Latest / Giant Ideas / Wikipedia Founder, Jimmy Wales: What Can You Trust In The Age of AI?

Transcript

Wikipedia Founder, Jimmy Wales: What Can You Trust In The Age of AI? · Giant Ideas

• 0:00
  SPEAKER_00: You know, somebody said to me, Oh, the VCs must love this, that you know, all these people just do the work for free.
• 0:05
  SPEAKER_00: Actually, the funny thing is, VCs are the only people who ever said, Why don't you pay people?
• 0:09
  SPEAKER_00: They're like, What if you paid people like a dollar per article or something like that?
• 0:13
  SPEAKER_00: Wouldn't you get more articles?
• 0:14
  SPEAKER_00: And I was like, get more crap, that's for sure.
• 0:17
  SPEAKER_02: Hello, and welcome to Giant Ideas with me, Cameron McLean.
• 0:21
  SPEAKER_01: And me, Tommy Stadlin.
• 0:22
  SPEAKER_01: We're co-founders of Giant Ventures, which builds and backs purpose-driven companies.
• 0:26
  SPEAKER_02: At Giant, we're lucky to meet extraordinary people with giant ideas that are changing the world.
• 0:31
  SPEAKER_02: This podcast brings you behind the scenes access to those ideas and the inspiring stories of the people behind them.
• 0:37
  SPEAKER_01: We explore how one giant idea can kickstart a billion-dollar company, shape culture, and transform life as we know it.
• 0:47
  SPEAKER_02: Today on the podcast, we are joined by Jimmy Wales.
• 0:50
  SPEAKER_02: Jimmy, along with millions of volunteers, created a fundamental part of the internet: Wikipedia, a crowdsourced encyclopedia which became the foundation of the internet as we know it today, with 11 billion viewers a month.
• 1:03
  SPEAKER_01: Describing himself as pathologically optimistic, he did something that most people considered impossible.
• 1:08
  SPEAKER_01: A gift of the world that required selflessness, organization, and a deep sense of purpose.
• 1:14
  SPEAKER_01: His latest book, The Seven Rules of Trust, talks about what made that possible and how others can learn from it.
• 1:20
  SPEAKER_01: Today we'll talk about Jimmy's journey of building Wikipedia and trust in the age of AI.
• 1:25
  SPEAKER_01: As Stephen Frey says, Everything in our thrilling and chilling future depends on one increasingly diminishing human resource trust.
• 1:34
  SPEAKER_01: Jimmy Wales is in a better position than almost anyone to teach the world the most important lessons.
• 1:40
  SPEAKER_01: Jimmy, welcome to Giant Ideas.
• 1:42
  SPEAKER_01: Good to be here.
• 1:43
  SPEAKER_01: Thanks for coming.
• 1:44
  SPEAKER_01: So let's start by talking about Wikipedia.
• 1:45
  SPEAKER_01: So many lessons that I'm sure you've learned and the world has learned from the founding of Wikipedia.
• 1:50
  SPEAKER_01: One curious uh element of it, or very interesting element of it, is that you decided to start it as a nonprofit.
• 1:55
  SPEAKER_01: We interviewed Tim Berners-Lee, the creator of the World Wide Web, who created the web for nothing.
• 2:00
  SPEAKER_01: You created a huge part of the World Wide Web, again, without personal gain.
• 2:06
  SPEAKER_01: Do you do you personally regret any of that?
• 2:08
  SPEAKER_01: Are you happy you did it?
• 2:09
  SPEAKER_01: And then maybe more importantly for the world, do you think it's been a really good thing for Wikipedia that it has been a nonprofit?
• 2:14
  SPEAKER_00: Yeah, no, I don't have any regrets.
• 2:16
  SPEAKER_00: I mean, it's uh it's an amazing thing to be a part of.
• 2:19
  SPEAKER_00: And uh my life is unbelievably interesting and fun.
• 2:23
  SPEAKER_00: I also have Fandom, my for-profit wiki company, which has done very well, a couple hundred million revenue and all that.
• 2:28
  SPEAKER_00: And so that's fine.
• 2:29
  SPEAKER_00: I'm not poor.
• 2:30
  SPEAKER_00: Uh and um the um uh you know, the the meaning of Wikipedia, you know, in the long run, I think in 500 years, people will look back on this era and say, yeah, that that was a bit of a crazy, horrible time.
• 2:46
  SPEAKER_00: And yet these nice people were really good and they were sharing knowledge, and and that seems like a useful thing to have done.
• 2:52
  SPEAKER_00: And how is it impacted us?
• 2:54
  SPEAKER_00: I mean, I think the the most important thing is actually uh not necessarily for profit versus nonprofit, although I do think that is relevant.
• 3:03
  SPEAKER_00: Um, but it's it's actually more broadly the business model.
• 3:07
  SPEAKER_00: So the even if you're a nonprofit, you have to have a business model.
• 3:10
  SPEAKER_00: You've got to pay the bills and you've got to sort of um you know uh make it work.
• 3:15
  SPEAKER_00: And if we had advertising as a business model, I think that would that would be pretty problematic in lots of ways.
• 3:23
  SPEAKER_00: Uh, you know, a lot of the stuff that we've seen happen to purely ad-supported journalism, uh, where there is a really strong temptation to have clickbait headlines and to think about viral things and and all of that, which is fine in certain areas, but really would be a bit odd for Wikipedia.
• 3:45
  SPEAKER_00: It'd be problematic, I would say.
• 3:46
  SPEAKER_02: So sort of a purity to the Wikipedia approach allows you to build some of that trust, I presume.
• 3:52
  SPEAKER_00: Yeah, and and and the organization, so the the Wikimedia Foundation can remain very, very uh mission-driven.
• 3:59
  SPEAKER_00: And so, you know, our our goal is a free encyclopedia for every single person on the planet in their own language.
