Can Musk Get Trump Elected? + Steve Ballmer’s Quest for the Facts + This Week in A.I.

25:24It's iconic. If you haven't seen it, go look it up. It's an amazing speech. But his new venture is very dispassionate, I would say. He is trying to basically just give people facts about the government and about the economy and about various other topics and sort of hope that that will educate them and inform them and make them better citizens. Yeah, and so why is that interesting to us? Well, you know, we live in a time where increasingly and partly because of tech, people are allowed to pick their own realities. You know, you pick a sort of partisan social network to hang out on. You pick a partisan cable news channel to get your news from.

25:59You pick other partisan media to kind of shape your reality, partisan influencers. And I think this kind of thing really frustrates Balmer because he's a tech guy. He was a CEO. He spent his entire career in meetings where people were showing him numbers, and he used those numbers to decide what is the best path forward. So USA Facts is his effort to bring a little bit of that CEO-inspired decision-making into American politics. Yeah, so I watched one of the Steve Ballmer videos that is coming out of this project because the project is not new. He's been doing this since 2015. But he did just start putting out a series of YouTube videos in which Steve Ballmer himself sort of appears on screen and basically gives like a little lecture about some part of the government, whether it's immigration or, you know, the state of the federal debt.

26:45He's just sort of patiently and calmly walking people through all the various information. And I watched one of these last night because I was like, I wonder what Steve Ballmer sounds like in this format. And it's very interesting. It's like a 13 and a half minute video that I watched about the federal budget and the deficit. And it's just it's sort of like Wikipedia level sort of summaries of where the money goes in the federal government, how much goes to Medicare, how much goes to Social Security, how much goes to the military. It is about as dry and neutral a presentation of the facts as possible.

27:17And it's just a very interesting contrast to people like Elon Musk, who've decided that the way that they're going to use their billions that they made in tech is by throwing their support behind one candidate or another. Steve Ballmer's basically saying, I'm just going to give you the facts and let you guys sort it out. Yeah, and look, I'll say, I think that this project, we might describe it as a bit quixotic, right? Like, I think, you know, even several years into this, this thing hasn't caught fire yet. And I'm just so fascinated by the fact that we have this one very rich, very powerful person who rather than go the full Elon route said, I'm going to do something completely different.

27:49And it just really made me want to talk to him. Yeah. So let's go ahead and bring in Steve Ballmer.

27:59Steve Ballmer, welcome to Hard Fork. Thanks. Pleasure to be here. So after being CEO of Microsoft, you could have done just about anything. How did you get interested in the media ecosystem? It's interesting that you call it the media ecosystem. I don't think of it that way. But given that we write things and produce videos, I guess you could call it the media ecosystem. Now, when I retired from Microsoft, my wife said, you know, it's time, dude, you got to help with the family philanthropy. I got it when you were out running Microsoft, but I've been plugging away. It's time for you to work. So I said, eh.

28:42And she was focusing on kids and families where, let's just say from the time of birth, the opportunity to move up economically is not that good. So what was my lazy-ass response? It was, hey, government does that. Government pays for those things. let's just pay our taxes and be happy about that and call it a day she said uh uh uh uh uh uh and guess what it was uh uh so i went to work with her on that but i had one of those little spousal i'm gonna see if i can prove i'm right so i went online and tried to snoop around and figure out really what was going on for wealth transfer, what the government was really doing, how to think about that.

29:32And it was a mess. It was a mess to do what I like to do, which is put together things holistically, you know, holistically who pays taxes, holistically what is the money being spent on. And that sent me out to try to work on that. I thought we had some interesting stuff. So we said, hey, let's publish that. People were kind of interested. And we've been doing that now for, I started in 2015, and we're still plugging away. So what I'm hearing is you go out and you get some data and you share it with folks and you know that different people are going to receive it. But the hope is maybe if we just sort of share some raw data, we can get some agreement.

