#210 Trust and Regulation in AI with Bruce Schneier, Internationally Renowned Security Technologist

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com.

AI pretends to be a person. AI pretends to have a relationship with you. It doesn't. And it is social trust. So in the same way you trust your phone, your search engine, your email provider, it is a tool and like all of those things, it is a pretty untrustworthy tool. Your phone, your email provider, your social networking platform, they all spy on you. They are all operating against your best interest. My worry is that AI is going to be the same that AI is fundamentally controlled by large for profit corporations.

Welcome to Data Framed. This is Richie the rapid advancements and hype around AI mean that many people now have conflicting emotions on the subject. On the one hand, there's a lot of optimism around the potential for AI to solve problems and make the world better. On the other hand, there is fear of new risks and problems from the technology. Reconciling these two things leads us to a philosophical how much should we trust AI? And and on a practical level, what can we do to make sure that AI is trustworthy enough? Today's guest is Bruce Schneier, one of the world's leading thinkers about security. He's a fellow at the Berkman Klein center for Internet and Society at Harvard University, a lecturer in Public Policy at the Harvard Kennedy School, a board member of the Electronic Frontier foundation, and Access now an Advisory board member of the Electronic Privacy Information center and verifiedvoting.org and also the Chief of Security Architecture at Inrupt. On top of this, he's a prolific writer. He's written 16 books and I must confess that I'm a long term fan of his work. I've been reading his Schneier on Security blog for nearly two decades. In recent years he's expanded his attention from security to AI, and I'm keen to hear his insights on its trustworthiness.

Foreign.

Welcome to the show.

Thanks for having me.

I'd love to talk about trust in AI, but before we get to that, can you tell me what trust means to you.

I wrote a book about trust. I don't see it on my bookshelf right now. It is an incredibly complex concept and it has many meanings. It's an overloaded word, kind of like security. So asking what trust means to them, I think is bad because it means many things in many contexts to everybody. There's a difference between trust and trustworthiness. In a security context, trust is often something you have to trust, not something that's trustworthy. I do write about the difference between a more intimate personal trust and social trust, but there's a lot of scholars about trust and they have very different definitions.

And.

And it really depends on what angle you're taking. So I tend to keep it open because many people have different needs when they talk about the need for trust. And as a security person, we often have to provide for them all. But in terms of AI, I think we should differentiate between interpersonal trust. You might trust a friend and a more social trust. You might trust a bank teller or an Uber driver.

Can you just elaborate on that? Like, what's the difference between interpersonal trust and social trust? Like, when would you need one or.

The other you might need? It's not really about need, it's about the relationship. So interpersonal trust is you trusting a friend. Right. It's based on your knowledge of them. It's less about their behavior and more about their inner selves. I trust a friend, I trust a spouse. I trust a relative. We know what that means. I don't know what they're going to do, but I kind of know it'll be informed by who they are. When I say I trust an Uber driver, it's a very different form of trust. I don't know this person. I don't even know their last name. I just met them. But I trust that their behavior means I'm going to get taken to my destination safely. And they could be a bank robber at night, I don't know, but it doesn't matter. So for social trust, there are systems of society that enable that. I mean, the reason I don't just get in the car of random strangers, but I get in the car of random strangers who are Uber drivers. There's an entire system based on surveillance, based on competition of a star, rankings based on whatever background checks Uber does, based on my history, based on their history, that enables us to trust each other in that interaction. Same thing when I hand cash over to a bank teller, I don't know who they are. I'm giving them my money. But I know there's an entire system that the bank has to allow me to trust that person in that circumstance. If we walk outside the bank down the block, I would never give that person my money, like ever. But in the bank I would. And that's the difference. That's a really important difference. Because social trust scales, interpersonal trust is only based on who I know. It's not going to be more than a hundred people in society and it'll probably be less than that. But social trust, I could trust thousands, millions of people. I flew in an airplane yesterday. Think of all the people I trusted during that process, including all the passengers not to leap over and attack me. And I mean, that's a little bit funny, but if we were chimpanzees, we couldn't do that. So social trust is a big deal. It is unique to humans and it makes society work.

