Legs Are Coming + $36 Burritos + Shocking AI News

Meta is in the fight of its life — and Mark Zuckerberg hopes VR is the answer. Hmmm. Plus, an existential threat to the gig economy and the wildest news in artificial intelligence.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T.com.

• 0:33Kevin Roose?

  Casey, anything else happen for you this week outside of Meta?

• 0:37Speaker 02

  Hmm, anything else happen for me? Um, meeting with people.

• 0:48Kevin Roose?

  OK, so no, nothing else happened for you this week? Just Meta? Your life is Meta?

• 0:59Kevin Roose?

  I'm Kevin Roos. I'm a tech columnist at The New York Times.

• 1:02Casey Newton?

  I'm Casey Newton from Platformer.

• 1:03Kevin Roose?

  And you're listening to Hard For.

• 1:05Casey Newton?

  And this week, my life really was Meta, specifically what I like to call Meta's annual talent show. So we're going to talk about that, the future of the gig economy, and then a truly wild week in artificial intelligence.

• 1:35Kevin Roose?

  So last week, we talked with Kashmir Hill about her experience in Meta's Metaverse app, Horizon Worlds. And this week, we learned a lot more about how Meta's bet on the Metaverse, this billions of dollars and thousands of employees that have been assigned to take us into the next age of computing. We learned how that's going. So Casey, what did we learn this week at Meta Connect?

• 1:58Casey Newton?

  Yeah, so once a year now, Meta throws a big annual conference where they show off their progress in this $10 billion project. And we learned a couple of things. One is that they are releasing a new high-end standalone VR headset called the Meta Quest Pro. It's going to ship October 25th, and it costs $1,500. And for context, the current sort of high-end headset that Meta sells, the Meta Quest 2, that sells for $400. So this is a really big swing at trying to get business customers in particular to buy these for their workers. And it comes with a bunch of other cool things. It's it's a sort of a much more

• 2:43Casey Newton?

  high-end development platform than than what we've seen to date. So that was kind of the headline news. But the thing that got everybody tweeting is that avatars in Meta's Metaverse are soon going to have legs.

• 2:56Kevin Roose?

  Legs. They will actually have an upper body and a lower body. And this was a big reveal.

• 3:03Casey Newton?

  I mean, it kind of was. And they poked some fun at themselves. They said, it might seem to you like legs should have been part of avatars from the jump. But they pointed out that in most of their peers, avatars don't have legs either. And apparently, it requires a lot of AI horsepower to use the various cameras and sensors in order to determine where your legs, you know, would be in this virtual world and animate them appropriately. So it was an unsolved problem in computer science, and it is closer to being solved.

• 3:34Kevin Roose?

  Check the box next to Metaverse legs. That part appears is done. Casey, you actually got to try this new high-end headset. I don't get invited to to demo Meta products anymore. I've made them mad too many times. But you did get invited down to Meta to test out the new Quest Pro. So What was it like?

• 3:57Casey Newton?

  Well, the first thing I want to tell you about going to Meta's campus in Burlingame, California is that there are outdoor speakers that play ambient music as like you're walking between buildings. So it's like, if you've ever been in the lobby of like the W Hotel, just imagine that, but it's playing around your corporate workspace. I'm still thinking about this a week later. Anyway, I don't know who the music was for. I thought it was like, it was like, sometimes like you walk down the street, like somebody's like on a boombox and you're like, oh, that's like kind of annoying. But it was like that, but like as an office campus. Okay, anyways, so I go in and I, there's a sort of nice presentation and there's some Meta workers and they tell me about, you know, everything that they've been doing to try to advance the state of the art. And after an hour or so of that, I get to go in and I get to do six demos over the course of a couple of hours that

• 4:40Casey Newton?

  show me everything they've been working on for these new virtual worlds that they're creating.

• 4:45Kevin Roose?

  And what was it like? What were the demos?

• 4:48Casey Newton?