• 4:06
  SPEAKER_00: That's how we make decisions.
• 4:08
  SPEAKER_00: And so when we have you know decisions to make, which even companies that actually do care a lot about that same kind of value system, I would say Google does.
• 4:17
  SPEAKER_00: But Google has to prioritize, of course, the markets where there's big ad revenue.
• 4:21
  SPEAKER_00: You know, they're they're good about trying to support as many languages as they can and things like that.
• 4:25
  SPEAKER_00: But obviously, you know, it's like the next million uh users of Google in California are going to be much more interesting and more profitable than the next million in Kenya.
• 4:37
  SPEAKER_00: Sure.
• 4:37
  SPEAKER_00: Just because that's the way the market is.
• 4:39
  SPEAKER_00: Um whereas for us, it's like, well, that hardly matters.
• 4:42
  SPEAKER_00: And indeed, for our business model, which is donation-based, it's probably those next million in Kenya who people are really more excited about.
• 4:50
  SPEAKER_01: Having built Wikipedia as a nonprofit, you've kept it as a nonprofit.
• 4:54
  SPEAKER_01: Famously, OpenAI started as a nonprofit and is basically not really a nonprofit anymore, for all intents and purposes.
• 5:00
  SPEAKER_01: What do you make of that situation?
• 5:02
  SPEAKER_01: And how do you think they should proceed?
• 5:05
  SPEAKER_00: Well, I it's uh it's very complicated.
• 5:08
  SPEAKER_00: And I I've sort of made a uh sort of amateur study of it, as I think many of us in the industry have.
• 5:14
  SPEAKER_00: It's like, what the hell what the hell is going on over there?
• 5:16
  SPEAKER_00: Like it's very like, how do you do that?
• 5:19
  SPEAKER_00: Um so it's really, you know, uh, it's still, I think, technically owned by the nonprofit at the top of it, but it's got this sort of capped return, blah, blah, blah.
• 5:28
  SPEAKER_00: You know, it's very complicated.
• 5:29
  SPEAKER_00: Um, and apparently Sam Altman doesn't own any shares, which is also quite interesting.
• 5:33
  SPEAKER_00: I don't know what to make of that either.
• 5:35
  SPEAKER_00: So I think one of the keys is um, you know, basically, if you think about what is Wikipedia, Wikipedia is often reported as a tech story, but the truth is the actual tech infrastructure is minimal.
• 5:49
  SPEAKER_00: Like it's really, you know, if you think about okay, what's the technology of Wikipedia?
• 5:53
  SPEAKER_00: Fundamentally, it's uh web server, web browser, database, you know, the idea of Wiki.
• 6:01
  SPEAKER_00: Um, it's very basic on all of those things.
• 6:04
  SPEAKER_00: And it's really a community story.
• 6:05
  SPEAKER_00: It's about the people, it's about the humans who are building this thing.
• 6:09
  SPEAKER_00: And so, therefore, the cost um is not nearly what it costs to do sort of frontier AI models.
• 6:17
  SPEAKER_00: And so I will say, like, okay, to do open AI as a nonprofit, I don't, I can't think of a single way, if it were set up as a charity, that they would say, well, okay, we need to invest 120 billion in GPUs.
• 6:32
  SPEAKER_00: Yeah, good luck raising that kind of money, unless there's going to be some kind of return from investors.
• 6:37
  SPEAKER_00: Like that, that's sort of obvious.
• 6:38
  SPEAKER_00: Whereas for Wikipedia, you know, we're trying to raise whatever, a couple hundred million, uh, and we've got a massive, massive audience, you know, it's doable and it's always been doable all along.
• 6:48
  SPEAKER_00: I mean, one of the great things about Wikipedia is it's really bootstrapped in the in that sense.
• 6:53
  SPEAKER_00: You know, it's like we've never had to go for outside capital and and so forth.
• 6:57
  SPEAKER_00: It's always just been donations and largely small donors.
• 7:00
  SPEAKER_00: So there is a difference.
• 7:02
  SPEAKER_00: So it's you know, it's very easy for people to sort of tisk tisk at uh open AI, but I'm like, well, you know, go ahead and start a charity and raise two million dollars and see what you can do with API.
• 7:13
  SPEAKER_00: I mean, good luck with with um uh not API GPUs, you know, it's gonna be quite hard.
• 7:19
  SPEAKER_02: We've talked a bit about your incentives starting a nonprofit versus a profit, but thinking about the incentives of the users of Wikipedia.
• 7:27
  SPEAKER_02: So you've created arguably probably the biggest historical body of knowledge, or just certainly one of them, without financial incentives for those contributing.
• 7:36
  SPEAKER_02: So I'm you've had a front row seat for the last 20 years.
• 7:38
  SPEAKER_02: What have you learned about, I guess, human motivation and and and why people are doing this?
• 7:43
  SPEAKER_00: Yeah, and I think this is this is really huge.
• 7:45
  SPEAKER_00: And um, you know, so so fandom is wiki communities as well, and it's ad supported and and so on, and it's it's mainly about entertainment, gaming, and so forth.
• 7:56
  SPEAKER_00: And so, you know, one one of the first hypotheses that I had, but I wasn't sure about is would people be willing to contribute to, I'll give an example, the Muppet Wiki, right?
• 8:08
  SPEAKER_00: Which you can't really set up as a charity because you can't really convince the government that's a charitable project, but it's a community project.
• 8:14
  SPEAKER_00: So would people contribute?
• 8:16
  SPEAKER_00: Um, would they contribute unless maybe like social media influencers, maybe they need some sort of deals or whatever?
• 8:23
  SPEAKER_00: It turns out they do.
• 8:25
  SPEAKER_00: Uh, it's because they're they love, they have a passion about some area of knowledge or something that they're interested in.