So it seems like social trust is then going to be incredibly important when it comes to AI. So in the same way that you mentioned the bank example, where there's hundreds of different people involved in sort of creating all these systems to make sure that social trust works, what's the equivalent that's needed for trusting AI?

So AI is interesting because AI pretends to be a person. AI pretends to have a relationship with you. It doesn't. And it is social trust. So in the same way you trust your phone, your search engine, your email provider, it is a tool, and like all of those things, it is a pretty untrustworthy tool. Your phone, your email provider, your social networking platform, they all spy on you, they are all operating against your best interest. My worry is that AI is going to be the same. That AI is fundamentally controlled by large for profit corporations, that surveillance capitalism will be unavoidable as a business model and these systems will be untrustworthy. We want social trust with them, but we won't have it. And because they are relational, because they are conversational, they will fool us. We will be fooled into thinking of them as friends and not as services. Whereas at best they're going to be services. So I worry a lot about misplaced trust in AI.

That certainly seems like it could be a very big problem when you sort of start giving away all your sort of most personal details or intimate thoughts to some AI that isn't actually as trustworthy as it is.

So yes and no. Already your search engine knows more about you than your spouse, than your best friends. We never lie to our search engines. They know about our hopes, our dreams, our fears whatever we're thinking about. Similarly, our social networking platforms know a lot about us, our phones, right? They know where we go, who we're with, what we're doing. So we are already giving a lot of personal data to untrustworthy services. AI is going to be like that. And when I think about AI digital assistants, I think it's one of the holy grails of personal AI. An AI that will be my travel agent and secretary and life coach and relationship counselor, concierge and all of those things. We're going to want it to know everything about us. We're going to want it to be intimate because it'll do a better job. And now how do we make it so that it's not also spying on us? And that's going to be hard.

So in terms of being able to know when it's appropriate to have those sort of trustworthy interactions or those intimate interactions, is there any way to gauge how trustworthy some AI is? Or have you just got to assume that it's not trustworthy?

We have to assume that the AIs that are run by Meta and by Google and by Microsoft and Amazon are going to be no more, more trustworthy than all of their other services. I mean, that would be foolish to think that Google, whose business model is spying on you, would make an AI that doesn't spy on you. I mean, maybe they won't, but that's not the way to bet. And surveillance capitalism is the business model of the Internet. Pretty much everything spies on us, your car, spies on you, your refrigerator spies on you, your drones. I mean, whatever it is, we see again and again how surveillance is pervasive in all of these systems. So we can't expect different here. And I think it'd be foolish if we did. If we want different, we're going to have to legislate. It's our only hope.

I think you're right that at the moment all the most powerful AIs are created by these large technology companies. So you mentioned the idea of legislation. Is there an alternative to this corporate AI?

I push for the notion of public AI. I think we should have at least one model out there. I don't need, I don't know, I just need one that is not built by a for profit corporation for their own benefit. And it could be a university doing it, it could be a government doing it, it could be a consortium, an ngo. And as long as it is public. And by this I don't mean a corporate model that has been made public domain. So the Llama model doesn't count. That is still a corporate model, even though we have access to its details. It needs to be a model built ground up on sort of nonprofit principles. This feels important. They get a different kind of AI. I don't need it to dominate. I don't need it to transplant all the corporate AIs. I need it in the mix. I need it to be an option that people can choose and an option that we researchers can study in opposition to the corporate AI and sort of understand where the contours are. You know, it's not a big ask. Models are expensive to build, but in the scheme of government expenditures, they're cheap and models are getting cheaper all the time. But this feels important. If we're to understand how much and how we can trust corporate AI, we're going to need non corporate AI to compare it to. Otherwise we're not going to be able to make good decisions.

