#335 Rebuilding Trust in the Digital Age with Jimmy Wales, Founder at Wikipedia

Generative AI is transforming industries at an unprecedented pace. But as AI changes how you or your team work, one thing is clear. Your skills also need to evolve. At datacamp, we offer everything you or your team need to adapt and thrive with AI. Whether it's business users looking to get the Most out of ChatGPT and Copilot, or developers and data scientists looking to fine tune models. You can learn the entire AI skills spectrum on Datacamp. Power your AI transformation today. Start learning@datacamp.com I think there is a huge opportunity for AI based tools to support the community, but I think it's really important. Like we're really quite adamant about human in the loop. Part of that is just simple practicality, like large language models. They're amazing and they're, you know, I love playing with them, but they're not good enough to write an encyclopedia article. And it's amazing that we were getting such great results. But it, it does lead to hallucinations and the hallucinations are just wrong. And in fact they are worse. One of the things that's sort of obvious to people is in a business setting, trust is really important. It's a lubrication that makes everything a lot easier. You've seen stories that to me are just mind boggling about companies that are, when people work from home, they're using software to monitor that you're still moving your mouse and clicking on the keyboard and I'm like, wow, that level of mistrust of employees is just deeply wrong. You don't want to work somewhere like that. They don't trust you and that's a hostile environment and it's just not good.

Welcome to Data Framed. This is Richie. Sometimes an idea is so successful you can't believe it wasn't always obvious. Wikipedia has been one of the top 10 most visited websites in the world for more than a decade. But when it first started out, it was perceived as a terrible idea. If any fool could edit the encyclopedia, then how could you trust the contents? Today we're going to discuss the techniques that Wikipedia as an organization used to build trust. It's a timely topic since AI is stirring a lot of conversations about what you can and can't trust and beyond trust. Wikipedia has long been a leader in automation for updating and managing content. So I want to find out how generative AI is changing things at the organization. Our guest is none other than Jimmy Wales, the founder at Wikipedia and now chair emeritus at Wikimedia Foundation. Jimmy's written a new book, the seven Rules of Trust, which forms the basis for today's conversation. So let's hear Jimmy's story. Hi, Jimmy, welcome to the show.

Thanks for having me on.

Yeah, great to have you here. Now I'd like to start by going back to the early days of Wikipedia. Now, it wasn't immediately considered to be a really good idea. I know there were a lot of jokes around Wikipedia. So can you tell me like, what was the harshest thing that was said about the site?

Oh yeah, it's a very good question. I would say, you know, obviously Stephen Colbert sort of quite famously made fun of us on his show, but that was actually, it was interesting because he played a character that was a right wing blowhard called Stephen Colbert and he is Stephen Colbert. Right. And so him making fun of you was not completely terrible. Right, because it was like, oh, he's sort of playing a buffoon and you know, he, he talked about wikiality and you know, if you don't like something in reality, just edit the wiki and it changes, you know, so it was a bit silly, but you know, I think mostly it was more like people just weren't sure. Like they're like, okay, like I don't, don't think that's going to work, but okay, whatever, have a try, you know?

Yeah, I mean, I guess there are worse forms of publicity than having Stephen Colbert do a routine about your, about your product.

Yeah, yeah, yeah. No, exactly. Although, I mean, I would say the worst wasn't poking fun. It was actually there was an error in Wikipedia and it was about John Sigenthaler Sr. And it said that he had been briefly suspected of having something to do with the Kennedy assassinations, which was absolutely untrue. He was actually a friend of the Kennedys. He was a pallbearer at Bobby Kennedy's funeral. So he saw this and he called me up and he's like, hey, you know, I'm like, oh, I'm really sorry. And I, you know, it was, it was down within minutes after I got off the phone with him and I checked the traffic logs and things like that and almost no one had seen the page. So I was like, okay, that was terrible. But. But then he wrote a scathing editorial in USA Today about us, which was kind of a big deal and really led to us, you know, led to me promulgating a biography of Living Persons policy which says any negative statement in a Wikipedia entry about a living person should be removed immediately unless it has a good source, I. E. What was happening at the time was people would say something like, oh, that's weird, that doesn't have a source. And they would go on the talk page, oh, this doesn't have a source. And then people would discuss, oh, we're going to get a source, saying, meanwhile, it's there on the site and shouldn't be. So. And, you know, that was when I would say we got pretty serious about human dignity of saying, like, if it's a biography, we have a really a higher level of responsibility, a biography of a living person. Like, you know, screw something up about just Julius Caesar, okay, well, his family's feelings won't be hurt by it. It's bad. We should fix it as soon as possible. But the, the duty of care is much higher when it's a living person. So.

Absolutely, yeah. And certainly if you got bad things being said about you on a Wikipedia, on a Wikipedia page about you, I mean, that's got to be quite hurtful.

I think, because it's definitely.

Yeah. So I love the idea of just like prioritizing. Well, these are the really important things that we need to edit very quickly. All right, so I'd love to. I guess Wikipedia is now is one of the biggest sites in the world. Do you want to just give me a summary of how you think you got this crazy success?

Yeah, I mean, I, I think there's a few things. So certainly in the early days, we met a need that was not being met online. So, you know, the, the idea of just very straightforward, factual information about everything under the sun is kind of an obvious use of the Internet, but it didn't exist yet. There were a few encyclopedia reference type of sites. Some of them were paywalled, some of them were covered with ads, and they were small. And, you know, but in terms of the, the, you know, a lot of times when people go on the Internet and they're searching for information, pretty much what they're looking for is a Wikipedia article. Like, that's kind of a thing that needed to be there. So that meant as the small community got started writing things, they got traffic. Like, it was interesting because it was like what people were looking for. And so that had a big part of it. And then, you know, over time, I think there are some of the early principles of Wikipedia which have been incredibly important to ongoing success and trust. As my book is about trust, I should talk about trust, including the idea of a neutral point of view, that Wikipedia should be neutral and shouldn't take sides on any major disputes. And that's obviously always something that's contentious. And lots to grapple with there. But I think that's been really important and something that we never should forget and we should double down on as much as possible, because it means that anyone from any side of the political spectrum or whatever belief system you might have should be able to come to Wikipedia and get a clear explanation of the state of debate about anything. And so that's really important. And then, you know, rule three in the book is purpose. So I think one of the things that helped make Wikipedia successful is it has a really simple, clear purpose. Like, we're going to write an encyclopedia. It should be high quality, it should be neutral, it should be in all the languages of the world, and everyone can understand that purpose. And so it really helped people come together because they sort of, we know what we're doing. Like, we're here, we're going to start working together. We know what we're trying to accomplish. If it were a little more mushy and vague of like, oh, we're going to write a bunch of stuff and, you know, like, people wouldn't know what to do. Like, oh, is it, Is it a joke aside? Or should we make funny things? Should we make serious things? Should we have funny cat videos? And no, we probably shouldn't. That's not really what we're here for.