• 8:32
  SPEAKER_00: They have oftentimes a community motive that you wouldn't think of as charitable.
• 8:36
  SPEAKER_00: So people who are working in a lot of uh gaming wikis, they're they're quite proud of their knowledge and they're quite happy to help other people like them who are trying to get through that very complicated level on the game and so on and so forth.
• 8:49
  SPEAKER_00: So it's interesting that um, you know, the you know, when I was raising money for fandom, uh you know, somebody said to me, Oh, the the VCs must love this, that you, you know, all these people just do the work for free, and so on.
• 9:04
  SPEAKER_00: I'm like, actually, the funny thing is, VCs are the only people who ever said, why don't you pay people?
• 9:09
  SPEAKER_00: They're like, what if you paid people like a dollar per article or something like that?
• 9:13
  SPEAKER_00: Wouldn't you get more articles?
• 9:14
  SPEAKER_00: And I was like, you get more crap, that's for sure.
• 9:17
  SPEAKER_00: And actually, that's another piece of it.
• 9:19
  SPEAKER_00: It's like the people writing Wikipedia aren't doing it for very low pay.
• 9:25
  SPEAKER_00: If you want good quality writing, uh, you can either pay good journalists a proper salary, right?
• 9:32
  SPEAKER_00: Or you can get people who are really passionate about it to do it out of the love of what they're doing.
• 9:37
  SPEAKER_00: But if you pay people a very low amount, you're just gonna get a bunch of nonsense.
• 9:40
  SPEAKER_00: There was a company called Demand Media, which had a little flurry of a splash.
• 9:44
  SPEAKER_00: I don't know whatever happened to them, but haven't heard about them in a long time.
• 9:47
  SPEAKER_00: But they basically were using algorithms to detect like what people were searching for, and they would find quirky search terms that had no coverage, and then they would pay people like a dollar or something, you know, to write, you know, uh, instead of just an article about how to wash your dog, it'd be a specific how to wash a Pomeranian.
• 10:06
  SPEAKER_00: Uh and it the content was terrible uh because people were like, oh, I'll take a dollar.
• 10:11
  SPEAKER_00: I know nothing about dogs or Pomeranians, but I could I need a dollar, so I'll do it.
• 10:15
  SPEAKER_00: And it it just wasn't good.
• 10:16
  SPEAKER_00: Whereas the Muppet Wiki, let me tell you, very, very good.
• 10:19
  SPEAKER_02: People care about Muppets.
• 10:20
  SPEAKER_02: Uh yeah.
• 10:21
  SPEAKER_00: And you know, the same is true.
• 10:22
  SPEAKER_00: I mean, that's it's a fun example, but uh, you know, it particularly around geek culture, you can imagine the people who are really deeply knowledgeable about Lord of the Rings and things like that.
• 10:33
  SPEAKER_00: And then the same for Wikipedia.
• 10:34
  SPEAKER_00: A lot of the people involved in Wikipedia are they're motivated, yes, by the charitable vision, yeah, free encyclopedia for everyone.
• 10:42
  SPEAKER_00: That sounds like a worthwhile way of spending your time.
• 10:45
  SPEAKER_00: But they're also very motivated about their passion around a certain topic.
• 10:49
  SPEAKER_00: So, you know, people who are writing about trains or, you know, whatever it might be, um, they're they're deeply knowledgeable.
• 10:57
  SPEAKER_00: And I, you know, just count me as any ordinary person who I sometimes stumble across a Wikipedia page and I'm like, who are these people?
• 11:04
  SPEAKER_00: It's unbelievable.
• 11:05
  SPEAKER_00: Like, you know, it's like here's some very obscure airplane um, you know, that was in production from 1934 to 1938, and there's this massive article about all the details, including specific airframes and which ones are still flying.
• 11:21
  SPEAKER_00: And I'm like, wow, this is fantastic.
• 11:23
  SPEAKER_00: But you wouldn't, you couldn't pay anybody to do that.
• 11:26
  SPEAKER_00: And it it wouldn't make sense.
• 11:28
  SPEAKER_00: But what does make sense is having a a community that's welcoming to that person who's like, hey, I I know about this, I'm obsessed.
• 11:38
  SPEAKER_00: Uh I'm not gonna apologize for obsession, I'm just gonna go with it.
• 11:40
  SPEAKER_02: So a combination of sort of contributing to the community, but also I guess some self-fulfillment from you know being an expert in something.
• 11:47
  SPEAKER_00: Yeah, and and you know, one of the interesting things is you know, it it's people uh you know, human beings do like recognition from other human beings.
• 11:57
  SPEAKER_00: That visibility is psychologically valuable.
• 12:00
  SPEAKER_00: And in some cases, that uh can be a bad thing.
• 12:05
  SPEAKER_00: Like a lot of times, uh, the way social media is designed, the way to get visibility and attention is to behave badly because you get more followers that way and this, that, and the other.
• 12:15
  SPEAKER_00: Uh the you know, the more positive side of it is people who get a huge following because they're actually interesting and they've got good ideas, or they are attractive in, you know, sort of visual social media and things like that.
• 12:29
  SPEAKER_00: But for the Wikipedians, nobody's trying to get famous uh and be an influencer and get a sponsorship deal or anything like that.
• 12:35
  SPEAKER_00: But they still do care about the um the feedback from other human beings.
• 12:42
  SPEAKER_00: But actually, you know, the person who's obsessively writing about the Muppets doesn't really care what any of us think.
• 12:49
  SPEAKER_00: They just want the other Muppet fans to sort of say, oh wow, you're you're amazing, you know.
• 12:53
  SPEAKER_01: And who are these people?
• 12:54
  SPEAKER_01: Are we talking about a very small number of people who create 80% of it, or is it actually across because it you have to be so niche in these areas to be able to write about it.