So this is a really interesting idea, just having this sort of counterweight to the corporate AI, just having a public model. So there are different sort of levels of openness. So I know, for example, the META models, they're sort of open weights, but they don't give you enough details to reproduce everything. So talk me through how much of it needs to be open and publicly available.

I think all of it built, ground up by the public, for the public, and in a way that requires political accountability, not just market accountability. So openness, transparency, responsive to public demands. So we know the training data, we know how it's trained, we have the weights, we have the model. And that now becomes something that anybody can build on top of. So universal access to the entire stack. Now this becomes a foundation for a free market in AI innovations. Now we're getting some of that with both the hugging face model out of France and with the LLAMA model out of Meta. But they are still proprietarily built and then given to the public, which is not enough. You're not going to get a model that isn't responsive in the same way to corporate demands. And maybe that doesn't matter, but we don't know if it doesn't matter yet. And I think we need to know. This is just too important to solely give to the near term financial interests of a bunch of tech billionaires. We're going to need a better way to think about this. And the goal isn't to make AI into friends. You're never going to get interpersonal trust with AI. All I want is reliable service in the way Uber is A reliable service, even though I don't know, or in this interpersonal way, trust anybody involved in that system. The way the mechanism works allows me to use Ubers anywhere on the planet without even thinking about it. And that's kind of the way trust works. Well, like when I get in an airplane, I don't even think about, do I trust the pilot, the plane, the maintenance engineers. I mean, I know because of this social trust that Delta Airlines puts well trained and well rested crews in cockpits on schedule. I don't have to think about it. When I go to lunch in about an hour or so, I'm not going to walk into the kitchen and check their sanitation levels. I know that there are health codes here in Cambridge, Massachusetts that will ensure that I'm not going to die of food poisoning. Are these perfect? No. People do occasionally get sick in restaurants. Airplanes occasionally fall out of the sky, but largely it's good enough. And there are Uber drivers that commit crimes against passengers. It's largely good enough. And actually I think Uber drives are interesting. Taxi driver used to be one of the country's most dangerous professions. It was incredibly risky to be a taxi driver in a big city. And Uber changed that through surveillance that enables this social trust.

There's a lot to unpack there. I didn't realize that taxi driver was such a dangerous occupation. But it's interesting that these sort of additional regulations, like so for example you mentioned with restaurants, there are regulations around sanitation and making sure that the food is going to be healthy and I'm going to put food poison people is sort of beneficial and helps in scale. I think a lot of people, when they say, okay, more regulation, their gut reaction is going to be, oh well, regulation stops innovation. So what regulation can you have for AI that is going to.

So let's stop for a second. Regulation stops innovation is a bullshit argument given to you by people who don't want to be regulated. It is not true. Do we have problems with innovation in healthcare? Do we have problems with innovation in automobile design? We put regulation in place because unfettered innovation kills people. So we're okay with it taking a few years for a drug to hit the market? Because if you get it wrong, people die. Same thing with airplanes. And if you wander around, what is the lack of innovation in restaurants in your city because of health codes? None. Zero. That is a fundamentally bullshit argument. Do not fall for it. That's one. Two is if it inhibits innovation, maybe we're okay with that. If innovation means people die, we want slower innovation. You can only move fast and break things. If the things you break are not consequential, once the things you break are life and property, then you can't move that fast. And if you don't like it, get into a different industry. I don't care. So I have no sympathy for companies that don't want to be regulated. That is the price for operating in society. And sometimes we regulate people out of business. But 150 years ago we said to an entire industry, you can't send 5 year olds up chimneys to clean them. And if it hurts your business too fricking bad, we no longer send 5 year olds up chimneys to clean them. Because we are more moral than that. You can't sell pajamas that catch on fire. If it hurts your business model, get a new business model. You can't do it. So I'm okay with putting restrictions on corporations. They exist because it is a clever way to organize capital and markets. That's it. They have no moral imperative to exist. If Facebook can't exist without being regulated, I mean, I want Facebook to spy. Stop spying on its users. If it can't exist because of that, maybe it goes away and gets replaced by a company that can do social networking without spying on their users. There's no rule that Facebook has to exist. Sorry, I'm a little strident on this.