Actually, I love the idea of just saying we're writing an encyclopedia because it just really distills things down into a simple sentence. If you're like, okay, we're going to write a thousand articles on these particular topics and you get overly weird.

Yeah, yeah, yeah, yeah. And, and in fact, you know, actually one of the things is like, the idea of a free encyclopedia for every single person on the planet in their own language. Like, that's a massive concept, right? That's a huge project. But actually, I think we got more participation early on because that huge project is quite inspiring. Like, if I had said, oh, I'm going to write geographical history of the United States, I'm going to have one article each for each of the 50 states, people would have gone, eh, okay, have fun. That's not that exciting. Like, whatever. Yeah, sounds like a reasonable thing to do. Go for it. Whereas this was like, oh, wow. Like, oh, wow. I can actually take part in something that is super big and super interesting.

That's kind of an important lesson. I guess if you give people like a big dream, then maybe that's the way to make things happen, rather than give people realistic goals and they're like, okay, fine, whatever, I'll hit Those. So I mean, your book's all about trust and there's such a hot topic at the moment. But one thing surprised me, like right at the start of chapter one, you got this picture of a triangle. I'm like, oh, this is Aristotle's rhetoric triangle. I was like, cutting edge, two and a half thousand year old idea. Talk me through how is that related to trust?

Yeah, so the triangle, we actually use Frances Fry's formulation. So she's an academic professor at Harvard who also has a lot of business experience. So she went into Uber when they were experiencing a real crisis of trust and helped manage them through that in the interim period between Travis leaving and Daria coming in. And at that moment, internally, they had all kinds of trust problems. And one of the things that she said to me was, was one of the most fun interviews I did for the book. She's a great person. So she said, you know, one of the things people say about trust is that once you've lost it, you can never get it back. And she said, that's just not true. There are things you can do to rebuild and get trust back. And so her triangle framework is empathy, logic and authenticity. Empathy, authenticity and logic. Empathy, authenticity and not logic. Yeah, I would have gotten there eventually. Yeah. Read the book, people. I have, I promise you, many times. But so, and, and like these three pillars, she says, whenever you have a crisis of trust or a problem with trust, one of these, or at least one of these is, has gone wrong. And so, you know, that provided me a great framework to start thinking. And that framework is her framework, but it's not unique. There are actually a lot of academics working in trust who talk about a three sided triangle. Those are just the terms she uses. Sometimes they use other terms. You know, I thought that was helpful because I was writing about trust and thinking about trust and the various things I've learned, but I didn't have a systematic way of thinking about it. And so bringing it together with that really helped me to go, okay, right now we've got a framework for thinking about this and talking about it.

Okay. Yeah, the same, very useful. And I mean, you mentioned Uber, I think, what was it? Maybe around 2016, 2017? They were having so many different problems with just everything. It was like people getting attacked in the cars and all sorts of horrible stuff. And it seems like they have managed to fix a lot of it.

Internally, they had a lot of trust problems. So the executives didn't trust each other and the, the drivers didn't trust the company and riders you know, like passengers were losing, losing trust and regulators definitely didn't trust Uber. And, you know, they had to really start over. I mean, I remember this amazing story about them tracking a journalist riding an Uber. Like, this was a really bad idea because you're not building trust with the media if you're behaving badly like that. And so, you know, it's sort of, sort of interesting. Okay. Like, you go from that point to being, you know, a reasonably trusted company. I mean, obviously everybody's always got questions about any big company, of course, but, you know, and it. It really required them to take step back and have a real think about trust.

Yeah, I love that. It's like, stop doing really stupid stuff like tracking people in the car, step one.

Okay.

So I guess the other side of things is, I mean, Wikipedia is very much based. I mean, I guess it's a kind of a social media. But a lot of social media is like, it's horrible and angry, whereas Wikipedia is like, it seems like a mostly positive community. Can you talk me through, like, how you built the community side of things?

Yeah, definitely. Well, so, I mean, one thing that I would say is, like, Wikipedia is not social media, but it is media and it is social. So it's sort of in that category a little bit. But I mean, and one of the big differences is that purpose. So one of the things that makes social media so toxic is, you know, whenever there's a box, that's basically what's on your mind. Well, some percentage of people have something really horrible on their minds. Right. And this is true, actually, of a lot of different. So it doesn't have to be a box saying that. It can be a place for uploading a picture. It can be video, YouTube, whatever. But that open to creativity brings with it a lot of problems. That doesn't mean we shouldn't have that. I'm not saying that, but I'm just saying the challenge is really hard. There's. And one of the things that made the challenge a lot easier for us is that purpose. Like, we're here to build an encyclopedia. And so, you know, one of the oldest rules, and this is in the, in the book, the chapter on civility. Your mother was right. Is no, no personal attacks. You know, the idea is we're working together to brighten encyclopedia and we shouldn't be attacking each other and insulting each other and behaving badly. We should have collegial debates. We may disagree, but let's treat each other with respect and kindness because it's really practical. Like it's the only way to get things done. And you know, in social media you don't see that. I mean, particularly, I would say Twitter's Sorry X is probably the worst, just in terms of it's a pretty horrible place with a lot of people viciously attacking each other for no actually useful purpose whatsoever. And because we have the purpose, then, you know, basically you get blocked. Like come in and say, you know, you're all a bunch of Nazis. Okay, hopefully you don't get blocked immediately. Hopefully someone will go, hey, you know, by the way, don't be a jerk. Like, that's not what we do here. Like, you know, like that's not really appropriate. But then, you know, if you keep it up, then yeah, you're just going to get blocked. I mean, it's like, that's not fun and it's not what we're here for. We're a friendly project to build an encyclopedia. So that's a piece of it. And then, you know, obviously there was a lot of detail along the way, a lot of things to learn. You know, one of the things that I would say about Wikipedia is that it is in many ways a child of the dot com crash. And at a moment when we were growing really quickly, there was no money and there was no possibility of raising money and had we had money to throw at the problem, we probably would have gone down a different path that wouldn't have scaled. We just wouldn't know what we were doing. So it's quite natural. If you start a website and you see some social problems, you think, oh, we need a, we need to hire some moderators, we need to hire community managers and it's going to be top down and we're going to make rules and we're going to enforce them and the company will be involved. This is exactly how all social media works is, you know, they've got moderators who, who enforce whatever rules instead. Like, well, there's nobody to hire because we've got no money. So we have to work out how do we do this, how do we manage this in a community? So we have the admins who are elected by the community and then, okay, but you can't let the admins become tyrants. So like, what are the checks and balances? So you can lose your admin ship if you're not doing it the right way. All your actions have to be transparent and you have to justify them. We have the arbitration committees in the larger languages who are, you know, elected to be a higher decision making Body. And, and you know, like, all of these elements had to evolve over time. So we had to involve, we had to evolve the, the institutions of a community in the same way that, you know, if you think about, like, what makes a good municipal government. Right. Well, you, you want, you don't want to get tossed in jail for criticizing the mayor, but you also don't want people accosting your grandma walking down the street. And so you need that sort of balance and you need sort of reason and thoughtfulness. And you, and you want the police to be subject to checks and balances, right? Yep. They're there to enforce the law, but they're not there to crack people in the head for no reason. And so all of that, those kind of ideas and principles which we fully get in general society, like, they also make sense online. You know, you can't copy everything word for word and line for line from the municipal code of conduct. But there's a lot of lessons to be learned and that's what we had to learn. So we just had to sort of make it up as we went along and think about it in a, in a thoughtful and kind way to say, like, okay, here's a problem. How do we deal with that?

