tried to think about it is they're not intelligent in a way that we would understand intelligence, but they have, they demonstrate intelligence, that they show a functionality and an output

that matches, or in some cases, surpasses already, you know, human effort and cognition. You know, the apocalyptic scenario, Janus, as you know, is what they call superintelligence, which is this sudden breakout. And, you know, this has been hypothesized that once they begin to, as you say, become more emergent and self-coding, that there could be a whole bunch of unintended consequences that we don't really understand, that they could suddenly have a step change in their functioning and power that then leads to a moment where they do something that isn't at all

what we intended them to do. That's one of the things I hope in this debate we try to explore, just because the word existential is in the resolution there purposely, right? The resolution on June 22nd, be it resolved, AI research and development is in a sense an existential risk. What's your feeling there? I mean, it's so hard to know, but it seems like as the characteristics of a technology, I think, as you say, maybe nuclear weapons isn't the right analogy. Maybe CRISPR or the manipulation of genes and genes.

editing could be the better analog to try to understand you know the risk and in the question you know gene editing it's you know the equivalent of super intelligence is the creation of a horrible pathogen by human hands that then escapes the lab and causes you know untold death and suffering yeah I think you're right that gene editing is a better analogy to what we're talking about and you know where does this discussion of super intelligence singularity the best way to get a feel for all this is to read some really good science fiction. Because the writer's been talking about this for 20 years.

Here's the fear at the bottom of this record. It's not only that artificial intelligence will be, and here's the way I think about it, way faster at doing very complex calculations, going to get there in a second. Well, first of all, I probably could never do it. And even if I could do it, it would take me forever. So it's going to outperform me on anything where extraordinarily complex calculation is going to matter. Now, that's one kind of intelligence, but it's sure not all, as we all know. That capacity is already there.

But what if when that happens, here we get to the fear, it wants to control me. And because it's so much faster than I am at these complex calculation, it figures out how to control me. At the root of the existential risk is the fear that we are creating a monster that ultimately will control and subjugate humans. But maybe the key thing there, Janice, just gets your thoughts on this, is that, again, it may not have the intention to control us. it just may simply have goals that evolve and become different maybe than the original goals that we, we ascribe to it. So that in then in pursuing those different goals, like turning everything in the world into a paper clip, that's not so great for everything that's not a paper clip.

Do you buy that kind of an idea? The key challenge here is, the key risk is how do we give AI goals? How do we ensure that it maintains those goals and those goals are generally associated with, like, human flourishing, as opposed to human non-flourishing? These seem like pretty complicated things to figure out in real time when most of this Janus is being done by highly competitive megacorporations that are effectively in a kind of arms race to deploy this technology into their products and into the market, to grab market share? That's the rub in and up there. The first process that you describe, Redyard, is called the alignment problem.

How do we make sure that these hyper-intelligent machines who, when I use the word intelligent, I just mean capable of incredibly fast and difficult computation. That's really what I mean. That's what they're doing. How do we make sure those machines are aligned with our goals? Because once it gets the capacity to learn on its own, it doesn't need to depend on our goals anymore. So how do we align?

And there's a really interesting book that members might want to read. We'll put in the notes afterwards called the alignment problem, which raises this issue. It's a good read. It would be a great way to get ready for that debate. You have to put guardrails around this, but you are smack up against the most innovative and the largest companies in the economy right now. They are mainly American, but they're also Chinese. Let's not forget that China is a massive investor in AI. And whatever regulation, the Europeans, Europeans are ahead in regulation here because they don't have any companies in the race.

So it's really easy for Europe to regulate. But were the Europeans and the Americans to regulate, that would still leave China as one of the leading countries in which AI is being developed, free to do whatever it wants. So I am, frankly, that's why I thought the pause was not the right answer. The alert is the right answer. But this genius out of the bottle, it's global. There is no pause.

here. And that, our political class has to get on this. If you're worried about climate change and we know how far behind, you know,

policy was global warming, that ain't nothing compared to the pace at which this is moving. Where we are today would have been inconceivable a year ago, even. And I have no idea where we'll be in six months. The pace of this, it's taken off in an exponential way. Yeah. The capacity.

And that's. Yeah. Let's take a quick break. We'll come back on the other side with our monk donors. And let's talk about the economic impacts. There was a report released by Open AI predicting some pretty significant economic dislocation. So I want to kind of talk, Janice, about how the heck do we interpret that. And then finally, let's touch on, you know, what are the challenges to our democracy? Big election coming up in the United States in 2024. This technology likely right in the center of that, we'll have that conversation right after this short break for our monk donors. Thanks for listening to this excerpt of the Friday Focus podcast to get full-length editions of each and every episode of this program. Simply go to our website, triple w, the monk debates.com. Click on

on the Friday Focus tab in our navigation on the top right of the site. Make a donation as little as $25 a year of 50 cents an episode. And we'll send you not only the full-length editions of each and every Friday Focus podcast, but all kinds of special offers, perks, access to events, and additional content. Again, you can do that right now by becoming a donor to the Monk Debates, at triple W. Monk Debates, MUNK, DebateswithanS.com.