No, I'm glad you're giving your opinions. So in that case, what sort of regulations do you think should be. Do you think should apply to AI in general?

My feeling is we don't need a lot of new regulations for AI because what we want to regulate is a human behavior. So if an AI at a university is racist in their admissions policies, that's illegal. But if a human admissions official is racist, that's also illegal. I don't care whether it's an AI or an AI plus a human or a non AI computer or an entirely human system. It's the output that matters. And that same thing that's true for loan applications or policing or, or any other place, you're going to see bias. If I am concerned about AIs creating fake videos in political campaigns, I'm equally concerned those videos are made with actors in the soundstage, like there's no difference. So in general, my feeling is that we want to regulate the behavior of the humans and not the tools they are using. Now, after saying that AI is a certain type of tool and we'll need some specific rules, just like poisoning someone is illegal. And also we make it harder for average People to buy certain poisons. We do both. So there will be need to be some AI specific rules. But in general regulate the human behavior and not the AI because the humans are the ones who are morally responsible with what's going on.

That certainly seems to help, I guess people or managers become more accountable. It's like okay, we're putting this AI tool interaction, but actually we are responsible for what goes on.

And it's the same if it's a non AI tool. Right? You put a tool in and a tool is an extension of your power and responsibility. So if it does damage, then it's your fault. And we have experience with this. A lot of talk is about what happens when AI start killing AI start killing people. You know, robots have been killing people for decades. You know, there's a steady stream, not a lot of industrial robot accidents where people die in the us in Europe and Asia. This is not new. So we have experience with robots and AI systems taking human life. And in general it is, it's the company, it's the maintainers. You know, courts are good at figuring out who's at fault and we as society have experience with that. That is not going to be a new thing.

So it seems like existing laws are going to largely cover the use cases of AI if we let. So how do you feel about the sort of new raft of AI regulations then? So we've got. The EU AI act is sort of the most recent, but there are quite a few on the way.

EUA act is really I think the only one that is real. I mean us, we have an executive order but you know, no one thinks Congress will be able to pass any IA regulation. We can't even regulate social media and we've been trying for what a decade. So I'm not optimistic about, about regulating AI. The EU AI act is good. I mean it's a good first attempt. You can tell it it exhibits the problems with regulation of the technology since technology changes. The EU A act is being written before generative AI becomes a thing. Then GPT hits the mainstream, they're frantically rewriting the law is kind of half and half. But what is the big thing next year and will the law cover it? And this is the problem of writing tech specific laws. The tech changes, it doesn't change are the humans. But in general I like the EU AI act. It's a really good attempt. I like the idea that they breaking applications into four different threat levels and regulating differently. You know, I think we can do more in banking as a good example we regulate large banks more heavily than small banks. So banking regulation is key to how big you are. We need regulation, but we don't want to strangle little banks. Now in tech, the big companies like regulation because, you know, it weeds out the competition. Facebook will able to meet any regulation. Any government throws at it, it's Facebook's competition that won't. That kind of thinking is needed here. But the tech is moving so fast. So we are going to need a regulatory environment that is flexible and agile and we're not good at that at society.

So it should be the larger companies that have more regulation or the larger AI models.

I think that's something we have to at least think about. I don't have answers here, but I kind of have been poking at the question. But yes, looking at some kind of tiered approach I think would be interesting here.

If we're just looking at regulations, then that's going to put a lot of burden on governments to create trust in AI. Are there any things beyond regulation we can do to increase trust?

I think not. I mean, corporations are fundamentally untrustworthy, meaning you cannot have interpersonal trust relationship with them. You can have a social trust relationship with them, but corporations are precisely as immoral as the law will let them get away with. That's the way it works. The government establishes the rules and the market plays on top of the rules. If we want to force more trust, you need regulation. No company, no industry has improved safety or security without being mandated by the government like ever. And it's planes, it's trains, automobiles, pharmaceuticals, workplace food and drugs, consumer goods. These are all industries that produce dangerous things until they were forced not to. The market doesn't reward safety and security in the same way. And the market rewards fake trust just as much as real trust. Especially when you have individuals working in a near term reward horizon. So you've got these market failures where you need non market mechanisms.