It's fascinating. And from the sounds of it, you've kind of invented like a whole system of government, I guess, within the science.

Yes, sort of.

Yeah, yeah. The police and all the delays of like, adjudication and whatever. So I guess the other thing related to this is that it can be very difficult to know like, what is true or online. And so this is certainly a case. If you're trying to write an encyclopedia, you need to know, well, is this correct or not?

Yeah. So I mean, for this we're pretty old fashioned about sources. So, you know, we, we want the sources to be quality magazines, newspapers, academic journals, books. So, you know, when we think about the sourcing for articles, then, you know, it's like, okay, well, we're writing about cancer. You know, which do we prefer? You know, do we prefer the National Enquirer or the New England Journal of Medicine? I think we'll go New England Journal of Medicine, you know, so it's that kind of pretty basic, pretty simple stuff. And even that, you know, it leaves a lot of room for discussion and thought and chewing on what really counts as a reliable source. I mean, one of the problems we have these days online is things circulate on social media and they get promoted by algorithms not because they're true, but because they're controversial or they're viral or they're funny or they're whatever they are. And so stuff. You know, it's very, very common on social media to see some infographic that turns out to be complete nonsense. You know, like none of it actually makes any sense or it's really cherry picked and so forth. And, you know, the Wikipedia is pretty immune to that kind of stuff because we basically spend our lives debating the quality of sources and, and deciding like, okay, well, what, what can we trust? How do we do it? The other thing is when I call. Actually, I, I don't call it this, but I like this phrase. I just heard it from Corey Doctorow, said it the other day in a podcast or, or a speech. I don't know where he said it. Somebody told me about it. Epistemic humility, the idea of, you know, it's not, it's not epistemic nihilism of like, we, we know nothing, we have no idea. There's no such thing as truth. No, no, no, no, it's not that. But it's like saying, actually truth is hard and getting it right is hard. And we need to be very thoughtful and accept that we might get it wrong. And we need to always be ready to chew on ideas, chew on criticisms of Wikipedia, like, okay, oh, you got a criticism of Wikipedia. We shouldn't become defensive and use PR strategies to combat it. We should go like, okay, hold on, let's have a look. Like, tell me more like, what could we fix? What are some sources that we should look at? And obviously we are better or worse at that at different times because we're human beings and things happen and the world's a noisy place. But that's the spirit that we should always have, is to say, oh, how do we make this better?

I love that. And it feels like a good life lesson beyond editing an online encyclopedia. It's like, yeah, accepting that you can be wrong sometime and trying to be better.

And, and you know, I think one of the, one of the core principles of Wikipedia is always assume good faith. And this is not a suicide pact, right? It doesn't mean you assume good faith forever and infinitely, no matter what anybody does, but it's actually a good bet. It's a good starting point. You know, you, you meet someone new. I mean, we do it every day in real life. Like you, you meet somebody new and you don't think, yeah, probably they're going to stab me. I mean, if you do think that, you know, you need some psychotherapy, like, you need some help, like Something's gone bad in your life and you're hurting, and that's unfortunate. And you know, like, let's. Let's see what we can do. But normally you meet somebody and you're like, oh, right, well, here's a new person. Like, what are they going to be like, maybe I'm not interested. Maybe I am interested. Maybe they're kind of annoying. Probably they're not. You know, and it's the same thing online is that that kind of assume good faith when we see people come new to Wikipedia. This is the main application. So somebody comes to Wikipedia and they're doing something and it's not quite right. Okay? You can either assume this person is a plant, an operative, you know, they're working for the KGB or the CIA or whatever crazy thing, you know, or it's like, oh, probably this person is trying to be helpful. They're trying to make Wikipedia better, but they just don't quite know what they're doing yet. Maybe we can help them. Maybe we can give them a pointer and say, oh, like actually, you know, when you wrote this sentence, you used a lot of words that are pretty conclusively, you know, biased in a certain way. Way we would write it would be a lot more reserved. We also need sources, you know, like, blah, blah, blah, I'm teaching you to be a Wikipedian. And that's great, because probably not always, but probably the person's going to go, oh, right, okay, that's great. Great, actually. That makes sense. Yeah, sure. Great, let's do it that way.

Okay. Yeah. So I love the idea of, like, just spend the time to teach people stuff before you start having a fight.

Yeah, yeah, yeah, no, exactly. And, you know, a lot of spaces online aren't really conducive to that. There are great places that are very conducive to that. I mean, one of the things that I write about in the book is there's a subreddit called Change My View. And basically what happens there, people post, you know, whatever their view is on something and they say, change my view. And then what you're supposed to do is whether you agree or not with that view, you're supposed to put forward the best argument you can to try to change their mind. So even if you agree with them, you can say things like, well, I do agree with you, but here's the thing that I think is the most challenging for our shared view. And then at the end, you know, people will say, oh, you changed my mind. You didn't change my mind. Or I, you know, often it's like, oh, yeah, I sort of changed my mind a bit on this part of it, and I still got the general, and it's great. And, and, and people there, you know, they've got really a sense of civility. I mean, it's not perfect, like, whatever, it's human beings, so people get angry and stuff like that. But generally it's quite good, and I think that is great. And we need more spaces like that online.

As a Brit, I find all these things are like, oh, this is how you have polite conversation. I'm very much in favor of this.

Yeah, yeah, no, I know. I mean, I think that's true everywhere. I mean, certainly, you know, Americans are very nice people, usually. I mean, not as nice as Canadians. No one's as nice as Canadians. But yeah, I try to avoid stereotypes, but only I keep the nice ones.