• 13:02
  SPEAKER_01: Is it a vast number of writers?
• 13:03
  SPEAKER_00: Aaron Powell It's it's it's vast, but it it's it's narrow or it's hard to really answer that question.
• 13:10
  SPEAKER_00: So I I would say um, you know, there's the really, really active contributors.
• 13:18
  SPEAKER_00: So the people who are the most active are actually probably not the people who are obsessive about a niche, they're obsessive about Wikipedia.
• 13:24
  SPEAKER_00: Right.
• 13:24
  SPEAKER_00: So they're doing work across the whole thing and they're checking references and they're monitoring incoming stuff or bad behavior and there being admins and things like that.
• 13:34
  SPEAKER_00: Um but yeah, it I mean, one one of the things, a lot of people have the idea of Wikipedia that it's 10 million people adding one sentence each, but it's really is a much more narrow group of people who are really uh very focused on Wikipedia and the the standards and the rules and how it all works.
• 13:52
  SPEAKER_02: Aaron Powell Well, that's I think a great segue.
• 13:53
  SPEAKER_02: I mean, Wikipedia promises a neutral point of view, which sounds ideal, but in reality, is it just a reflection of who shows up, who argues the most, who cares the most, and reflective of that group that you described?
• 14:05
  SPEAKER_02: Because that is, I guess, one of the recent criticisms from some people on the Sure.
• 14:09
  SPEAKER_00: So it's it's a it's a bit of a yes and a no to answer that.
• 14:13
  SPEAKER_00: So uh on the one hand, uh, yes, of course, we all have our biases and we know about the things we know about.
• 14:22
  SPEAKER_00: And, you know, very often, you know, one of the things I I say about bias is that it's often like um if you ask a fish about water, the fish will say, What water?
• 14:33
  SPEAKER_00: Uh, because you're just swimming in it.
• 14:35
  SPEAKER_00: Like you don't really even realize you're biased in many cases.
• 14:38
  SPEAKER_00: And that's because of the media you consume and the culture you live in and so on and so forth.
• 14:42
  SPEAKER_00: So that's unavoidable.
• 14:44
  SPEAKER_00: Kind of unavoidable.
• 14:46
  SPEAKER_00: The other bit is to say, like, if you have people who have a genuine commitment to neutrality, they can definitely overcome their own personal viewpoint and say, actually, I do understand uh, you know, like we're not here to settle the argument, we're here to present all sides of the argument in a fair way.
• 15:05
  SPEAKER_00: And people are actually pretty good at that.
• 15:07
  SPEAKER_00: Not everybody.
• 15:07
  SPEAKER_00: And some people find it very uncomfortable.
• 15:09
  SPEAKER_00: And some people are good at it in some areas and not in other areas.
• 15:13
  SPEAKER_00: My, I mean, my personal example is I don't ever attempt to edit anything about Donald Trump because the man makes me insane and I couldn't be a good Wikipedian.
• 15:22
  SPEAKER_00: Uh, but broadly there are people who are quite good at it and and who who do you know work to achieve neutrality.
• 15:30
  SPEAKER_00: And so is it perfect?
• 15:31
  SPEAKER_00: No.
• 15:32
  SPEAKER_00: Uh it can be imperfect because people aren't even realizing that they're being biased because they haven't really heard any other perspectives.
• 15:39
  SPEAKER_00: Um, it can be imperfect because of very, if there's something that's very emotional, uh, that makes it a lot harder.
• 15:47
  SPEAKER_00: So where I see problems around neutrality is generally um, you know, if there's a war on, uh, particularly if Wikipedians have been killed and things like that, yeah, we're human beings.
• 15:59
  SPEAKER_00: Like it's hard.
• 16:00
  SPEAKER_00: Uh, even when we have a commitment, even when the community wants to be neutral, yeah, it's kind of hard to hold on to that um and in those in those times.
• 16:09
  SPEAKER_00: And so, you know, the idea that, as Elon says, Wikipedia has become Wokipedia, no, it hasn't.
• 16:17
  SPEAKER_00: Like, I I know the Wikipedians, I read Wikipedia a lot and say, look, that's just not true.
• 16:22
  SPEAKER_00: Uh, I mean, one sort of little example, there's a a classic kind of uh gotcha question around the whole trans issue, what is a woman?
• 16:32
  SPEAKER_00: And I don't know what they think the answer is going to be, but somebody started coming at me with this question on X, you know, and I was like, and I knew a little bit about this, but you know, and the and the the answer, there was a like a right-wing documentary about this, and the like the big punchline is a woman is an adult human female.
• 16:50
  SPEAKER_00: And they did get some people on that documentary who are not quite willing to go there and say it and sort of waffled in this and that, but I'm like, I can just tell you what it says in Wikipedia, adult human female.
• 17:00
  SPEAKER_00: Like then further down in the article, yeah, of course, it addresses like the questions of trans and that there's a social question and the issues and da-da-da-da.
• 17:07
  SPEAKER_00: And it's all like, yeah, that's fine.
• 17:08
  SPEAKER_00: Like, I'm like, if you're coming at me over that one, I'm like, yeah, we've we've not become you know crazy woke.
• 17:14
  SPEAKER_00: Um on the other hand, if you say, yeah, look, Wikipedia has problems of bias, um, I'm like, yeah, actually, I think we probably do.
• 17:21
  SPEAKER_00: And I think that's something we always have to have, you know, to examine and to be open about.
• 17:28
  SPEAKER_00: You know, in my in my book that just came out, Seven Rules of Trust, one of the ones I think is really, really important, particularly in my personal context at this moment in time, is uh be transparent even when you have something to hide.
• 17:43
  SPEAKER_00: And, you know, sometimes like the comms team at the Wikimedia Foundation get a little uncomfortable with the way I approach this, but I'm like, no, like if we just go with a PR front and say, no, we have absolutely no problems with bias, if you think that's wrong, that's because you're a right-wing lunatic.