So if you need regulation, is it going to have to be every country is sort of making its own regulations or do we need some kind of global synchronization here?

Global synchronization doesn't exist, so it doesn't matter what you need. You're not getting it. I'm fine. Fine with countries doing their own things. This is again, companies complain, there's, you know, lots of regulations and hard to meet them all. And again, my response is too freaking bad. Get used to it. You're a corporation, Internet, you're a multinational corporation. If you can't handle it, don't be a Multinational corporation. This isn't hard. So I'm okay with a patchwork. I'm okay with a patchworker states. We have state privacy rules that are different in every state. Meta's doing just fine. They complain a lot, but they're doing just fine. So I'm good with a patchwork. I don't think you're going to get international harmonization like anytime soon because we can't get it on even easier things. And this is just moving so fast now. We don't have planetary government, I suppose.

Yeah, certainly. Probably pros and cons of having a planetary government. And yeah, maybe the patchwork is best, at least for now.

No, I'm actually a fan of planetary government. I think in general, we as a species have two sets of problems. We have global problems and we have local problems. We no longer have like France sized problems. So I tend to like government that is very big and very small. Right now, the medium size seems to me it was very important in an industrial age, but it seems less important today. That is a bigger conversation than this.

Yeah, I feel that's a whole separate podcast episode. All right, so one thing that seems to have come up in conversation a few times already is social media. And it seemed like social media went from like being the sort of darling of like, this is going to save democracy and with things like the Arab Spring and all this sort of like 10, 15 years ago and being kind of like, yeah, this is causing a lot of problems. And yeah, it's widely sort of, well, there's disillusionment anyway. So are there any lessons that the AI industry can learn from social media?

So yes, and actually I wrote an essay, it appeared in the MIT Tech Journal like last month called 5 Lessons AI Can Learn from Social Media, where I talk about our inability to regulate social media and causing all these problems and how we can learn from their mistakes. And it's things like virality, it's things like surveillance, lock in, in monopolization. I think that's the big one, that the biggest problem with social media is that they're monopolies. And that means they just don't have to be responsive to their users or their customers, which are the advertisers actually, but their users most importantly. And anything we can do to break up the tech monopolies will be incredibly valuable in all of this. And that's the biggest lesson.

So you think in this case, I guess for anyone who's thinking about regulation, you should think, well, okay, these AI companies, they're going to be monopolies and therefore you need to regulate them as monopolies. Is that the idea?

Yeah, I mean, I mean monopolization as a. If you were a monopoly, you have a lot more power to shape the market and not respond basically to market demands. Right. You can operate outside, you know, the basic tenets of a capitalist market system and that breaks the system. So I need there to be competition. I need, I need sellers to compete for buyers. Right. That is how I get a vibrant market. And if sellers aren't competing with each other for buyers, you don't have that dynamic. And that's what's really important.

And do you think that sort of that monopoly market is going to be inevitable for AI, particularly for generative AI, or do you think there will be that inevitable?

It's only enabled by the ability of corporations to ignore antitrust law. We have laws in place to try to prevent monopolies. We didn't enforce them for a few decades, which is why we have the big tech monopolies. They are starting to be enforced again now. But the power imbalance is great, so we'll see how it goes. But EU is doing better than the US and in a sense the EU is the regulatory superpower of the planet. I look to them more than the US to keep these companies in check.

It occurs to me that we've been talking about corporations as being this sort of like single entity, but actually they consist of people. And in terms of people, but that's.