  Yeah. So some of them were very silly and simple. Like there was a kind of virtual Jenga game, which showed off how you sort of have very fine grained motor controls now using these upgraded sensors in their, in their controllers. So you're able to pull these sort of very small virtual blocks out of their position without toppling the whole tower. I got to learn how to DJ just a little bit with this app called Tribe, where a musician who was sort of observing me remotely, had me standing in front of a virtual DJ console, which is intended to be an identical copy of the real rigs that DJs use. And with a

• 5:30Casey Newton?

  combination of like some rhythm games and some instruction from the musician, I was, you know, mixing a set and frankly, sounding fantastic. So that was kind of fun, too.

• 5:42Kevin Roose?

  Where's your, yeah, when, when does the mixtape drop? When does your SoundCloud come?

• 5:45Casey Newton?

  I'm going to do a Vegas residency and I will make sure to keep you posted. But you know, the really kind of like the highest stakes demo that I got, I would say, was for their, their workrooms product, which we actually used last week to talk to Cash Hill. But with the Quest Pro, it has some new bells and whistles. And one of the big ones is that if you turn it on, it is switched off by default. But if you turn it on, it will track your face and it will enable facial expressions. So I was getting this demo from, uh, from a executive or a, you know, a product manager at Meta. And, um, he's like smiling at me. He's making eye contact with me. At one point he winked at me just to show me that he could.

• 6:29Kevin Roose?

  That's creepy.

• 6:31Casey Newton?

  And I will say it, it made for a much, uh, richer, dare I say, experience than what we had done last week. Right. Because you sort of feel like, okay, yeah, this person is actually looking at me. They're nodding, you know, when I speak, they're smiling at me. And it felt a little bit more like being, uh, you know, in a meeting face to face with someone.

• 6:50Kevin Roose?

  Yeah. And I, I didn't demo the headset like you did, but I did watch the event this week. And, you know, they showed off a lot of these features, a lot of those sort of fine grained stuff that they're doing around motion tracking and face tracking and things like pancake lenses that I still don't really know what those are.

• 7:08Casey Newton?

  But they're delicious.

• 7:11Kevin Roose?

  They rattle off all these new features and then they start talking about all these partnerships they're doing. They announced that they're doing a partnership with NBC to bring like content from the office and other NBC franchises to the metaverse. They announced something with Microsoft. Uh, they say that Microsoft's products are coming to the metaverse. There will be Microsoft Teams and maybe eventually you'll be able to just do Excel spreadsheets in the metaverse.

• 7:40Casey Newton?

  Yeah.

• 7:40Kevin Roose?

  For whatever reasons you might want to do that. And all this was sort of nominally aimed at developers, right? This is a developer conference. The people who are listening are mostly people who are building things for VR and AR. But I think it was also supposed to be kind of a progress report, um, to show the public that, you know, over the past year, since Mark Zuckerberg announced that Facebook was changing its name to Meta, uh, that they actually have made material progress on, uh, on the metaverse. And did they convince you of that? Did what you saw at the event make you think that the metaverse is a year closer? Than it was last year?

• 8:17Casey Newton?

  In some ways, yes. Like, let me tell you the stuff that I think was really positive. Number one, this new headset is so much more comfortable to wear than the last one. Like, I actually tried to watch the event yesterday using my, uh, my Quest 2, and I just couldn't get the fit to a point that didn't make me want to talk, like, tear it off my head and throw it against a wall. The Pro had, it's, it's really well balanced. The battery is at the back, so it just kind of like sits a little bit better on your head. And, you know, even though I wore it for two hours, nothing about the fit of it was bothering for me, which really is an achievement. The other thing, too, is that the visual fidelity is just much higher. So I always thought that the, the Quest 2, it just looked a little muddy. You know, it's like, if you're used to looking at an iPhone or like a relatively good laptop display, like, the Quest 2 is just not very good.

• 9:02Casey Newton?

  The Pro was much better. Still not as good as an iPhone, but like, kind of getting there. So I thought between those two things, it felt like a really meaningful difference. I could imagine myself spending more time in this environment. And, you know, you were sort of saying, like, you know, would you actually work on an Excel spreadsheet in VR? They, they sort of showed me what that might look like, right? In one of the demos, I set up a virtual workspace. So I was sitting at a real physical desk, but then I pulled up three giant virtual monitors in front of me. I pulled up a web browser. I could, you know, go to any website that I wanted and then read the website in VR. This was not an amazing experience. Like, this would not be my top five ways that I would want to read any website. But it does work. So I think you sort of have to keep two things in your head at the same time.