30:16Maybe we can have sharper policy debates and that sort of thing. Like you said, you've been at this now for almost a decade. What's happened so far? How do you feel like that's going? I would say I am warmed by our progress and I am daunted by the progress I want to make. Look, we have, you know, kind of regularly, you know, we're doing just over a million a month in terms of visits. It's not bad. But if you really want to educate people, we need to get to more people. And I want longer interactions. You know, we're over 200 ,000 newsletter subscribers. Not bad. Not bad. Not bad. But, you know, when I think about, as I said, millions, and I think about the kind of interaction that lets you be thoughtful, which takes a little longer, I am daunted and excited about the potential.

31:14So the thing that's new this year is that you're pivoting to video. Is that right? You're putting up videos on YouTube. Are you doing any other platforms? And I guess, is that born out of a feeling that you can just reach a bigger audience through video than you could through text? Or what prompted the shift? Pivot sounds like we took a hard left and we'll only do videos. So I would call it additionally, if you will. and i'll tell you we're real this was really born in 1992 uh i'm looking at you guys and say you probably don't remember what i'm about to say but okay casey's very old so he was a teenager but i was in grade school i was in sixth grade okay so so here's the deal 1992 ross perot who had started a tech company called eds decides he's going to run for president as an independent and ross starts buying TV time.

32:09And he does these little poster boards, almost like from science fair. And he's trying to explain government by these poster boards, by the numbers. So this friend of mine who'd been kind of around the beginning of USA Facts, I'd known him since college, he and I have always admired the Perot poster boards. So he comes at me this year and said, come on, man, It's an election year. We should take a winger on this. Let's try to do the modern day equivalent of the poster boards. Let's buy remnant TV like Ross would have done and, you know, run 15 minute videos that, you know, really are comprehensible.

32:50People can watch on TV. We'll drive them to YouTube. Let's go do. I'm not running for anything. So I'm not Ross Perot to the max. But we said, you know, let's go. And so I'm heartened. We'll see how we do with the next four videos. And what do you want people to do with the information that they get from these videos? Is it about sort of sharing it with their friends? Is it about talking to their elected officials? Or what do you what do you hope the effect is? Yes. Yes. And voting. Let me just take immigration as an issue that we've already done. Yeah. At the end of the day, we're going to have some immigration.

33:30We are. Nobody would say not. So in a way, the question becomes how many immigrants, not I'm against it, I'm for it. I mean, that's kind of a silly discussion in my opinion. It's, hey, let's step back and think how much immigration do we want with people who have what kind of labor skills and what really is accepting people who have pain in their home country. And let's make sure our systems line up with our strategy. I mean, we're, you know, we have, you know, let me check the number. We have basically. Facts only, please, Steve. Yeah, that's all. We have about 1.3 million people who were released or paroled into the United States, most of whom are seeking asylum.

34:27And I can't remember off the top of my head, but let's say it's 50 ,000 people who were granted asylum last year. And there's many people who have 10 years before they'll even get their asylum court case. This does not compute. If we want to bring in a certain number of people, maybe we want to bring in 2 million without a path to citizenship, another 1.3 million with a path to asylum, whatever it is. It's just numbers. And, you know, I don't – some people may think the number of immigrants we should let in is very small. Some people would think it's very large. I don't think anybody thinks it's zero.

35:08So debate. Debate to the facts. Debate to the facts. Not, you know, just, hey, I don't like it. Hey, I do. Build the wall. Don't build the wall. Tear the wall down. And I'm a, you know, sort of a nonpartisan complainer in that. Yeah. The idea that you can change minds with data feels very tech-inspired to me. I can imagine that as a CEO, you were often shown data that changed your mind about what decision to make. Do you feel like this approach is inspired by the way you tried to run Microsoft? Yes, and look, dudes, I was a math guy. Actually, in second grade, the math teacher told my parents that I was going to struggle in math.