Kind of a. I mean, what's his name said that? Romney said that. I mean it's true and it's bullshit. Charlie Stroh, science fiction writer, talks about corporations as slow AI. A really interesting parallel that yes, they're people, but they are this socio technical organism and they have things they do are greater than the individual people. It's like saying that a car is like metal and screws. I mean, yes and no. And if you think about it, let's take Meta. I mean Meta could decide tomorrow not to surveil its people. And if they did, the CEO would be fired and replaced with a less ethical CEO. The people can't operate with full autonomy because of the system they're embedded in. So corporations are not just a bunch of people in a room. They are a highly structured multi human entity and you cannot reduce them to just people.

I suppose that's true. There are certainly limits on things I could say like I would never tell people it's a terrible idea to learn about data and AI because that's what I do. And they're immortal.

I mean, they're immortal in a way that the people aren't. Right? They outlive the people in them. The people in them come and go. It's like your skin cells. The cells in your body come and go, but who you are is greater than a pile of cells. Maybe that's a better analogy.

We're skin cells as part of a human corporation, something. All right, so just human being from above, though. So within the corporations, there are people who are sort of building these AI tools who are working on these things. And we have a lot of them listening in the audience. So do you have any advice for the people who are building AI or making use of AI at work? Like, what do they need to do to create more trustworthy AI? Or can they have some sort of effect?

I think really pay attention to applications. It matters. Applications. If I have an AI that's a political candidate chatbot and it says we should nuke New Zealand, that's a gaffe. If I have a AI chatbot that's helping someone fill out immigration paperwork and it makes a mistake, they get deported. So the use matters a lot. What is the cost of failure? What is the trust environment? Is it adversarial? Is it not? If I have an AI that is advising me in corporate negotiations, how much does it matter if the AI's parent company is on the other side of those negotiations? So pay attention to the use case a lot, and that really determines whether it makes sense. Like, AI assistants are doing a lot of legal work. As long as there are human lawyers reviewing it, that's fantastic. Makes lawyers more effective. So think about how the human and AIs interact. Think about the systems and the trust that needs to be in the system. The cost of getting it wrong, in addition to how often things get get wrong. And notice that that will change all the time. This field is moving incredibly fast. What is true today won't be true in six months. So any decision you make needs to be constantly revisited.

I think that's quite important. Like, think about what goes wrong. Because often when you're building something, you think, well, just, can I make something that gives the right answer sometimes? But think about how it can be misused. Think about what happens when it fails is equally important. And how about for people who are just using AI, is there anything they can do in order to make sure that AI becomes more trustworthy over time?

Better laws is the answer, but really no. I mean, like AI is already embedded in your mapping software. The AI is Giving you directions. What could you do? You either use it or you don't. AI is controlling your feed on TikTok or Facebook. What are your options? There aren't any. So really, for us as consumers, the AIs are handed to us embedded in systems. And that's pretty much like all tech. And we either choose to use the systems or not. This is where the monopolies are our problem, because often we have no choice. I mean, I mean, I can tell you, like, don't get a cell phone, but that's like dumb advice in the 21st century. It's not real, that's not viable. So again, I need government to step in and ensure that, you know, you can use the cell phone without it being too, too bad. And most people believe they have more protections in their consumer goods than they do with phones and interconnected cars. And there's a story that broke a couple weeks ago about GM spying on its drivers and selling information to insurance companies. People were surprised there. I was surprised. Kashmir Hill, who writes about privacy for the New York Times, was surprised. But should we be surprised? No. But we believe we have more protections than we do.

So it seems like if we don't have those protections, then that's going to break down the social trust. Then to go back to your original.

Point, and it does. And I think this is why you're seeing what we mentioned a little bit earlier, this backlash against social media. We thought it was all good. Now we think it's all bad. The truth is in the middle, but we thought we could trust them.

It seems like there are some sort of good possible futures and some bad possible futures. What's your sort of ideal situation here? What happens next that you think will make things go well?