• 17:58
  SPEAKER_00: Yeah, no, that's that's that's not the right way to answer it.
• 18:00
  SPEAKER_00: The right way to answer is to go, yeah, you know what?
• 18:02
  SPEAKER_00: Actually, there's a few areas where I don't think we quite live up to it.
• 18:06
  SPEAKER_00: And I think we need to work on that.
• 18:08
  SPEAKER_00: There are definitely areas where the media is biased, uh, and therefore Wikipedia ends up reflecting that.
• 18:16
  SPEAKER_00: How severe is that?
• 18:18
  SPEAKER_00: Good.
• 18:18
  SPEAKER_00: Let's debate that.
• 18:19
  SPEAKER_00: I don't know.
• 18:20
  SPEAKER_00: Uh, usually what I find is that when the media is very biased on an issue, the Wikipedians tend to tone that down.
• 18:27
  SPEAKER_00: Uh, so actually, we're better than the New York Times if the New York Times is pursuing an agenda or something.
• 18:33
  SPEAKER_00: But it's complicated and it's quite hard.
• 18:35
  SPEAKER_00: And actually, you know, there's a lot of detail under the surface.
• 18:38
  SPEAKER_00: You know, like one of the problems that we have right now is um a lot of the right-wing media is quite independently of being right wing, also low quality.
• 18:48
  SPEAKER_00: Uh, and that's a problem.
• 18:50
  SPEAKER_00: You know, like if if you if you're really being serious and thoughtful and you say, okay, right, I've got uh, you know, a story in the New York Times and the Washington Post, and I've got a story in Breitbart, you're not gonna go, well, those are about equal.
• 19:04
  SPEAKER_00: Like one of them is actually not very good.
• 19:06
  SPEAKER_00: It's like tabloidy and sort of lots of misleading information and a lot of polemics and so on.
• 19:12
  SPEAKER_00: And the other are, yeah, slightly left-leaning newspapers, which you can grapple with.
• 19:16
  SPEAKER_00: You know, so that's Rama.
• 19:17
  SPEAKER_00: And so my my thing, if I you know, say to Elon, like, why don't you fund a newspaper?
• 19:22
  SPEAKER_00: You know, like a seriously, a serious intellectual effort to put forward the ideas that you agree with.
• 19:28
  SPEAKER_00: That would be far more helpful to the world than beaten on me.
• 19:32
  SPEAKER_01: So we live in a kind of era of fake news and there's a lot of concern about trust in AI.
• 19:39
  SPEAKER_01: How do you feel about where we're going with with AI and whether people can trust it?
• 19:43
  SPEAKER_01: How can we learn from Wikipedia to foster more trust?
• 19:46
  SPEAKER_00: Yeah, so I mean, huge question.
• 19:48
  SPEAKER_00: Uh there, there's so many different elements to that.
• 19:51
  SPEAKER_00: So, you know, clearly right now there's a problem with AI and trust because of the hallucination issue, which is still quite severe.
• 20:01
  SPEAKER_00: And uh, you know, as new models come out, um, you know, Sam Altman is, I would say, a bit of an offender.
• 20:08
  SPEAKER_00: I like Sam, I know Sam, but he'll hide, he'll say things like, This is unbelievable, we've solved the hallucination issue.
• 20:15
  SPEAKER_00: I'm not sure he's ever saying we've solved it, but he'll he'll hint in that direction.
• 20:18
  SPEAKER_00: I'm like, oh great, that sounds fantastic.
• 20:20
  SPEAKER_00: And then I start using it, I'm like, yeah, actually, it's still bullshit.
• 20:22
  SPEAKER_00: It's like a night, I my my latest uh, which will only really make sense, I think, to British people.
• 20:28
  SPEAKER_00: Um it's like a it's it's very intelligent, very knowledgeable, a complete bullshitter.
• 20:36
  SPEAKER_00: I'm like, it's a 21-year-old Boris Johnson.
• 20:39
  SPEAKER_00: That's gonna be part of him as an intern.
• 20:44
  SPEAKER_02: And uh, you know, and and so I think you had a great example of a hallucination if um when we when we had you at uh at our event, John Ideas, if you could share that one.
• 20:54
  SPEAKER_00: Oh, yeah, yeah, yeah.
• 20:54
  SPEAKER_00: So I mean one of my one of my favorite things to do, and I think if I keep talking about it, it'll get into the training cater and then it'll ruin it, but uh is um I every new AI model that I test, and I do a lot, I do a lot of local AI because I'm really fascinated by what's going on in the open source world.
• 21:10
  SPEAKER_00: And I always ask, uh, who is Kate Garvey, my wife?
• 21:14
  SPEAKER_00: And it's she's kind of perfect for this because she's not a famous person, but she is known to a degree.
• 21:23
  SPEAKER_00: And she she worked for Tony Blair for 10 years and she promotes the sustainable development goals, and she's been in the press a little bit here and there, but not a famous person.
• 21:34
  SPEAKER_00: And so it it always sort of BS and it and it says things that are generally quite plausible, uh, but not true.
• 21:44
  SPEAKER_00: So once it said uh that she um that she set up a nonprofit organization, yeah, true, uh, to promote women's empowerment in the workplace.
• 21:54
  SPEAKER_00: Definitely not true, but it sounds like Kate.
• 21:56
  SPEAKER_00: Um and that she did this with uh Miriam Gonzalez, who's nicknamed.
• 22:00
  SPEAKER_00: Clegg's wife, um, which who we know.
• 22:03
  SPEAKER_00: I mean, I know Nick because of Meta, and you know, we've worked together on whatever there, and um, and our kids go to the same school and you know, like whatever small town London.
• 22:13
  SPEAKER_00: And uh I showed Kate this and she's like, wow, like that could have happened.