35:52and so I studied all summers so that in third grade I wouldn't be behind. But it turns out I like these numbers. I went to a math camp one summer. I went to a science camp one summer. I came into college saying, hey, I'm going to major in physics or math and math, math, math. So I have a long pre-CEO heritage of being a numbers lover and a math lover and I also think that's very helpful. I don't know how you'd be a CEO. You don't have to be me, a math camper, but I think it's very hard to be a CEO unless you're going to dig into some numbers. Yeah. Steve, I want to sort of interrogate your theory of change here because I think there's this idea that Casey just alluded to that sort of people change their minds based on the quality and the quantity of data that is available to them.

36:49And I know that's a very popular idea among sort of business leaders and people with experience running companies, but I don't feel that out in the culture. And, you know, take economic data, for example. That is a domain where we actually do have very good, reliable, high-quality government data. You know, we've got the Bureau of Labor Statistics that's constantly putting out reports in the Federal Reserve. But when you ask people about the state of the economy, they don't cite that data. And in fact, they often get the data wrong. There was this survey that was done earlier this year by The Guardian where they basically asked people like, do you think the stock market is up or down this year?

37:30Are we in a recession or not? And some huge percentage of Americans believe falsely that we are in a recession and also believe that the stock market is going down when it's actually going up. So I guess my question to you is, like, what makes you confident that telling people, hey, here are the numbers, look how well the economy is doing, here's the data that proves it, is actually going to change their minds or cause them to behave any differently? Remember, I'm not trying to change their minds. We are trying to educate. It's a little different. Do I fundamentally disagree with what you're saying?

38:08In a way, no. People will extrapolate from their personal experience. I'm also maybe not as aspirational as you guys would have me be. Let's say we could get 5 to 10 million people. We know there's about, what, 150 million, 160 million people who vote. if we can get some people to ground that is bigger than a breadbasket, but smaller than a refrigerator, as my mom used to say, if we can do that, I view that as successful. I mean, sometimes journalists try to do these sort of like very like fact-heavy stories like, hey, you know, this is like, we think it's like really important. And maybe it's not about a subject that you're super interested in, but we just think that you should know that.

38:50And we often call that, you know, making readers eat their vegetables. You know, it strikes me your whole enterprise is about making people eat their vegetables. And my guess is you're not able to do a lot of clickbait, right? Like I went to your website. There are no headlines. It's like, you know, this one government statistic will absolutely shock you. But you've sort of given yourself, I feel like, a tough challenge just by the nature of how sort of rigorously neutral you're trying to be about everything. My wife gives me, she says, look, what do you think we should do? People are interested.

39:23To which my response in this context has to be, could we do more? One of my dreams is eventually we'll figure out how to do a point counterpoint. We get people from both sides. We'd kind of referee to anchor the thing in the numbers. So you get a little more pizzazz, a little bit more energy, but it's not our opinion. It's our holding people honest, but a little point counterpoint might give us, you know, it might make us, uh, what shall I say? Uh, more of the amuse-bouche. Uh, it's not exactly your dessert, man, but it wets your whistle. Um, you know, and look, there are places I think people do a pretty good job of this.

40:08Take the economist. The economist does pretty good job actually on their charts and numbers. and they feature them. And for that reader, I don't think it feels like vegetables, but they, as you know, they have a very particular audience. I'm curious, Steve, there's been a lot of chatter this year about the sort of tech elite shifting toward Donald Trump and sort of becoming more activated on the right. We've seen a number of influential people in tech come out in support of him. I'm curious, what do you make of all the Trump curious tech CEOs and power players out there? One question I'd throw out is, and let's go back pre-Trump, really?

40:54Was the tech industry all Democrats? The answer is no. It's just everything wasn't so out there. Things were not so emotionally charged. I would say on balance, you would hear more from the, you know, sort of more de-partisan folks than the are partisan folks. But I know at Microsoft, we had people who were happy to welcome Republicans on campus to speak as well as Democrats to come on campus and speak. So I don't think it's completely different. I'm not saying it's not different, but I don't think it's completely different from days past. It's just everything's more sensationalized. That's one thing.