• 9:48Casey Newton?

  One is, this really is way better than the last thing that they made. And two, it's not good enough for most people. And that is like the most frustrating point, I think, for any tech company to be in. But it's also the only way that progress happens, right? Is you just try to get it a little bit better every year.

• 10:04Kevin Roose?

  Well, I actually, I've tried to approach the subject of the metaverse with an open mind. I think, you know, talking to Cash Hill last week made me, you know, sort of realize that like, you know, it is possible to have positive experiences in the metaverse and to feel like your time there is well spent. This event did not convince me of anything, except that Meta is really Really desperate to make this happen. I was counting at one point the number of times that someone who was on screen said that they were excited or so excited or very excited for something in the metaverse. And it was just like so much forced enthusiasm, so much like scripted, like infomercial style pitches for these products.

• 10:48Kevin Roose?

  None of it felt genuine. It all felt like a bunch of people who are trying very hard to convince you that something is good and exciting, but not actually like being palpably excited about it themselves. Am I, am I reading that wrong?

• 11:05Casey Newton?

  Yeah, well, at the same time, Kevin, I happen to know that you had spent weeks before this talking to nothing but disgruntled Meta employees about how this was going for them. So I imagine you were bringing that filter to things as you watched the big talent show.

• 11:19Kevin Roose?

  Yeah, I did come out with a story just shortly before the event with my colleagues, Ryan Mack and Shira Frankel. We did not just talk to disgruntled Meta employees, but we did talk to a number of people inside Meta, current and former employees, about how this metaverse push is going inside the company. And the overwhelming takeaway from talking to these people was that it's just been very messy and chaotic and frustrating. That there's a lot of disagreement inside the company about whether this is a smart strategy. So there's just a lot of skepticism of this effort and a lot of people who are sort of internally kind of disgruntled with the fact that, you know, they are doing this thing that they don't necessarily believe in. So just a few things that we reported.

• 12:05Kevin Roose?

  One of them is that, you know, Meta employees are not by and large jazzed about using the metaverse for, for meetings that actually the company earlier this year told them that they should be taking meetings in Horizon workrooms. But a lot of employees actually didn't use VR. They didn't, they hadn't gotten their VR headsets or hadn't gotten them set up yet. And so there was this kind of scramble at the last minute to, to get these things set up so that their managers wouldn't know that they hadn't been using it. We also reported that there's been a lot of turmoil around Horizon Worlds, the, the sort of flagship social app that Meta hopes will be kind of like the, the sort of metaverse equivalent of Facebook, the place where you'll, you'll just kind of like hang out with your friends and family.

• 12:50Kevin Roose?

  And that is going so poorly that the VP in charge of it, Vishal Shah, just wrote a post in which he declared that they would have a, quote, "quality lockdown" on Horizon Worlds, which would last the rest of the year, which is basically like, okay, we, this thing, it's buggy, it doesn't really work that well, it's not that popular with users, and so we are just gonna basically focus ourselves intensely on making this thing better, because right now, it's not working.

• 13:18Casey Newton?

  And I actually thought that was really inspiring, and so I'm announcing right now that Hard Fork is gonna be on quality lockdown through the end of the year, and we're gonna, we'll be fixing the bugs, we will be addressing worker morale, anything that we can do to get this podcast into shape, we're gonna do. So, very interesting reporting, that memo in particular had some memorable lines in it.

• 13:42Kevin Roose?

  And then, one more, one more point I wanna make on this, is that even among the high-ranking folks at Meta, there is a lot of tension and disagreement with Mark Zuckerberg about this, this Metaverse strategy. So, we got some posts from Workplace, which is sort of Facebook's internal version of Facebook that they use for all their Workplace communication, by John Carmack. And I don't know how familiar people are with John Carmack, but the, basically, he's kind of this, like, godfather figure in the tech industry. He was a game developer for many years, he was the force behind games like Doom and

• 14:25Kevin Roose?

  Quake and Wolfenstein, if you remember that.

• 14:28Casey Newton?

  Of course.

• 14:28Kevin Roose?