Yeah, absolutely.

Some Canadians out there. How dare you? Like, we're not that nice. Okay, okay. Most of the Canadians are very nice. Just not you. But okay.

Okay. So I mean, you mentioned, like, having civil discussions. So I think maybe perhaps the most vicious parts of Wikipedia are things like for talk pages where you have people arguing about things, and then also you have edit wars where people will change things, they change them back. When people fundamentally can't disagree on things, how do you resolve that?

Yeah, well, it's quite hard. Like, one of the fundamental techniques is to go meta, to step up one level and say, look, we're not going to agree about this fundamental issue, but we can characterize the debate. We can be fair about that. We can describe the issue. So my classic example is, you know, think about a kind and thoughtful Catholic priest and a kind and thoughtful Planned Parenthood activist. They're never going to agree about abortion. They could fight about it until the end of time, but they can step back and say, okay, but we're just going to write about the controversy and we're going to. We're going to sort of explain what the different views are. Turns out people are quite good at that. Quite often people can do that. Another one that I think is really important and that I've been lately, it's sort of one of my campaigns is like, avoid what we call wiki voice. In other words, don't say things in the voice of Wikipedia unless you've got really pretty much overwhelming consensus in the community. If any significant minority in the community disagrees, you should probably step back a bit and start to characterize the debate. Don't say it in the voice of Wikipedia. So there's lots of things you can say. Like, it's quite easy to say Paris is the capital of France. Like, that's fine. Like, that's not in dispute. Like, literally, I don't think there's a single person anywhere who could put forward any plausible disagreement with that. So you can just say it. Like, you can say it. You might need a footnote, you know, but it's fine. Like, that's not controversial. Loads of statements are. Loads of statements are things that you may believe very strongly, but other people in the community may not. And I'm not even talking here about, do we disagree about the fundamentals? Sometimes Maybe you and I agree about the fundamental thing, but I may think, wow, we. We shouldn't say this in the Wikipedia voice. We should sort of talk about the different sides. So even though we agree, we should acknowledge that there's another side. And you may say, like, no, the other side's crazy. I know. In that case, I would say, look, if we can't agree, then it shouldn't be in Wikipedia's voice. It should just be describing the issue. So that. That's one of the techniques, because that element of neutrality, I always say it serves two purposes. So one is the epistemic or the cognitive purpose, the knowledge purpose. Like, I don't want to go to an encyclopedia and only hear one side of the story. I want to hear about the debate, but also the social purpose of, like, oh, how can people who disagree actually work together? Well, this is a great technique, is to say, okay, we're not going to agree, but we can write about what we disagree about and we can agree about the debate and how to characterize it.

Okay, that's interesting. So you've mentioned neutrality a few times, and it seems incredibly important for an encyclopedia to have a neutral point of view. Do you think the neutral point of view is always a good idea? Or when would you want to be opinionated? When you're putting things across?

I think the encyclopedia should be neutral, full stop. And that's sort of top to bottom. That's the end of it. Other aspects of life. Absolutely not. No. I mean, certainly I wouldn't expect Tom Friedman, who I interview in the book. He's a columnist for the New York Times. He's got opinions. His columns are about his opinions. Great. He should be opinionated. He's actually an interesting analyst. And sometimes I disagree with him, but I'm like, I always have to grapple with what he says, even if I disagree, because he's a smart guy. And he knows a lot of people and he's talked to, he's lived through a lot of things and that's great. And certainly, you know, in my life, I, I definitely would say we at the Wikimedia foundation, but and me personally as a person who's an activist for an open Internet and free software, open source advocate for access to knowledge, then yeah, I've got a lot of things to say. They're quite opinionated like, yeah, basically you shouldn't be censoring Wikipedia. You shouldn't block Wikipedia because you don't agree with what it says like that. I've got an opinion there and I'm ready to fight for it. So yeah, the, the idea of neutrality is something though that even in those kind of contexts. Right. It's actually, I think helpful to people if you are an activist for something. Like you really think I want to change the world in this way, let's say the environment. Right. It still behooves you when you are learning about some particular sub issue within the environmental cause to take that sort of Wikipedia and mindset to say, okay, well the first thing I need to do is understand. I need to understand what are the debates, what's the argument here? And if I don't do that, I'm going to be a less effective advocate because I'm going to be missing some of the facts, I'm going to get some things wrong and therefore I won't be able to convince the people who disagree with me because they're going to be like, you don't know what you're talking about. Yeah, I think it comes in a lot of different places where reason and sort of thoughtful chewing on ideas is probably our most important human responsibility.

Absolutely. So yeah, I love the idea that like, if you've got a good opinion, then yeah, go for it. But yeah, in the, in the context of encyclopedias. Yeah, let's do neutral things. You mentioned human responsibility and there's been a lot of talk like community with respect to humans. I think obviously like we've reached a point with AI. We've had like some people talking about like we're not far off the point of having AI colleagues. I know Wikipedia's like long had like a lot of automation. Like there've been bots for many, many years to automate stuff. Have you seen this progress over the last few years?