I think that AI as assistive tech is phenomenal. A lot of what goes well is human plus AI, and a lot of what goes poorly is AI replacing human. least today, few years it might be different. So the more we leverage these technologies to enhance humans and really to enhance human flourishing, the better we're going to do. Now, that is not necessarily where the market's going to push. Market will push towards replacement because that is cheaper. But that has a lot of downstream effects. Massive job loss is just not good for society. And we might want to think about ways to help people that aren't tied to their jobs. In the United States, all of your stuff is tied to your job. Unlike Europe. In Europe, they got healthcare through the political process. In the us, we got healthcare through collective bargaining. And that didn't matter in the 50s, they both worked, but here we are in this decade. It means your healthcare is tied to your job in a way it's not in Europe, and that's not serving us well right now. So if we're going to see massive unemployment because of AI, we need to figure out some other way to deliver basic human services that aren't tied to your job. That's not great because we as a society really have trouble doing all these things. So that's kind of meandered a bit here. Get me back on track.

No, that's. We started off going down the happy path and then sort of turned into disaster zones.

I think there's an enormous power in AI. I am mostly optimist. I mean, I know I'm a security person and I say a lot of pessimistic things, but I am mostly optimistic here. I think this will be incredibly powerful for democracy, for society. So that's not what I'm writing about now, and I think it's true.

So maybe we'll try and finish on a happy note. So what is it that you're most optimistic about?

I think there's enormous power in AI as a mediator and moderator and consensus builder. And a lot of the things I think AI will do for us are things humans can probably capably do. We just don't have enough humans. So putting an AI moderator in every Reddit group, in local online government meetings, in citizen assemblies on different issues, be incredibly powerful. AI doing adjudication. I think AI is sense maker explaining to people political issues. AI is an infinitely patient teacher. So instead of reading a book, you engage in a conversation. I think that makes you a better person. If we can get that working at scale, the enormous values of AI as a doctor. There are parts of this planet where people never see a doctor because there aren't enough doctors. But an AI assisted nurse will be just as good in almost all cases. I mean, there's phenomenal potential there. AI is doing research, especially research that is like big data pattern matching. Where are you seeing articles about AI and drug discovery? I think there's a lot of potential there. So really I look for places where there aren't enough humans to do the work. And AI can make the humans that are doing the work more effective.

Lots of incredibly positive things there. I really like the idea of an AI moderator. That's something that hasn't cropped up in many discussions, really, so that. That's pretty.

So.

Yeah, a few months ago I.

Was interviewed by a podcaster Back like last summer. ChatGPT is becoming a big thing initially. And they said, I went to ChatGPT and asked a bunch of interview questions. Asked it to come up with interview questions for you, and here they are. And they were fantastic interview questions. Like, one of them was, if you were an action figure, what was it? What would your accessories be? I'd never gotten that question before.

That's brilliant. Okay, so don't ask that. I guess my days are numbered as a podcast.

Maybe.

Look out for AI Richie.

Isn't there an NPR episode on AI where they had the AI come up with a podcast and it asked questions and it came up with a little sketch on the topic as most like three part look it up. I. I think it was some NPR program. Might have been All Things Considered.

Absolutely. Actually, I just recently saw Reese Hoffman interviewing an AI version of himself, and that was exciting. Yeah, it's getting very close to. Yeah, Richie is replaceable. All right, so just to finish up, do you have any advice? Is there, like, some action you think people should take in order to get towards this happy path of AI being good?

I mean, to me, this has to become a political issue, and nothing will change unless government forces it. Government will not force it unless we the people demand it. I want these things discussed at presidential debates, and I want them to be political issues that people campaign on. Like, that matter in the same way that inflation matters, unemployment matters, and US China policy matters. It needs to matter to us. Otherwise the tech monopolies are going to just roll over the government, because that's what they do. They have the money, they have the lobbying, and it's very hard to get policy that the money doesn't want.

It's really hard, I think, call to action there for everyone to start writing letters to their local representative.

To get their AIs to write letters. The local representative.

Oh, yeah, yeah, they should write letters. There we go. Technology something again. Nice. All right. Thank you so much for coming on the show, Bruce. That was brilliant.

Good luck. Thank you.