  Or my personal favorite, Commander Keen, from back in the day, that was also one of Carmack's games. And people who know him just kind of talk about him as like, maybe the greatest software engineer of all time. So he spoke at this event this week, he got his own, like, hour-long speaking slot after Mark Zuckerberg had done his thing, and people were making comments about him, like, one person called him a Level 99 Technomancer, like, this is a guy who's just done- Everything there is to do in the tech industry. He got bored making games and started an aerospace company. Now he's doing like artificial general intelligence and he still works at Meta one day a week.

• 15:11Kevin Roose?

  And he has this kind of singular role at Meta where he's not directly like managing large teams. He's not like working directly on their VR products. He's kind of just like, I think they basically call him a consultant, but from what I can tell, he just provides feedback on what Meta is doing in the metaverse. And a lot of that consists of like going into these internal chats and just like criticizing a bunch of things. And he's been very critical of this vision of the metaverse. He told a podcast host earlier this year that the amount of money that Meta was spending on the metaverse made him, quote, "sick to my stomach." He's made all these internal posts about how bad their VR hardware is.

• 15:55Kevin Roose?

  And basically just sort of saying like, this thing that you all are spending billions of dollars doing and assigning thousands of people to, it's not working. And so I found it fascinating that not only does that relationship exist, but that they're not shutting him up, that he's basically getting away with saying this stuff at a time when it doesn't seem like a lot of other people feel empowered to raise doubts internally about this metaverse strategy.

• 16:20Casey Newton?

  Well, so, I mean, if I'm Mark Zuckerberg, I actually think that that could be a great use of John Carmack, right? Because they're in this moment where they need to get better quickly. And what better way to get better quickly than to have a level 99 technomancer in the chat being like, "Hey, do better. This sucks," right? You know, it's not enough for just Zuckerberg to be saying it. Like, you need people who have this kind of credibility. So, I mean, look, you've painted a picture of a company that's having a hard time. I think it's undeniable that this is one of the most challenging times. In fact, I interviewed Andrew Bosworth, who is Meta's CTO, that I published a piece on. And I asked him, I was like, "Is this the hardest time in the history of your company?" And he said, "It's definitely in the pantheon." The other two times he brought up were like 2008 during the financial crisis,

• 17:04Casey Newton?

  when they had to like raise money and, you know, they didn't have a sort of proven business model. And then the other time was in 2012, when they were trying to shift from desktop software to mobile phones. And they were going through an IPO, which was really, really hard on them. So they've really only been two times as hard in the history of the company, right? And at the same time, like, duh, right? Like, of course, this was going to be hard. And I think it's important to step back and say, well, what were Meta's alternatives here? They could have stayed as Facebook. They could have continued with like Facebook, the app as their core product forever. They could have tried to compete with TikTok and just hope that that went well. But you know that that story is going to end at some point, right? That eventually people are going to get tired of their brand of social networking.

• 17:49Casey Newton?

  They're not going to be-

• 17:50Kevin Roose?

  Right, every social, every, every social media app dies eventually.

• 17:54Casey Newton?

  Yes, Valor Margulis is they, why did I say that? I don't know. It just sounded like you were saying something from Game of Thrones briefly. But every social, every social network dies, but not every social network truly lives. Now we're doing Braveheart. Anyway, the point of the story is that they had to invent something. And given the problems that they've had with Apple over the past couple of years, Zuckerberg knew he had to build his own platform. He had to own that platform. And he had the virtue of having this cash printing social network that would enable him to spend almost unlimited amounts of money to try to build it. So, you know, I, I think everything that you've reported on here is really interesting. And at the same time, I'm like, well, yeah, of course it was going to be hard.

• 18:39Casey Newton?

  Of course, a lot of the employees were going to think it sucked. They joined to do something completely different. You know, it's interesting, like, I'll talk to, you know, like people like Meta, like communications people. And they have pointed me, like, three or four different people have pointed me over weeks to this meeting that leaked where Zuckerberg was basically like, realistically, a lot of you don't belong here, which was an uncharacteristically harsh thing for him to say. But I think they keep telling me that because they want me to repeat it so that more of their employees leave. I'm not even kidding. Like, I think they want the people who are not all in to just get out now. And they want the people who are left to be the ones that think that there is something interesting about Project Metaverse.

• 19:18Kevin Roose?