Yeah, I mean, so it's really interesting. So our, our bots. So there's kind of two things. So one, most of the bots aren't Even remotely AI related. Like they're, they're doing things like, you know, like if you see an article go from 20,000 bytes to 4 bytes, probably that's vandalism. So a bot might revert that instantly and notify a human to be like, oh, something weird's happened here. That's not AI, that's just a basic automation of some, some things. We also have some machine learning tools and algorithms which are somewhat used in the community. So there's an ORES scoring system which you can turn on and filter recent changes. I sometimes play with it. One of my favorite is edits that are bad, but probably in good faith from new users because I'm always interested in that. Like, oh, okay. And can it accurately detect that it's okay? Like it's just using this is pre large language models. So it's just using some statistical things. Like actually yeah, I find it's not perfect but sometimes you look at it and you're like, yeah, that looks like to me as a human, that looks like a good faith edit that's not very good by a new user. So then how do I respond to that? Okay, well that's a whole category of like, oh well I should probably go talk to them and, and, and this and that and you know, identifying spam, that can be automated pretty well. You know, it's like spam is spam and it's not that hard to detect in terms of like what we're looking at now. I think there is a huge opportunity for AI based tools to support the community but I think it's really important. Like we're really quite adamant about human in the loop. Part of that is just simple practicality. Like large language models. They're amazing and they're, you know, I love playing with them but they're not good enough to write an encyclopedia article. The way the technology works. And I understand your audience is a lot of machine learning and AI people. So for them, sorry, I'm boring and hopefully I don't get it too far wrong. But you know, this, you know, predicting the next token based on, you know, probabilistic guessing what the most plausible next thing is is amazing. And it's amazing it works at all. And it's amazing that we, we're getting such great results. But it, it does lead to hallucinations and the hallucinations are just wrong. And in fact they are worse in my experience. And I think this is uncontroversial to say they're much worse on more obscure topics. So the Further, you get from very simple claims into more obscure things. Like, I, I saw an example which I haven't tested much, but I, you know, if, like, if you ask, who is Tom Cruise's mother? Most of the models get it right. If you ask that person, I can't remember her name. If you ask, who's her son? No idea, like, or they make something up. And why is that? Well, it's because they've read it one way. Lots of times they've read about Tom Cruise and they, you know, but they've never read the other way. And so like, for a human, and, and human reasoning, it's like, well, the minute I can ever remember Tom Cruise's mother's name, I'm gonna have to, if I use this anecdote, I'm gonna have to do that. But I know if I know who your mother is, I know who her son is. Like, that's, that's like tautological. Like, that's very simple. But the models don't work that way. They're not really understanding. They're just predicting tokens. And so as a result, on less common topics, they, they hallucinate more. Which means for writing the encyclopedia, they're not good, but there's stuff they are really good at. And you know, one of the things I've played around with, and I hope to get back to it after the book tour and all of this stuff that I'm doing right now, I wrote a little script just to say, okay, like, I can feed it a Wikipedia entry, a short entry, because it, it's not really good at doing a long one short Wikipedia entry. And the sources, maybe there's five sources. And I ask it, is there anything in these sources that should be in Wikipedia but isn't, or is there anything in Wikipedia that's not supported by the sources? Make some suggestions. And you know what? It's pretty good. It's not perfect. It's pretty good. And I haven't even done any like, prompt engineering or playing around to try to improve it, but I feel like it's probably good enough to give some pointers. And so now you can imagine if you had a more sophisticated system. This is just me banging out a script and I'm a very bad programmer, so whatever. But a more sophisticated system that somehow, you know, let's use multiple models to ask the same thing, let's cross check it, let's actually make sure that it's not hallucinating the sources. Let's do this and that, and let's tell it don't make a suggestion unless you're really, really sure that's actually one of the harder things to do. Then what if you could do that and run it massively over all of Wikipedia and leave little notes on the talk page saying, hi, I'm your friendly neighborhood robot. And I noticed that three of the sources mention this person's birthplace, but Wikipedia doesn't have their birthplace. And usually we do. What do you think? And then a human can go, oh, right. Because one of the things we know, like a lot of Wikipedians, they come to Wikipedia not to, you know, like, they're just like, I'm a Wikipedian. I'm like, what am I going to do today? I'm going to do something good. In Wikipedia, they're always looking for something to do and they sometimes just click random article. Like people enjoy doing that. Click random article and see if you can improve it. Well, this would be a little bit better than random. It'd be like, oh, here's a list of things where the AI thinks there's an easy fix. They'll take a few seconds for a human to approve, have a go at it. I think people will enjoy it. If they don't, they won't use it. So that's what's great, is like, let's just make some things that the community might find useful and see if they do So.

I love that idea of having AI assistants for just inspiring humans on what to edit. And I guess, I mean, there's probably some of our data frame listeners can probably help out building that sort of thing. So, yeah, challenge for the audience, I think go to Wikipedia, build some tools for.

Oh, yeah, that'd be lovely. Yeah, yeah, you can come to my User Talk page and I'll chat with you about it because I'm very interested.

Wonderful. Okay, challenge for the audience. But you mentioned you don't think having a completely AI generated encyclopedia is a good idea. In that case, I got to ask, what's your take on Grokipedia?

Well, I mean, I haven't had time to do a proper analysis, so I'm always a little reserved of all, I say. But that's because I'm a Wikipedian and that's the way we are, you know. But if you ask me to predict what it would be like and then you put that up against all the news reports I've read, I'm like, yeah, that's pretty much what I would have thought. So a lot of errors, a lot of hallucinations, certainly. Although Elon loves To promise that it will be more neutral than Wikipedia. It seems to interestingly agree with some of his more curious political views. And that's not most people's definition of neutrality. Like agreeing with Elon isn't the core of neutrality.

So I mean, unless you're Elon.

Unless you're Elon, yeah. And so, you know, it doesn't surprise me. I mean, I think it's an interesting concept, right? Could you do this? Could you use a large language model to generate quality content? And obviously we're all going to be grappling with that over the next few years in lots of different ways. So already we see in journalism there are bad uses and good uses that are taking place. Like one of, one of the ones I actually like is, I'll give two examples, but one is earnings report story. So a story about earnings report. I don't know to what extent it's large language models, but I know that a lot of the financial publications are using technology to semi automate making those stories. And it's fine like because basically those are very formulaic stories anyway. It's sort of like, oh, Analyst Expectations were two point, you know, $2.10 earnings per share, but it actually ended up $2.12 beating earnings. The stock responded thusly. You know, like it's pretty basic stuff and for humans to have to do that is really boring. And you know, it's like, yeah, great, like that's completely fine. I hope humans checking it because it might hallucinate and it might get it all wrong. But another example that I would argue could be useful is a certain type of sports story, the less interesting ones. But you've got in a, in the sports section of the paper you've got to have both like some pretty basic reporting. Like here's what happened in the game last night, like, here's the score, here's the, here's the step by step that's actually pretty tedious. Then there's the more interesting thing, the human flare, the angle. Like the funny thing that happened, the interesting bit like how did the players react? What are the fans saying? That you still need like proper journalism, but maybe pieces of that could be helped. And so that as we see that and then you know, like we know the financial model of journalism is under pressure, has been for a long time. There will be publications who want to stretch it a little too far and like, oh, you know, if only we didn't have to pay all these reporters, we could churn out twice as much content. And that's not going to end well. I think we're all going to be grappling with that sort of thing. And then we can also talk about bad actors who actually don't care at all about quality, our correctness and yeah, I could imagine them turning out lots of stuff that isn't very good.

Yeah, it does seem like there's a, a sort of economic trade off here. So you've got like cheap AI. Well, I guess a lot of AI is cheaper than humans for this kind of stuff. And then you got like difference in qualities of models. So like the really cheap models, they get it wrong a lot more time, but they're again, you saving money on, on your tokens and then fact checking is just expensive in general. So. Yeah, I guess then you pay a premium for having truth.