41:40The second thing I'd say is, you know, tech CEOs are very business focused and that means different things to different people. And both parties try to say things to appeal to various constituencies for sure. Now I'll just be completely nonpartisan. During the Trump administration, corporate income taxes went down. That's a fact. And not only did you see it in the rates, if you go, I mean, when people talk about tax rates, that's a little bit of BS. What you should really look at is total taxes paid because that's the reality. The rates are a tool to get there, but the rates don't always drive.

42:27And I could talk more about that, but revenue went down. Well, if you're a corporate CEO, you might say, Hey, that's a good thing for my shareholders. So I salute that. I support that. Uh, so there are issues, concrete issues in which, you know, I can understand why a CEO would say one thing or another. Is that why the people who are most vocal are really speaking up, whether it's on the D side or the R side. I'm not sure that's why people are really speaking up. But, you know, we generally, when I was at Microsoft, we avoid, I avoided doing things that were political, but there were issues that were important to our employees.

43:09We had an issue around gay marriage, I think it was. And there was a local pastor in Bellevue, Redmond, Washington, who was stirring that pot. And we stood for our employees on that issue, but it was an issue. It wasn't a partisan issue per se. So, you know, this is a tech podcast, but I'm sure we have plenty of sports fans who listen to us as well, Kevin. And Steve, you own the Clippers and you have a new arena opening this week. We're talking to you the week that you're opening the Intuit Dome, this massive project in Inglewood. And so I guess I just wanted to know, are there cool tech things that you're doing in your new arena?

43:51I think very much so. But I want to make a point. I said when we started this thing, which was also nine plus years ago, by the way, things take time. I said to the guy, look, you may think I'm a tech guy, so we just have to do tech because we have to do tech. And I said, that will never be the mindset of this project because you, you, you see that, Oh, I've got to apply my background because my background's mine and I got to strut my stuff. No, I said, we're not going to do that. But as we got into things that we wanted to do, the tech got important. I'll give you two or three examples. First, the simplest example of all.

44:33We have an acre of scoreboard. The typical NBA scoreboard's 9 ,000 square feet. We did it all in 4K. We couldn't go 8K. I wanted 8K. I want to see the individual pores on the players' faces when I'm watching. You should come take a look. No, but there's a question. Is the display quality good enough for the long run? And so you get to how does your eye see it from the distance? Remember, we're not as close up as somebody looking at their phone or computer. Okay. Then you go to, hey, I really want to reward people who get in their seat early. Well, how do we know who's in the building? How do we reward them?

45:19How do we know if they're sitting down or not? During the game, we want to reward them if they're standing up and making noise. And so we went through all the things we would do from a tech perspective to do that. So we have a system built in place that down to the individual seat level, we can tell how noisy you are. We put proximity sensor into the seat so we can tell, are you there or are you not there? We put power. We wound up running power to every seat so people could charge their devices. But as soon as you have power there, then you say, what else can we do with the power? So we put lighting into every seat that we can control.

45:59We put a mini controller, a four-button controller in there so people can interact with the big board sense of community. Once we had some of this stuff, huh, we have a huge screen out in our plaza, different screen. Let's get some digital art done. What does that look like? So actually the artist takes statistical data, including, for example, data on how our team played, and then creates art. using some AI techniques and some not to create the art that goes up digitally on the big board. So did we wind up with some things that people would consider technological? I think the answer to that is yes.

46:41Steve, have you ever considered combining your two passions of owning the Clippers and government data and maybe having at the Clippers game, instead of like the half court shot at halftime to win a car, Like you have to list three figures from the most recent Congressional Budget Office report and win a car. Like you could really have some synergy here. That's just a free idea. I'll consider doing that over 43 ,000 square feet with four buttons. Well, Steve, this has been fascinating. Thank you so much for coming on Hard Fork. I really appreciate it. It's been a lot of fun. You guys are great.

47:19All right. Thanks, Steve. Thanks very much. Bye-bye.

47:24When we come back, we'll tell you what happened that made one software engineer say, the entire night I was hearing the boop, boop, boop.