  But the danger there, right, is that you get into a situation where everyone inside the company is a true believer, they're totally bought in, no one is giving you critical feedback, which is why I think the figure of John Carmack is actually kind of fascinating here. He's this person, he's totally like alone in his ability to kind of speak truth to power inside Meta. He doesn't need their money, he doesn't need their paycheck, like he could, you know, very easily retire tomorrow, and he's sort of just in there as kind of a fly in the ointment. And I thought it was really interesting that they elevated him and gave him his own sort of speaking slot yesterday. And I don't know if you actually watched his talk. Not yet. But it was fascinating. I mean, he was basically saying like, you know, he's running down some of the new features, but then he also like was pretty open about the

• 20:03Kevin Roose?

  fact that he doesn't really like their strategy. I mean, he says he's always advocated for doing like smaller, cheaper mass market headsets when they're going in the opposite direction, making this $1,500 Pro model headset. You know, he's, he at one point just said like, I'm gonna spend a little time being grumpy, and then just started rattling off. My favorite thing was he said that one of his internal posts had gotten reported for being insensitive. Like, he was so mean to the poor developers on the VR team that they actually took down his post. Like, on one level, a lot of the criticisms he's

• 20:43Kevin Roose?

  making are valid. On the other hand, like, imagine being the like 26-year-old junior developer on the team and having like the best software engineer of all time, just like looking over your shoulder and constantly just laughing at how small and pathetic your project is. But I think that it's worth teasing out, just for a few minutes here, these two competing visions of the metaverse. One coming from Mark Zuckerberg and the people who agree with him, and the other one coming from John Carmack and the people who agree with him, because I think it's actually kind of the entire story in a nutshell. So, very briefly, like, Mark Zuckerberg believes in what could be called like a tools-first approach to the metaverse. So basically, if you want to build the metaverse, first you have to start by building all of the capabilities.

• 21:27Kevin Roose?

  And the tools and the apps that people are then going to be able to use to like build stuff in the metaverse. You have to get the hardware, you have to get the software, you have to get all of the pieces and assemble them and give them to people and then they will make things that will entice other people to show up. This sort of tools first approach. John Carmack, however, is saying, basically, if you want to make the metaverse work, you have to start with something fun. You have to start with essentially a game. And there's some evidence that this is true, like Fortnite, Roblox, these are the largest sort of what we might call metaverse platforms. And they started off as games. And the thesis here is with the game first approach is that you build something, you make it very popular,

• 22:11Kevin Roose?

  hundreds of millions of people show up. And then you start expanding the capabilities of the game as Fortnite and Roblox have done, such that now when you go into the game, there's a lot more that you can do. And it kind of becomes a social media experience, even though you didn't explicitly set out to create that. So John Carmack's version is basically, let's build the popular thing first and then turn it into the social network of the future. Whereas Mark Zuckerberg is saying, let's build the social network of the future now.

• 22:41Casey Newton?

  Fair enough, although we should also point out that they have sold something like 15 million headsets so far. And it seems like the main reason that people are buying them is to play games. So like, yes, they don't have like a Fortnite or Roblox unto their own, but they have enabled a lot of other developers to sell their own games. And they have bought a lot of games, some of which do sort of resemble Fortnite and Roblox, right? There's one called, I think, Population One, that is somewhat Fortnite-esque. So I don't know that those are actually two diametrically opposed viewpoints within the company. I think arguably, you need to do both. And, you know, the argument, we should start kind of smaller and scrappier,

• 23:23Casey Newton?

  really just doesn't appeal that much to Mark Zuckerberg. I think Mark Zuckerberg is in a moment in his life where he wants to go very, very big.

• 23:30Kevin Roose?

  Well, and I get why he wants to do that. Like, the market that Meta is going after is bigger than just the gaming market, right? The biggest gaming companies in this space would be something like Activision Blizzard, which is this big gaming studio. It has tons of super popular games. And as of today, its market cap is about $57 billion. Whereas Meta's market cap, even after this 60% stock drop in the past year, is more than $340 billion. So it's a lot bigger than Activision Blizzard. And so I think it has to aim bigger than a gaming company if it wants to replace all of the revenue that it's going to lose as Facebook and Instagram continue to decline.

• 24:14Casey Newton?