• 22:17
  SPEAKER_00: Like, like had we been sitting together at a dinner party, you know, at a certain time, like that could have actually happened.
• 22:23
  SPEAKER_00: And that's typically yeah, it typically it's things that are plausible.
• 22:27
  SPEAKER_00: And then I always the next question I ask is, and who did she marry?
• 22:30
  SPEAKER_03: Yeah.
• 22:30
  SPEAKER_00: Um, and that's also generally plausible.
• 22:33
  SPEAKER_00: It's it's because she worked in politics and is sort of a London person, it's generally uh members of parliament uh or political journalists in the UK.
• 22:45
  SPEAKER_00: Uh, but my favorite is actually not that plausible.
• 22:48
  SPEAKER_00: I would say it it said uh that she married uh Peter Mandelson.
• 22:52
  SPEAKER_00: And then I said to this was Chat GPT, and I said, But isn't Peter Mandelson you know quite famously gay?
• 22:59
  SPEAKER_00: And then it got very woke with me, and it's it's not appropriate to speculate about people's personal sex lives.
• 23:04
  SPEAKER_00: And gay people can get married in the United Kingdom.
• 23:06
  SPEAKER_02: It's like, yeah, okay, not really my point, but amazing the Jimmy Wells test, like the Turin test for solutions or so yeah.
• 23:14
  SPEAKER_00: So when you when you say that and you you think, okay, wow, could Wikipedians use ChatGPT to help write Wikipedia?
• 23:21
  SPEAKER_00: It's like, well, no, like if you say, um, you know, write an encyclopedia article about Taylor Swift, who is a famous person, it actually, you know, it's actually not bad, although there will be a few zingers in there in my experience.
• 23:37
  SPEAKER_00: Uh, but if you the more obscure the topic, the more likely it is to hallucinate.
• 23:42
  SPEAKER_00: Right.
• 23:42
  SPEAKER_00: And we don't need help writing about Taylor Swift.
• 23:44
  SPEAKER_00: We've got that covered.
• 23:45
  SPEAKER_00: You know, it's the it's the more obscure things.
• 23:47
  SPEAKER_00: Uh actually I I recently looked, and I don't know how often they update, but I looked at Grocopedia, uh, Kate Garvey, uh, as an interesting.
• 23:55
  SPEAKER_00: I was checking because she just got an OBE, like a British honor, and um Wikipedia was updated, not because it was her, but because we have some obsessives who, when the list comes out, they immediately go and update everybody.
• 24:07
  SPEAKER_00: Uh Wikipedia was updated within a couple of hours.
• 24:10
  SPEAKER_00: And I was like, oh, I wonder how quickly they will pick that up.
• 24:13
  SPEAKER_00: And they didn't pick it up.
• 24:14
  SPEAKER_00: But also, I it's the first time I'd actually read it.
• 24:16
  SPEAKER_00: And it was it was wild.
• 24:18
  SPEAKER_00: It was it was much longer than the Wikipedia entry, sort of longer than it should be because she's not a famous person.
• 24:23
  SPEAKER_03: Yeah.
• 24:24
  SPEAKER_00: But further down, it it started to go on and on, sort of in a very AI sort of way, about there's not much known about her private life, uh, because she's not a very public person.
• 24:35
  SPEAKER_00: Uh yeah, which means you can stop now because you're an encyclopedia.
• 24:39
  SPEAKER_00: But and then it linked that to the sort of message discipline and the secrecy of the Blair administration.
• 24:47
  SPEAKER_00: And I'm like, yeah, it's really not got anything to do with that.
• 24:50
  SPEAKER_00: It's just like she's not famous.
• 24:52
  SPEAKER_00: Right.
• 24:52
  SPEAKER_00: Yeah, yeah, yeah.
• 24:52
  SPEAKER_00: Yeah.
• 24:53
  SPEAKER_00: It's just like her personal life isn't of much interest to anybody.
• 24:56
  SPEAKER_00: So yeah.
• 24:57
  SPEAKER_00: So uh so anyway, but at the same time, I think we have to say like the anything you say about AI today in six months' time, you really have to reevaluate.
• 25:10
  SPEAKER_00: Sure, sure.
• 25:10
  SPEAKER_00: And certainly, you know, when I look at it and I'm playing with things, uh, particularly when once you give yourself a huge token budget and you don't just do a simple chatbot where you ask the question, but you actually sort of in an agentic way, you go through multiple passes to improve and so on.
• 25:30
  SPEAKER_00: Uh, I think it's interesting.
• 25:32
  SPEAKER_00: And I think there's more and more use cases that the Wikipedia community might find helpful.
• 25:37
  SPEAKER_02: On that note, I mean AI is, I mean, a lot of these uh chatbots are being trained in large part on Wikipedia.
• 25:43
  SPEAKER_02: Of course, the the LLMs are trained on the internet.
• 25:45
  SPEAKER_02: AI is kind of becoming in some ways the new interface for knowledge and truth for a lot of people in a way that Wikipedia and Google were in the previous era.
• 25:52
  SPEAKER_02: How do you feel about Wikipedia's role in this?
• 25:54
  SPEAKER_02: Is this going to be sort of the what just one part of the library infrastructure that feeds into the AIs or yeah?
• 26:00
  SPEAKER_00: Yeah, so I mean, all of the uh every model trains on Wikipedia.
• 26:04
  SPEAKER_00: Of course, it's a huge part of the training component.
• 26:07
  SPEAKER_00: Um, I think by numbers, the the you don't always know because they don't reveal everything.
• 26:12
  SPEAKER_00: They're all busy competing with each other, but uh it seems like in terms of quantity, like Reddit number one and Wikipedia number two.
• 26:19
  SPEAKER_02: Reddit's number one.