47:48Okay, so you know what? I'm sort of sick of talking about politics this week. Me too. I know. It's hard to escape because we are in an election year. But there is also a ton happening in the tech world. And so this week, I thought we should bring back a segment that we haven't done in a while. Let's do This Week in AI. Play the theme music. This Week in AI. That was beautiful. Thank you. Yeah. Kevin, what's been happening this week in AI? Well, first up, the AI industry calls for a pause in regulating AI. Oh, yeah. It's about time we stopped all of these onerous regulations that have been passed everywhere.

48:24I'm just kidding. There have not been any of those. Yes, but this week, the biggest story in Silicon Valley that people are talking about that is not the election is this thing called Senate Bill 1047. This is a bill that has been proposed in California, a state bill that is technically called the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act. And Casey, people are really upset about this. Yeah, this is one where rather than the federal government trying to set like one standard for AI across the country, California has come in, as it often does, and tried to offer a template.

49:02And while I think sometimes, you know, those approaches have been very successful with things like auto emissions, this is one where basically every single company that works on AI says, this is terrible, please do not do this. Yeah, so this bill passed the California State Senate in May, and this week it's been making its way through the California State Assembly Appropriations Committee. The next step is it's now expected to go to the full assembly for a vote. So what's in this bill? So it's a long bill. It's got a lot in it. But basically it says that if you are making what they call a frontier AI model, one of the most advanced AI models out there, you have to conduct safety testing on it.

49:40That is the biggest sort of central piece of this. You have to tell the government that you're doing this. You have to conduct safety testing before you release it to the public. The bill also empowers the state attorney general to take legal action against AI companies if their AI tools cause what they call critical harms, which they define as things like a massive cyber attack that does a lot of damage or up to and including a mass casualty event. You could then be sued if you are an AI company who makes one of these models. There's a lot more in there, but that's sort of the thing that got everyone in Silicon Valley's hackles up and made them want to oppose this bill.

50:16Yeah, and basically, like, it's a who's who of AI when it comes to who opposes this bill. So Meta, Google, OpenAI, Microsoft, and Anthropic all oppose it or have offered some lengthy proposed amendments. Ycombinator, the big startup accelerator here, released a letter from about 140 startups who oppose this bill. On the other side, some of the godfathers of AI, including Jeff Hinton and Joshua Bengio, do support it. So, Kevin, what do we think of California's effort to regulate AI here? Yeah, I think that there's been a lot of kind of inflated and hyperbolic claims being made on both sides of the debate.

50:50But I think this is a really important conversation to have because I think everyone agrees that AI needs to be regulated. But then, you know, as soon as a bill comes along that would regulate AI, it's like, oh, well, not like that. We didn't mean it like that. I'm a little skeptical that any of this is going to actually reduce the harm risk from AI. I just don't think the regulators can move fast enough to keep up with this technology. And so I think they're always going to be coming from behind a little bit. But what do you make of it? Well, my feeling about this is we had the Biden executive order last year, which attempts to do some of the same things.

51:27And so far, it seems like the AI industry has proceeded along without really hurting anyone too badly, except for all the copyright owners, of course. My preference would be let the national government lead here. Like, I do not want a patchwork of 50 different state laws regulating AI. And I don't see any pressing need for California to pass this bill now because I'm still not sure what specific harm they think that it would address even within the next few years. Yeah, I would love to see more regulation around transparency when it comes to how these models are trained, what data are they using, what does the pre-training process, the post-training process look like, how are they, what kinds of decisions are these companies making about what their models can and can't do.

52:14Like, that's the kind of thing where I think regulators could actually move the needle in the short term. But I think you're right. I mean, these harms, some of them are still very theoretical. I think that's actually not an argument against this bill. I think we should be trying to avoid harms before they happen. But I think in this country, the way that things usually go when it comes to regulating industries is that something terrible happens and then we pass a bill to prevent it from happening again. I think the risk here that the AI safety folks are worried about is like they don't think we will have that luxury.