  Yeah, and that's why, by the way, they're going so big on this productivity thing. You know, I think we should say, like, this Microsoft partnership is really interesting and maybe a big deal because, you know, we haven't talked about Satya Nadella yet on the show, like, very, very smart CEO. And basically, what he's done is swoop in and say, "We're going to leverage the $10 billion that Zuckerberg is spending to figure out this hardware, to get it in people's hands, and we're just gonna, like, make our software available on it. And like, maybe we become the default productivity operating system of VR without having to solve all of these incredibly difficult problems that Meta is going through right now."

• 24:52Kevin Roose?

  Yeah, I mean, you notice what he's not saying is, "We're gonna also spend $10 billion developing the metaverse." He's basically saying, like, "Look, buddy, if you're picking up the tab, Mark Zuckerberg, for building all of this hardware and solving all these hard problems, like, we will show up to the party, but we're not gonna, like, throw this party with you because that would be crazy."

• 25:14Casey Newton?

  Yeah, that's on you.

• 25:16Kevin Roose?

  So where do you land on this, where the metaverse is? Because I, you know, I don't, this is not a metaverse podcast. I don't expect that we'll talk about this every week or even close to every week.

• 25:24Casey Newton?

  We promise.

• 25:26Kevin Roose?

  Where are you right now in your thinking about Meta and the Metaverse?

• 25:30Casey Newton?

  Yeah, so I think you lead a slightly too negative on it like overall, but my skepticism is essentially that I think Meta is maybe just too early to this, right? So I'm willing to believe that VR, augmented reality, mixed reality, that these really do become the next big computing platform. I don't know that they'll become bigger than laptops and desktops, but I can see a world where most people have one of these things in their house. But if you even listen to the interviews that Zuckerberg was giving this week, he'll tell you like, we're three, four, five years away from this being a mainstream thing. It's a really long time in technology.

• 26:12Casey Newton?

  It's a really long time for a company that is struggling as hard as Meta is right now, even just to compete with TikTok and the sort of competitors of today. At some point during those next five years, Apple is going to come out with a headset. Apple, a company that people like much more, trust much more, has a record of sort of hitting it out of the park with their 1.0 hardware products. And so I think there is a world where we look back and say, Meta did a lot of important work in that kind of early Metaverse to figure some things out, to develop some kind of best practices, to do some of the early innovation, but that they may not capture most of the upside because they got to it just a little bit too early and other people kind of came along and knocked them off the chessboard.

• 26:57Casey Newton?

  So, you know, I talked to somebody who like worked on Metaverse stuff over there and left not all that long ago. And for them, this was essentially a big reason why they couldn't get excited about it, was that they worried that the company was just too early. So I share that worry.

• 27:11Kevin Roose?

  Yeah, and I think that's a difference that we have. And I'm trying to be open minded about this and I, you know, I am trying not to just write the whole project off because I think you're right that there is, there is something there. It's true that computing is going to be more embodied in the future than it is today. I don't think like the rectangular smartphone is the evolutionary endpoint of all technology. I just, I think that I'm landing sort of on the John Carmack side of things. A lot of what was shown off at this event this week was very impressive technologically. None of it looked particularly fun. And I think if I'm gonna spend a lot of time in the metaverse, it's gonna be because there's some amazing killer app

• 27:54Kevin Roose?

  or killer game or something that I like can't wait to get out my headset and go into. And then that thing gradually expands over time to the point that I'm spending many hours a day inside of it. But I think it's gotta be fun first.

• 28:08Casey Newton?

  So, I mean, I think there are fun things to do in VR. Beat Saber is a fun game. Like you can play games. You might also just like finish those games. And then maybe there's not another one that you think is as fun. I think VR needs to be 10 times better at something than something that you're doing on another device. And that's what I think it isn't yet. You know, there's a lot of things you can do in VR and almost all of them I would rather do on a phone or on a laptop. So I think that's when you're gonna know that this thing is really cooking, when you think, oh, yeah, well, of course, we're doing that in VR because it would be dumb to do it any other way.

• 28:36Kevin Roose?

  Totally.

• 28:43Kevin Roose?

  When we come back, the future of the gig economy.