Yeah, yeah, No, I mean, I think, and I think that's inevitable because, you know, it's an amazing technology and I mean I use it all the time. I mean I'm playing with it, experimenting, using it for personal use. I think the key, as with everything. And I, I, I, as I say this, I, it echoes things that I remember saying about Wikipedia when it was a brand new technology, which is we don't, it's not about do we use it or do we not use it? It's about let's teach people how to use it. Like what are the good and the bad uses? So, you know, like sometimes you used to hear teachers would say to their students, don't even look at Wikipedia. And I'm like, well they're 100 going to look at Wikipedia. Like kids love it. You know, instead you should be saying, oh, here, Wikipedia, right? It's super interesting, it can be very valuable. But by the way, when it says the neutrality of this article has been disputed, they mean something about that. When it says the following section doesn't cite any sources, they mean something like that. If you read something that sounds surprising, it might be vandalism. Check the sources. Like, let's teach people the right way to use Wikipedia. Similarly with AI, Like, I use it like one of the things I love using it for is cooking. I love to cook. I fancy myself a decent cook. And you know, I ask it for recipes and I also ask it for links to, you know, famous cookery websites. And it has in the past absolutely hallucinated the URLs and you know, apologize to me that it doesn't know which websites are real or not. Like, great, okay. But I always warn people, I'm like, yeah, it's fun and Useful to do that, but I wouldn't recommend it if you're brand new to cooking. Like, I know how to cook. So when it gives me a recipe that has way too much salt, I have a bit of an instinct to go, ooh, that seems a bit much like maybe you're getting that wrong. Can I challenge you on that? But, you know, it's, it's very useful. I also use it for. I'm, I'm a programmer, but not a very good programmer, and I can absolutely see why. Stack Overflow, for example, has, as I've read, 10% of the traffic it had some years back. And it's because as a programmer, a bad programmer, you hit a lot of error message and you're like, oh, how do I fix this? And you Google it and you get to Stack overflow and somebody's discussing it and you get the answer. Whereas now, you know, Gemini, Claude, ChatGPT, they're all quite good at coding and they're quite good at explaining a basic syntax error or even a more subtle error. And it's just more efficient as a, as an end user. It's like, this is a better product for me. Like, I'm trying to write something and I'm getting it wrong. I need some help and I want to ask a question. It's, it's decent. I mean, although it is far too prone, in my opinion, to tell me to sort of rewrite JavaScript. I'm like, no, probably it's my bug. Probably it's not a bug in JavaScript. Like, that's probably not it. Like, I don't know why you're telling me this, because it's telling me also I'm a genius, you know, so it's.

Nice to be heard that you're a genius. But, yeah, chances are you probably made a mistake yourself.

Yeah, yeah, no, it's often like, oh, that's a very insightful question. I'm like, literally I have enough self respect to know that was not a, Like, I asked a really dumb question. And I'm okay with that. Right. You don't have to butter me up. Like, it's fine.

Absolutely. So, I mean, you mentioned that Stack Overflow traffic has decreased because people use all these AI tools. It said, how has Wikipedia been affected by this? Because I guess people want to look up, like facts about Tom Cruise's mom, but, you know, do they go to ChatGPT instead of Wikipedia?

Yeah, so there's a couple of things. So I, I would say not necessarily chatgpt, it's A different use case and it's a different paradigm and it doesn't directly, it may compete. Everything competes with everything in some sense. I say Google AI summaries is probably the more impactful, potentially. So. There was a Pew Research study not long ago that, that looked at for according to their study and I can't really vouch for it completely, but it seems plausible that in terms of like in search results, the first page, the, the classic 10 blue links, Wikipedia is cited on the first page 3% of the time. Sounded low to me, but it depends on what you're searching for. And in AI summaries we're linked to 6% of the time. So we're getting twice as many links to Wikipedia, but people click through half as often. So why do you not click through? Because, well, it answers your question. So you say, oh, who is Tom Cruise's mom? And then it goes so and so, and then there's a link to Wikipedia but you don't need to click it because you got the answer you needed. On the other hand, according to Google, when people do follow those links, they're much more engaged. So they didn't just click into Wikipedia to find out the name of the mom and then bounce. They actually like, oh, that's his mom's name. Oh, great. Also, I was thinking about this and so I'm gonna go read about Tom Cruise. So they stay longer. We haven't verified that, but that seems plausible to me. So as a result, that's kind of a wash. You know, getting twice as many links but clipped through half as often is kind of a wash. But we have seen, we just had a blog post about this recently, an 8% drop year over year in human visits to Wikipedia. That doesn't mean our traffic has dropped, but we try to analyse, you know, analyze bots versus non bots. And that's imperfect, of course, but we, we have some idea. And so that's concerning, but not, not a disaster at the moment. And then I think a part of that to understand is for Wikipedia, we're a charity and our business model, we don't have ads. So our business model is about as long as you know the information you got is from Wikipedia, then when you do see that banner saying, hey, use Wikipedia all the time, you should probably chip in, you'll respond and that will keep the lights on, that keeps us running. And so we're not so obsessed with page views. Like if we were, you know, funded by ads, obviously an 8% drop in page views is an 8% drop in revenue. And that doesn't really match for us. Like, what we really need is people's trust. We need people to love Wikipedia. We need people to say, yeah, Wikipedia, that's one of the good things should chip in. And so that's what we have to defend. Not so much just the page views. Obviously we care about page views to degree and all of that, but it's not our primary motivator.

Okay, that's reassuring. But you mentioned there's an increase in bot traffic, I presume. Is this all the foundation model companies scraping your site then? I mean, I know Wikipedia text is like, basically all of it is in most of the large language models these days.

Yeah, yeah, yeah. So all of the major models use Wikipedia quite heavily. And that's fine. That is what it is. Like, we're freely licensed and that's kind of. We're happy to be part of the infrastructure of the world. But one of the things we've come to understand, I've come to understand recently, other people got this long before I did, but is like, bot traffic versus human traffic is very different in terms of what it costs us. Why is that? How could that possibly be? And the answer is humans tend to visit the most popular pages and they, particularly, if something's big in the news, will get millions of views. So when the Queen died, obviously millions of people came to read about the Queen. Well, what does that mean? That means we, we build the page out of the database and we have it cached in memory and we just blast it out. Costs very little. Whereas if a bot comes and reads every obscure page of Wikipedia, well, we, we have to have more servers, either more RAM for caching, we have to have more database, you know, calls and, and construct pages and all that. It actually costs us a disproportionate amount. And what we feel is awkwardly unfair about that. It's not like it's wrecking us, but it's like, gee, the people who are giving an average of, as I understand I was told recently, $10.38 a year, like that's the average donation to Wikipedia. Well, the people chip it in their ten bucks or ten quid, ten euros, whatever it is, they're thinking, great, I love Wikipedia, I want it to survive. They're not thinking, I really want to subsidize OpenAI crawling Wikipedia, right? That's not what they're there for. And so that fact that it costs us more and it's being funded by ordinary people chipping in their money doesn't Seem right. So we have an enterprise product. Google is a good customer. So we're happy with Google, but we're still working on the other companies and we're not yet blocking them or things like that. We're quite easygoing and friendly and they're like, hey, you know, come and do the right thing. So hopefully, you know, they will pay their way. And that seems fair. Like, that's kind of like the fairness model of Wikipedia is like you use it all the time, you should probably chip in, your bot uses all the time, you should probably drip in a lot.