52:45They think there will basically be these harms will arrive in some catastrophic form in the next few years and we basically get one shot to stop them. And if we're late, then we're doomed. I respect the point of view. This is just not how I want to solve this problem right now. Yeah. All right. What else is happening in AI this week, Kevin? Okay. Well, the second story from AI this week is that Humane's AI pin is the must-not-have gadget of the summer. This was a story, an amazing story, The Verge, last week, which reported that Humane, who built this AI pin, that was basically a kind of brooch that you'd wear on your shirt or your jacket that would let you talk to AI and project a little laser screen.

53:29The Verge reported that while Humane had brought in$9 million in sales of their AI pin, they had more than a million dollars worth of returns, and that between May and August, according to the internal sales data that The Verge got a hold of, more AI pins were returned than purchased. Casey, is that good? Well, Kevin. I'm no mathematician, but that does seem very bad and like a validation of what the critics were saying, unfortunately. Yeah. I mean, look, I think a lot of people wanted to try this thing out, right? There was kind of something cool and futuristic and sci-fi about it. And this is the great thing about early adopters.

54:08They will hand you$700 just to sort of maybe have an experience of what the future will be like. It seems like Humane's return policy was maybe a little too generous. because you know there's some things you can't return you know it's like if you buy underwear at the department store you can't actually return it i i wouldn't be surprised if humane puts out another device if the return policy isn't much more limited well and making things worse apparently according to the verge they can't humane can't actually do anything with the returned ai pins like they because there's some kind of limitation on the t-mobile contracts that are associated with these devices they can't just like refurbish them and and resell them to other customers it's just a whole you know goat rodeo I truly feel bad for these people they were trying to do something new a different cool it's hard to do things that are new a different and cool and they did talk a very big game and the effect of talking a very big game is they create a lot of people who are rooting against them and so now that it does seem like everything is falling apart like man it just has to be rough days over at Humane alright our next story from the Week in AI GPT-4-0 is dead long live GPT-4-0 oh my goodness what's happening Well, Casey, you will remember that several months ago, OpenAI released the new update to their flagship AI model.

55:23This was called GPT-40. And this week, we got the news that they have quietly updated the model behind ChatGPT, which is now called, and I promise I'm not kidding about this, ChatGPT-40-latest-2024-08 -06. zero six. Wow. We're one step away from open AI releasing chat GPT four Oh final, final, final, final, final, final, final. Use this one. Yeah. It's so true. So they have officially taken over the, uh, the title of the, the, the worst naming conventions for AI models, a title previously held by Google. Um, but, uh, but yeah, this is, um, this is an update that they have made quietly. They did not draw a bunch of attention to this.

56:10And I think this was related to a sense that we've actually heard about from some of our listeners who have written in to say, hey, does anyone notice that ChatGPT has just gotten really bad recently? Do you hear this? Yes, people have written into us about this. And for me, it's always a hard question to gauge because, you know, I don't know what you're inputting. I don't know what output you're getting. Like, I don't know what is making you say that it seems dumber. but the fact that these models are just sort of like getting swapped in and out and like still have basically the same names they did before is obviously very confusing you know i mean like i feel like such a like a an ai hipster being you know like back in my day these things were called gpt2 and then gpt3 and it was really easy for me to keep track of nowadays it's like 4.0 is gone and it's being replaced by 4.0 you lost me yeah yeah and it also feels like you know release notes when they update an app, they'll say like bug fixes and improvements.

57:05And sometimes they'll have like a little list of things they fix. They never do that for these AI models. People just sort of notice, oh, this thing seems smarter than it was or dumber than it was. It's just kind of this like, we're like trying to understand this unanswerable thing. As always, if you know what's going on at OpenAI, please get in touch. We're trying to figure it out. Have you been following the strawberry discourse? Yes, I have. My boyfriend is obsessed with strawberry discourse. So it's become a hot topic for us lately. Okay, I'm gonna try to explain it to our listeners who are not as terminally online as you and your boyfriend and I are.

57:36But correct me if I'm getting something wrong here. So for, I would say, more than a year now, there's been this sort of, I don't know, cottage industry of kind of people proclaiming to have inside information from the big AI companies, and specifically with open AI. There are these accounts on X that sort of purport to be insiders or people with access to unreleased models who will sort of give these like vague, mysterious statements about like something big is coming, like no one's ready for what's about to hit the internet. And there's this norm on X where if you're an AI person, you can have an anonymous account.