• 28:57Speaker 00

  Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O U T S H I F T dot com. AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real time guardrails by team and model,

• 29:42Speaker 00

  helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 30:01Speaker 04

  Ripple has been powering modern finance for a decade. It's not our first trip around the blockchain, and it won't be our last. We're leading the charge to bridge the gap between existing financial systems and the future of finance. You can have it all: legacy, stability, and digital innovation. Systems that work with your other systems. Because with the right partner, the question isn't crypto or traditional finance. The question is simply, why choose? A new standard of modern finance. It's happening with Ripple. For informational purposes only. Availability of products can vary and may be provided by different Ripple entities.

• 30:36Kevin Roose?

  All right, everyone. Set a case here, you're rolling. I'm rolling.

• 30:40Speaker 05

  Oh, can I also just say, my dog is here and he's really barky. So if he jumps in and starts barking, we might have to stop.

• 30:50Kevin Roose?

  Okay.

• 30:50Speaker 05

  I feel like, as long as no one comes with a delivery or something, we should be fine.

• 30:56Kevin Roose?

  Okay, if we have a bark incident, we may have to re-record, but we will hope that he stays napping for now.

• 31:02Speaker 05

  Yeah, okay.

• 31:04Kevin Roose?

  All right, so we are here with Aaron Griffith and Noam Shiber, who are both reporters for The New York Times. Aaron, Noam, thank you for coming on our show. Yeah, great to be here.

• 31:14Speaker 05

  Thanks for having us.

• 31:15Kevin Roose?

  So, Aaron, you cover startups for The Times, and Noam, you cover labor and the economy for The Times. And we asked you here today because the areas that you cover are intersecting in a pretty dramatic way right now when it comes to the gig economy. These companies like Uber, Lyft, DoorDash, et cetera, that have been growing for many years, but that are now facing this new challenge of kind of a worker backlash. So, Noam, you've written about this happening at the regulatory level. The Biden administration, as you wrote this week, is now proposing new rules that would make it much harder for companies like Uber and Lyft to classify their workers as independent

• 31:55Kevin Roose?

  contractors and would force some of them to become employees, which would entitle them to overtime and unemployment insurance and minimum wages, et cetera.

• 32:06Kevin Roose?

  And Aaron, you've written about a different way that workers in the gig economy are starting to change the power dynamics in that industry by finding these kind of ingenious loopholes and ways to kind of game the system. And you wrote in particular recently about an app called Para that I just found totally fascinating and wanted to start by talking with you about. So what is Para?

• 32:30Speaker 05

  Para is a kind of funny, kind of interesting app that was created by a former Uber employee. His name is David Pickrell, and he worked in Las Vegas. And he would talk to drivers and realize that even though he, sitting behind a computer, had access to all of this information about, you know, if we turn this incentive up by 5%, we'll get X many more drivers on the road. And he had basically a dial that he could turn to try to balance this marketplace. Meanwhile, the drivers kind of were in the dark about, you know, when they were going to be getting certain bonuses and incentives, what hours they should be driving, where the rides were even taking them. And then all of the other costs that

• 33:12Speaker 05

  go into this hourly wage that Uber claims that its drivers make, I think it's $30 an hour, that does not take into account your gas, your insurance, your depreciation on your car, any of those kinds of costs. And he was just kind of appalled that, you know, these drivers, this information is available to them, but they don't have it, especially when they're driving. And so he created an app to sort of push back on that and give drivers a little bit more control and information about what they're doing. And so Para went viral last year with one of its features, which allows DoorDash drivers to see the tip in advance. DoorDash, if you leave a really big tip on a DoorDash order, they

• 33:53Speaker 05

  will sometimes hide it to prevent drivers from all just waiting back and only taking the high-tip orders. But basically, Para is an app for drivers to let them toggle between multiple apps and have a little bit more of the information advantage that these apps that are dispatching them, you know, hold.

• 34:14Casey Newton?

  It's amazing to me that something like Para was even able to be made because it it seems strange to me that this data is just available. Like, how is Para, is Para like inside Uber headquarters, you know, with a USB stick? Or where are they getting this stuff?