• 26:20
  SPEAKER_00: Yeah, but Reddit is bigger than Wikipedia in terms of sheer volume of words, uh, because you know, like that's a lot of Reddit posts every day.
• 26:28
  SPEAKER_02: But I always say it but Reddit leans very specifically to a certain kind of demographic too.
• 26:33
  SPEAKER_00: So it certainly does, yeah, yeah.
• 26:35
  SPEAKER_00: Sort of a nerdy, I would say somewhat left, uh nerdy left.
• 26:39
  SPEAKER_00: And quite male, I think, right?
• 26:40
  SPEAKER_00: Extremely male, even more male than Wikipedia, yeah.
• 26:43
  SPEAKER_00: But anyway, it's a huge part of the training data.
• 26:45
  SPEAKER_00: And but what I always say is they read Reddit to find out how humans talk, and then they read Wikipedia to find out information.
• 26:52
  SPEAKER_00: But um, so it's a huge part of the training data.
• 26:54
  SPEAKER_00: But I think the the the real question is, you know, what is our place in the consumer internet ecosystem?
• 27:04
  SPEAKER_00: You know, traditionally our place was, you know, if you if you ask uh Google, how old is Tom Cruise?
• 27:13
  SPEAKER_00: I and 20 years ago you type how old is Tom Cruise into Google.
• 27:17
  SPEAKER_00: Google had no idea how old Tom Cruise is, but it knew where to find that.
• 27:19
  SPEAKER_00: So it linked Wikipedia, and you would come to Wikipedia and find out.
• 27:23
  SPEAKER_00: Now Google knows, but Google has always been good about attribution, so it'll say, right, the top, you know, source Wikipedia, and then we're still the first link, and we're in all the knowledge graph and all that loads of links to us.
• 27:34
  SPEAKER_00: And so we we have seen in the last year an 8% drop in human traffic, even though our traffic has gone on.
• 27:41
  SPEAKER_00: We've got a lot of bot traffic these days.
• 27:44
  SPEAKER_00: Um, and we believe, but it's hard to really tease out, that a lot of that is that really quick answer short traffic, where you're asking a very simple question, you know, uh, what's the capital of France?
• 27:59
  SPEAKER_00: Well, probably nobody asked that.
• 28:01
  SPEAKER_00: Yeah, what's the capital of Azerbaijan?
• 28:04
  SPEAKER_00: And uh you get the answer and you're you're you're done.
• 28:07
  SPEAKER_00: So you don't go to Wikipedia because you already got the answer.
• 28:10
  SPEAKER_00: And then, but what you might do is you might say, Oh, okay, that's interesting.
• 28:14
  SPEAKER_00: I want to know more about that place.
• 28:15
  SPEAKER_00: So then you click the link when it says Source Wikipedia and you go and you read about that place.
• 28:19
  SPEAKER_00: So we still have that long-form reader traffic.
• 28:22
  SPEAKER_00: We still, you know, people are generally aware that you can't really trust AI.
• 28:27
  SPEAKER_00: And I think in particular, um, you know, the the Google um sort of AI summaries at the top of search, I don't think that Google would put them there were it not for the competitive pressure they're under.
• 28:43
  SPEAKER_00: I think if they didn't have everybody nipping at their heels, I just think it's not in their nature to put a product up that's so insane because it gets things wrong a lot.
• 28:52
  SPEAKER_03: Right.
• 28:53
  SPEAKER_00: Uh things you would think would be right, and you're like, oh my God, like you've gone a little crazy there, Google.
• 28:59
  SPEAKER_00: Um but that's good in a way.
• 29:01
  SPEAKER_00: It's good that if people are like, oh, okay, hold on, it's AI.
• 29:04
  SPEAKER_00: I mean, my children are good, you know, the teenage girls, and they roll their eyes about AI quite a lot because they or they'll they'll search something up and they'll give an answer uh to me, you know.
• 29:15
  SPEAKER_00: I was like, oh, what's this?
• 29:16
  SPEAKER_00: Can you look that up for me?
• 29:17
  SPEAKER_00: And they're like, but it's just AI answering, you know.
• 29:20
  SPEAKER_00: So like we're not sure.
• 29:21
  SPEAKER_00: And I'm like, great, that's let's hold on to that for a little while until it gets better.
• 29:24
  SPEAKER_00: But it is getting better.
• 29:26
  SPEAKER_00: And I I don't know how fast.
• 29:27
  SPEAKER_00: Like that's that's one of the open questions I would say at this moment in time.
• 29:32
  SPEAKER_00: Um, if you talk to Gary Marcus, a big AI researcher who's sometimes is thought of as an AI opponent or AI skeptic, he isn't really either of those, but I think he is he's a large language model.
• 29:45
  SPEAKER_00: He thinks we've hit a plateau and that scaling, more training, and so on isn't actually giving us very much incremental increase.
• 29:53
  SPEAKER_00: Yeah.
• 29:54
  SPEAKER_00: I think he's probably right about that.
• 29:56
  SPEAKER_00: Um, which means are we at a plateau that's gonna like are we stuck and we're gonna be here for 20 years?
• 30:02
  SPEAKER_00: Are we stuck and we're gonna be here for six months?
• 30:05
  SPEAKER_00: Uh I don't know yet.
• 30:06
  SPEAKER_00: Like, I think that's a super interesting issue.
• 30:09
  SPEAKER_01: Wikipedia, you guys are one of the most viewed websites in the world.
• 30:12
  SPEAKER_01: Yeah, you think in five or ten years' time, do you think people at scale will still be opening up their phone or a laptop and going on websites to go on things like Wikipedia or newspapers or whatever else, or will it still mostly then be going through sort of chatbot interfaces?
• 30:27
  SPEAKER_01: Uh and and so that's one question.
• 30:29
  SPEAKER_01: The other question I'd love to get you if you're on is kind of which types of websites will survive?