58:15And like some of those anonymous accounts have actually like gotten OpenAI's attention and those people have been hired into the company. And so whereas normally you'd see an anonymous account and you would think this is obviously fake, In this particular case, it actually could be real. And so what's been happening over the past week or two is Sam Altman, CEO of OpenAI, recently posted a photo on X of his garden or a garden with some strawberries growing in it. And the caption was, I love summer in the garden. Normal post. Seems like a pretty normal post. So the sort of AI movement went nuts over this.

58:58There was a theory that was being advanced by this anonymous account called I Rule the World that this was actually a clue that Sam Altman was dropping that the company's newest next flagship model, which is codenamed Strawberry, was about to come out and that this was going to be AGI and that the world was not ready for this. Right. And iRule the World didn't just do that right. They had a sort of very long post on X where they talked about it in somewhat like cryptic and, you know, I would say incomprehensible terms about what exactly was going on. But the gist was essentially AGI is very, very close.

59:37And I will just say I have truly no idea what to make of any of this except to say Sam Haltman did respond to this person on Twitter without really confirming anything. But in a way that I think did stoke a little bit of speculation that, well, maybe this person is on to something. Well, totally. And he's also dropped breadcrumbs before. Like, remember before the demo of the AI voice mode, he posted her on X. And that was sort of a breadcrumb that they were going to be doing this like voice update. So like he is sort of fueling some of these theories in a way that I think he I think he's mostly just having fun.

1:00:13But there is this weird community of like people on Reddit and X and other social platforms who are just like trying to decipher all of the clues and everything that Sam Altman says. which I do think just speaks to the fact that like if you want to get people interested in your company these days you have to create an alternate reality game where you're sort of putting clues around the internet and there's there's stuff for the reddit posters to discuss and analyze and tease apart so they've been very effective at that and I guess we'll have to see if any of this is real in the meantime though I do think that this works as psychological warfare against Sam's enemies because if you're working at Anthropic or Gemini and you're on X and you think, wait a minute, are they about to do?

1:00:53It just sort of puts that little, you know, fear into the back of your head and it messes with you in a way that I think would greatly please Sam Altman. Yeah. When you just start dropping some more, like, Easter eggs. Well, let me just drop an Easter egg right now, Kevin. Watermelon. And I'll say no more. I will say no more. But let me just say, those who know, know. Yeah. If you know, you know. You know, yeah. All right. Before we move on to our next item, here's our standard disclaimer that the lawyers want us to make. Every time we mentioned OpenAI, which is that the New York Times company has sued OpenAI for copyright infringement.

1:01:27And Platformer has not. And that's the difference between the two publications. Watermelon. Watermelon. Period. Okay, next story. This week in AI, Donald Trump is accusing Kamala Harris of deepfakery. So this was a story that went around this week. Donald Trump claimed falsely that Kamala Harris has used AI to make the size of the crowds at one of her rallies look larger than it actually was. So this was a photo that was posted online last week of a rally in Michigan, where basically there is a plane, Kamala Harris's plane, and there's a massive throng of people celebrating her arrival on the tarmac.

1:02:14And this was posted online. And Donald Trump apparently thought that this was a little suspicious. And he said on Truth Social, has anyone noticed that Kamala cheated at the airport? There was nobody at the plane and she AI'd it and showed a massive crowd of so-called followers, but they didn't exist. Now, this is false. The crowd actually was there. We have well-documented evidence that this was not manipulated to make it look bigger than it was. But this is now apparently a thing that you can just accuse your opponent of in a political campaign is using AI to fake the size of their crowd. Is this the first instance of a political candidate using AI as a verb?