So yeah, I feel like Sam Alton can afford a $10.38 donation to keep.

I mean, a little proportionality would be good, but. Yeah, yeah, exactly.

All right. So it seems like, you know, negotiations with the foundational model companies are kind of ongoing. So I'd love to move a bit beyond Wikipedia actually, because it feels like a lot of the stuff we talked about with trust, these can be useful like, well beyond just creating an encyclopedia. So talk me through, like how do you use these principles or how can they be used like in day to day life in your work?

Well, I mean, I think, you know, one of the things that's sort of obvious to people is like in, in a business setting, trust is really important. It's a lubrication that makes everything a lot easier. Like if you are, I'll tell you, like here's, here's an example. You've seen stories that to me are just mind boggling about companies that are. When people work from home, they're using software to monitor that you're still moving your mouse and clicking on the keyboard. And I'm like, wow, like that level of mistrust of employees is just deeply wrong. And I think if you are an employee who works at a company like that, well, every time you're not jiggling your mouse to pretend to be at work, Please be on LinkedIn and looking for a job. Like, you don't want to work somewhere like that. They don't trust you and that's a hostile environment and it's just not good. Whereas, you know, if you can build a culture where you could say, you know, to the staff, like, yeah, work from home. I trust you. What really matters is we gotta get the work done. Like we've got customers, we got responsibilities here. We're gonna, you know, do this and I'm not gonna watch every single thing you do because that's like weird and annoying and wow, you know what, you'll get much better work out of people because you've built a healthy environment and people are like, yeah, actually they treat me properly, like a decent human being. So I'm gonna treat them back like that. I'm going to go the extra mile. We try to do the best for our customers. Like, we take pride in that. All of that kind of good stuff really works in an environment of trust. And when you don't have trust, everything starts to break down. Certainly if you imagine, say business to business partnerships, like there's a lot that is not written into a contract that is part of the relationship and part of the implied agreement. And that's good, right? Writing every single thing down and then ticking off to the letter. We're only going to do what we're contracted to do. That's like when the relationship's gone very, very wrong, like, that's pretty bad. And it, it means less success in business and so on. So, you know, all of these things, like, this is why trust really matters. And then, you know, how do you build it? Well, some of it's kind of obvious. I mean, we talk about transparency and make it personal and have a, have a good purpose, you know, so all of these rules are designed to be practical, to be things that you could check off. And like, one of our big recommendations in the book is for companies is take a trust inventory. Like really sort of sit down and go through the seven rules and go like, okay, what are the things that we maybe could be doing better? Where are we failing? Like, if we know, like one of the issues we have, a lot of companies have, this is like, oh, we need to make this really hard change and it's really tough. And maybe we're going to have to let some people go and so forth, but nobody trusts us. They think we're up to no good and so on and so forth. And okay, you're going to have a big problem. But if it's like, yeah, people will accept, like they're going to be unhappy they left, that they're losing their job, but the other people are going to be like, yeah, give us fair. The decision was made in a way that was reasonable and obviously it's a shame, but we're going to have to, you know, reconfigure the company. Like, wow, trust really matters a lot in those cases that you aren't being malicious and capricious and all of that. And so you've got to build it.

Absolutely. So it seems like the big places to watch out for this are when either when things have gone horribly wrong or they're about to go wrong. And that's the point where you really need to think about how do we make this process trustworthy? I mean, certainly like if you're having to make people redundant, it's like, well, are you doing it because the company finances are a mess or is it because like, oh, executive needs to hit their bonus this year. So yeah, you want to be.

Yeah, yeah, yeah. No, well then also, yeah, think about, yeah, that's like employees and the management and the company. But there's also like in a business to business situation where my company is supposed to deliver something for your company, but we're going to miss the target. Right. How are you going to react to that? And if we have a relationship of trust and I can go to you and say, yeah, unfortunately these things happen. We're going to be two weeks late. Are you going to go, you suck, this is terrible. I'm out of here, I'm going to find somebody else next time around. Are you going to be like, oh, okay, Like, I know you did your best. I can see the problem. Let's think about how we're going to solve this so it doesn't happen again. Like, you've got the trust of like, oh, you're a proper partner. Yeah, all of that super important.

Absolutely. Definitely better than just being like, okay, we're going to miss this. Like, let's not communicate with the customer. It goes to two weeks and yeah, yeah.

And that, that's exactly, exactly where you into a really big problem. Right. Because you didn't follow a basic principle of transparency of like, you know, like, oh, actually like one of the things is be transparent, especially when you have something to hide. That's one of the chapter headings, you know, and it's like, yeah, it's easy to be transparent when everything's going great. It's harder when it's like, oof. Actually you're going to have to go hat in hand and say, yeah, actually something bad has happened and we need to tell you about it. Like, that's harder. But it's actually the moment when trust is built. It's actually the moment when people can see it's like, yeah, it's easy to be, you know, like everything's fine, but actually to come in and say it when you know you'd rather not have to say it. Yeah, that's, that's incredible. It's a great moment to build a better relationship going forward.

Definitely a bit of a test of character, but it seems very worthwhile. All right. So I guess, are there any places where you think your rules about trust ought to be applied that aren't at the moment? Like, are there any areas where you think, okay, we need to really increase trust here?