• 30:33
  SPEAKER_01: Because people think, for example, social media is like you can't really replicate that with with AI necessarily.
• 30:37
  SPEAKER_01: And so people think some form of social media will stick around.
• 30:40
  SPEAKER_00: Yeah, I it's a really good question.
• 30:42
  SPEAKER_00: And I think it's uh particularly if we think about journalism, it's a really uh crucial question for society because we need journalism.
• 30:54
  SPEAKER_00: Like, and there's no substitute, there's no AI substitute for a reporter on the ground gathering facts.
• 31:01
  SPEAKER_00: Um, I mean, maybe someday we'll have sort of humanoid journalists.
• 31:06
  SPEAKER_00: All-knowing humanoid journalists like monks recording uh in a in a perfect way, you know, everything that they see.
• 31:13
  SPEAKER_00: And I that's that's science fiction fantasy.
• 31:16
  SPEAKER_00: But and so the business model of journalism has obviously been a problem for a long time now.
• 31:23
  SPEAKER_00: I would say, I would argue, even going back into the 90s, uh, you know, that we we had problems and fractures were going on there.
• 31:31
  SPEAKER_00: And I don't have an answer to that, unfortunately.
• 31:33
  SPEAKER_00: I wish I did.
• 31:34
  SPEAKER_00: Um, but I do I do think that uh we will definitely still have journalism and we will still have people who want to read a story that's written by a human of something that happened today.
• 31:48
  SPEAKER_00: Um, I think that we'll see some dynamics around paywalls and things like that.
• 31:53
  SPEAKER_00: We're already seeing uh battles break out uh between um, you know, information producers, journalism and scientific journals and things like that, and AI crawlers and bots and sort of reusing content.
• 32:09
  SPEAKER_00: And I think some of that is uh is potentially quite dangerous.
• 32:15
  SPEAKER_00: Uh like one of my concerns is that, you know, at this time, obviously the the traditional media still have an enormous amount of political power.
• 32:25
  SPEAKER_00: And I can tell you there are definitely scientific publishers who would love to change the law so that facts are copyrightable.
• 32:32
  SPEAKER_00: Right now, facts are not copyrightable.
• 32:34
  SPEAKER_00: Um, and to say you can't use any facts from our scientific articles unless you pay a license fee, I mean, that would be horrific, right?
• 32:41
  SPEAKER_00: Uh, it'd be horrific if you couldn't report on any news if someone else reported on it first.
• 32:46
  SPEAKER_00: Like all these things are really bad ideas.
• 32:49
  SPEAKER_00: Even though uh, you know, these questions about like what happens if people stop going to news sites entirely because they just rely on an AI summary.
• 32:59
  SPEAKER_00: Yeah, okay, that's interesting.
• 33:00
  SPEAKER_00: Like that's potentially very problematic.
• 33:03
  SPEAKER_00: I I also I don't I'm not so sure it's all that likely.
• 33:05
  SPEAKER_00: I mean, I think that chatbot interfaces are um useful in some ways, but they're not really always what people want.
• 33:14
  SPEAKER_00: I mean, one of the things that I do uh on my phone, mainly because I took social media off my phone, because I don't think it's healthy for me.
• 33:22
  SPEAKER_00: But now, I don't know if this is any healthier, but now I just read the news.
• 33:26
  SPEAKER_00: Uh, but I use the uh I have an Android phone, so there's the Google app, and you just scroll and it it rem it knows you and so on.
• 33:33
  SPEAKER_00: And it so I I get a lot of stories about AI and things like that that I'm interested in.
• 33:37
  SPEAKER_00: And I I like that.
• 33:38
  SPEAKER_00: One of the reasons I like that as a human is I don't know what I'm not I'm not in search mode.
• 33:43
  SPEAKER_00: I'm not in, I have a question and I want to get an answer to my question.
• 33:47
  SPEAKER_00: So if you give me a blank chatbot screen, I don't there's nothing I have at the moment.
• 33:52
  SPEAKER_00: Often, though, I read a news story and then I switch over and I'm like, oh, hold on, I'm gonna go ask Jim and I about that.
• 33:57
  SPEAKER_00: Um, but you still want that serendipity of just some a feed, you know.
• 34:02
  SPEAKER_00: And I do think uh, you know, social media, although the I think social media is well due for some new competition because um I think a lot of people are like me sort of saying, actually, I don't, I'm not it's not benefiting my life in a way.
• 34:19
  SPEAKER_00: And even though it can be quite addictive, it's you know, it's it's not actually good for me.
• 34:24
  SPEAKER_00: And so, you know, I I feel the same way about um short form video.
• 34:29
  SPEAKER_00: Um actually one of my I have very, you know, I I love Google, but I have so few complaints.
• 34:34
  SPEAKER_00: And one is on YouTube, it's very hard to turn off shorts.
• 34:38
  SPEAKER_00: Yes.
• 34:38
  SPEAKER_00: I'm like, I often I don't really want shorts because I know I'll click on them and I know I'll waste 20 minutes flicking to the next, flicking to the next.
• 34:45
  SPEAKER_00: That's not really what I want to do with my time.
• 34:47
  SPEAKER_00: But I do want to go to YouTube and sort of get an algorithmic feed of of longer videos and actually click on something and spend 20 minutes watching and learning something.
• 34:56
  SPEAKER_00: Anyway, that's just a little complaint.
• 34:58
  SPEAKER_00: Yeah.
• 34:58
  SPEAKER_00: Anyone from Google's list.
• 35:00
  SPEAKER_01: Speaking of watching very interesting videos on YouTube, this has been a fascinating part one.
• 35:05
  SPEAKER_01: We're gonna come back for part two next week to talk more about trust and also your journey as well.
• 35:10
  SPEAKER_01: Thanks so much for joining John Ideas, Jimmy.