1:02:54I think so. I think that's the first time I've ever heard anyone use AI as a verb. He AI'd it. I mean, look, everybody knows that Trump is obsessed with the size of his crowds, and he seems to have been getting more obsessed with it as Harris's crowds have grown. My initial instinct was to dismiss all of this as very silly, but it did kind of captivate the online right this week. And there were a lot of people sharing that and sort of, you know, trying to analyze the photo in Zapruder film-like detail. And at the end of the day, of course, there's nothing to analyze. This rally happened. There was so much evidence out there, people posting all sorts of photos, videos from every angle indicating that, yes, there were many, many people there, but it didn't matter.

1:03:37And I read a really great post on this this week from a guy named Mike Caulfield who studies sort of misinformation and related topics. And he said, this is actually really important in part because the idea that Democrats actually have small crowds is a core belief among the right. And people like Trump promote it because it means that should Harris win, people will have been primed to believe it can't possibly be real because all along the way, they've been saying, well, look, no one is even at a crowd. Oh, you thought there was a big crowd there. There wasn't. It was all just AI. So while on the surface, this is an absolutely ridiculous story, we have to remember that these kinds of lies led to the violence of January 6th, and we are unfortunately seeing them again in this election.

1:04:25Yep. Yeah. All right. Last story from the Week in AI. This one is about the hottest new club in San Francisco is the Waymo parking lot. Oh, yeah. It is popping off in the Waymo parking lot, Kevin. So this is a story that came out earlier this week after people started posting on the internet about a parking lot in San Francisco where a group of Waymo self-driving cars were gathering at night and honking at each other. Did you see this video? I did see this video, and I love this video. It's so good. It's called When It's 4 a.m., but the Waymos are getting aggressive in the parking lot. and if I can just describe it for our listeners there's a parking lot on the roof of a building someone's sort of filming from a nearby apartment building and it just shows these Waymo self-driving taxis kind of parked in a line but then also kind of some of them are backing up and trying to sort of get around each other and they're just honking and flashing their lights at each other.

1:05:23They're having a meeting. They're having a meeting. Can we watch this video together? Yeah, let's watch this. So you can see they're just kind of all

1:05:35saying hi to each other.

1:05:40It's fucking 4 a.m. Help.

1:05:50What a great piece of commentary there. So, Casey, what did you make of this? Look, I always think it's beautiful to see robots getting together and just sort of talking through their differences, and this was no different. So, you know, what did Waymo say about this, Kevin? They told The Verge that they are aware that in some scenarios, our vehicles may briefly honk while navigating our parking lots. They also said that they've figured out what's causing the behavior and they're working to fix it. I mean, it does answer a question that I had, which is where do all the Waymos go? That's one of my favorite children's books.

1:06:21Where the Waymos go? They have to charge, right? Because they're all electric. But then they're not, you know, it's not constant demand throughout the day. So, you know, at 4 a.m. when no one's trying to get home from the club anymore, where do they go? And it turns out they go to this one parking lot. I just, I love that the person who posted this video, Sophia Tong, has actually set up a live stream complete with lo-fi study beats. So at any time of day, you can just pop over to the live stream and check out what the Waymos are doing. Sophia, you're a legend. That's incredible. Amazing. Amazing work.

1:06:56It's true. I love it. We love it. You know, Casey, we should hang out in the parking lot at 4 a.m. We really should. Can we go hang out with the Waymos? I actually would like to do that some night. Almost do like a sort of David Attenborough-style field recording of like, At night, while San Francisco sleeps, a group of cars gathers to share their thoughts. And it sounds like this. Oh, yes, I love it. That's this week in AI! That's this week in AI. A lot going on.

1:07:43Hard Fork is produced by Whitney Jones and Rachel Cohn. We're edited by Jen Plian. We're fact-checked by Caitlin Love. Today's show was engineered by Chris Wood. Original music by Alicia B. Etoop, Marian Lozano, Sophia Landman, Rowan Nemisto, and Dan Powell. Our audience editor is Nell Golotli. Video production by Ryan Manning and Chris Schott. You can watch this full episode on YouTube at youtube.com slash hardfork. Special thanks to Paula Schumann, Wee Wing Tam, Dahlia Haddad, and Jeffrey Miranda. You can email us at hardfork at nytimes.com.