Well, I mean, yes, there are, but I, I'm hesitating just a little bit, but I'll tell you why. So one of the areas that we have a big problem and one of the side effects, and we've been talking about the stuff I prefer to talk about, which is, you know, like practical matters in business and personal relationships and all that. But also we know that politics is an absolute mess these days and trust in journalism, trust in politicians is at all time lows and in many cases quite right. You know, we, we have. And one of the issues there is like whenever people are voting and they think, well, you know what? I'm not gonna, I'm not gonna rank my voting in any way based on whether I trust the person, because they're all liars and they're all dishonest. So I'm gonna use something else. I'm just gonn feelings about their claimed policies. Right, well, that's a problem. Like, that gets us into a world of mess. We need to actually, at the ballot box, punish people who aren't trustworthy. Like, don't vote for people if you can't trust them. And that the reason I hesitate is, well, like, okay, great, yeah, people might go, yeah, Jimmy, you're right. But that isn't like a proactive, simple step to like build trust. It's just like what we can do as consumers of politics. But it's an area where the breakdown of trust in society is a big problem. And I do think it makes sense to say let's be a bit more demanding about trust. Like, let's demand that politicians the same way we expect a company like, you know, like Amazon, okay, Amazon, I love Amazon. Amazon comes all the time to my house. Do I have complaints about, you know, sort of big abstract complaints about Amazon is going to be, of course, like, everybody could have something that they, you know, like, how are they treating people? Is it, is it okay? All that, fine, great. But you know what? When I order the thing on Amazon, it's going to come. And if it doesn't come and I contact Amazon, they're going to send me another one. Like, you trust them in that sense? Like, they just do the right business things. And I feel like in politics, we're not even sure they're going to do the right things. Like, you know, in the US There was just this longest ever government shutdown because they couldn't make a simple agreement on the budget. And it's like, okay, gang, you're not even doing the thing. Like, it's the one thing you're there to do. And you're all squabbling so much and turning everything into a culture war and being so vicious. You're not actually governing the country in any sensible way. How about a little compromise, a little give and take, a little, like, trust each other a little bit to say, like, you know what? I'm going to give you this one, but you're going to owe me next time, you know? Or I'll give you this, but you need to give me that. Okay, we can find a path forward. No, it's just like, stonewalling, breakdown. It's a lack of trust, and it's not good.

Okay, yeah, certainly. I mean, politics is very tricky in terms of trust. But I mean, I guess if Uber can turn it around. You talk about, like, ways to rebuild trust, maybe we can fix politics as well.

You know, there's a great quote in the book about a French leader who said, this is during the height of the Cold War. And I'm not naming the names because that's not important. It's like a French leader said of the U.S. president, well, if the president says it, that's good enough for me. Like, there was trust, even though they didn't agree on lots of things politically. And I remember when it was Obama versus McCain. And I remember, like, at the time, I wasn't thinking about trust and all that. I didn't have those words to use yet. But I. I said to lots of my friends, I'm like, you know what? This is great. I'm really proud of the country. Like, I have agreements and disagreements on policy with each of them. I know I'm gonna vote one way. But you know what? They're both proper people. Like, they're both sensible. They're not crazy radicals. They're not dishonest. They've got problems, whatever. You know, it's okay. Whereas more recently, I'm like, yeah, this is. Isn't okay. Like, this isn't just like, oh, I have some disagreements of policies. I'm like, this is not proper behavior. And, yeah, and I live now in the UK and, you know, I'm the first to say in the last election, it's. And I don't want to bore people who aren't in the uk, but I will anyway. But in the last election, the Tories lost and I think they lost more than labor won because they had lost the trust of the country. Not about policy, not about this and that. It was about that party in Downing street when we were all feeling for the Queen sitting all alone the next day at her husband's funeral. And we find out later, meanwhile, the night before, while the Queen is sort of in isolation because of COVID and they're having parties in Downing street, like, this is not okay. Like, you're not behaving like proper people. And I think that feeling was more important to voters. Maybe I'm wrong. Maybe some political scientists will do surveys and prove me wrong, but I bet I'm right. It meant more to people than the specific details of policy. Like that. It was just like, I don't trust you anymore to be doing the right things. You're just having a laugh at our expense. It's not okay.

Absolutely. Yeah. So I guess really, I mean, the lesson there is like, you, you've got to behave as a trustworthy human if you want people to respect you and they won't want you to trust.

Yeah, yeah, yeah. When you. When you go out driving during COVID don't say, I had to go driving with my children in the back to test my eyesight. I don't know. That doesn't seem like a plausible answer to me. Sorry.

Absolutely. Yeah. Maybe just. Yeah. Don't lie to the general public. Really good tip for building trust. All right, super. So just finally, I always want new people to learn things from. Who are you following right now? Whose work are you interested in?

I'll tell you who I'm interested in. I'm afraid to say I'm not sure I can follow his work directly because he's too smart and all of that. But I really am a big fan and a friend of Demis Hassabis. So Demis is the. The CEO of DeepMind, Google's big AI, you know, unit, and he just won the Nobel Prize in chemistry for the work on protein folding. But, like, the stuff that. That DeepMind are doing, and it's beyond just the commercial large language models and all this stuff which they are at the forefront of. But I'm really fascinated about, you know, all of the amazing possibilities, the exciting, positive things that are going to come out of, you know, using AI techniques, using large, like machine learning to crack some really hard problems in biology and health. I think it's an amazing time to be alive and it's an amazing time to maybe live a little longer and a little bit happier and a little Bit healthier. And so I think that stuff is really fantastic. And so it's the kind of thing that although for my professional work, I have to think a lot about large language models and stuff that's directly sort of coming into my world. I'm keeping one eye on the other stuff. Driverless cars is another example. Like, I'm fascinated. Like, I just think, you know, the idea that, you know, in. I'll tell a funny little story. When my kids were little, I have a Volvo. I'd just gotten my brand new Volvo. This is several years ago now. It's now my not so brand new Volvo. And the car beeps a lot. Like it's very. The kids call it a scaredy cat. But also I said to them, yeah, actually like, I'm in a parking lot, if I get too close to another car, it will slam on the brakes. And they said, so I don't reckon anything. And they, One of them said, but don't all cars do that? Like. Cause they're little kids, they're like, obviously, like cars shouldn't like wreck into stuff. Like, why would a car wreck into stuff? It's got cameras. Like, we know it's got cameras. A lot of cars have cameras. Why do they wreck into stuff? It's like, yeah, you're right. And so I envision, you know, in not that far away, not as quickly as Elon promised us. Like, it's a little bit speaking of trust, you know, but in a very near future, like the idea of dying in a car accident, which is a quite common thing, or being permanently, you know, life changing injuries, is going to seem like really old. Like, really. It's like, wow, like, you know, dying of scurvy or something. You know, like something that just doesn't happen anymore because we figured out like how to solve it. That's exciting to me. And so I do watch that as well. Like I'm really interested in the technology around automated vehicles and the possibilities for safety and sort of better travel and transport.

So I love that. I guess there's a common theme there in like solving causes of death that seems. Yeah, definitely.

Yeah, it's a good thing. Yeah, let's do that. Let's. Let's have a lot more of that and a lot less yelling at each other on Twitter, you know.

All right, I think it's a wonderful way to wrap up then. So. Yeah, thank you so much for your time, Jimmy.

Yeah, lovely, brilliant. Thank you, Sam.