• 34:30Speaker 05

  Well, so, yeah, they basically found like a sort of hacky tech workaround in order to get this information. And then DoorDash cut them off. Because in order to use the app, you have to log in with your DoorDash credentials. And they kind of got into this little cat and mouse game where they would push an update to their app. A few hours later, DoorDash would do an update and and cut them off. And eventually, they realized this wasn't sustainable. And so now they're kind of retrenching and figuring out a new, a new way. Like, right now, basically, the app doesn't work for DoorDash. But they do still work with a number of other apps. They work with Uber. Uber recently, actually, after I asked Uber for comment, sent Para a cease and desist. So one of the big things is like letting drivers switch between the apps and

• 35:13Speaker 05

  sort of manage to be, to take the best orders from whichever app is available. That's actually kind of grown out of the pandemic. Because, you know, it used to be there was like Uber, maybe Lyft, and maybe DoorDash. Now, there are so many different apps that are using this. I mean, Amazon has something called Amazon Flex. There's a bunch of different, you know, Instacart. During the pandemic, a lot of drivers started driving for multiple apps. And so they're all kind of like, driving while also toggling between a bunch of different apps on their phone, trying to choose the best thing, trying to figure out which one is going to, you know, make the most money for their time. And so apps like Para, and there are a few others that that charge drivers are helping them kind of basically just navigate this.

• 35:55Casey Newton?

  Yeah, so it seems like the general trend here is that workers are getting more power. And many of these companies are not really consistently profitable. And I do think it sort of raises the question of what are these apps going to look like in two or three years? Should we expect them to still be around in their current form? What do you two think?

• 36:16Speaker 06

  You know, one thing that's interesting to me about this is, you know, there are three potential people who can subsidize this. The investors who were doing that for a long time, the consumers who more recently have been asked to do that, and then the workers. And I think who we identify as the party to subsidize this, kind of changes and is very contentious. And I think, you know, we've seen consumers revolt when they've been asked to subsidize a service. We've seen investors revolt by demanding that these companies get profitable. And we've seen workers certainly revolt in apps like Para have really helped workers revolt. And kind of where this lands and who ends up getting stuck with the bill, I think is kind of the fascinating drama that's going to play out over the next few years here.

• 36:58Kevin Roose?

  Yeah, it kind of seems like basically all of these services have like,

• 37:04Kevin Roose?

  there's sort of like a pick two thing. Like, they can either be, you know, number one, affordable for consumers, number two, fair to workers, or C, profitable as businesses. And you can kind of pick whichever two of those you want, but it seems like you can't really have all three at the same time. Aaron, does that sound right?

• 37:26Speaker 05

  Yeah, I mean, what I think is kind of interesting about this from the driver's perspective is like, this is just kind of like a gamified environment that they're living in. And they basically are trying to fight back and win, and figure out any little way to kind of get an advantage on this. And talking to drivers, they really like the flexibility. They don't necessarily want to be employees. I don't think they are all totally clear on what that would entail. And I guess if you work in this sort of like gamified environment where your boss is an app, and it's like pinging you constantly, telling you like, take this crappy ride, or if you do this, you get more money. And you're like constantly having to weigh these incentives in your mind. People are going to find ways to cheat the system, basically.

• 38:10Speaker 05

  And you see that from the marketplace side, too. Like DoorDash got in trouble a few years ago for what people were calling tip skimming. They were basically taking, if you left a really big tip, they would take part of the tip and use that as the base payment for the delivery.

• 38:26Kevin Roose?

  Horrible.

• 38:26Speaker 05

  And so people end up with like a lower tip. And they stopped doing that after there was kind of outcry from it. But this sort of thing with a hidden tip is another version of that. And another thing they do is they'll bundle a really bad order together with a good order to try to get people incentivized to take them. And like, from their perspective, it's like they're just trying to balance their marketplace. They want, when you hit that order button, for it to actually show up. That's the whole premise of it. And like, you know, the drivers. They don't care about that. They just want the the highest pay and us customers want to pay the least amount of money to get our food delivered to us. It's these incentives. You know, it it kind of, it kind of makes me feel like these companies, they're tech companies or they're valued like them, but their real innovation was just kind of like the arbitrage

• 39:11Speaker 05

  of figuring out how to put themselves in the middle of this marketplace. And we're starting to see that unravel a little bit. And at the same time, it's not going away. In the pandemic, we learned to be creatures of deliver what I want at the top of a button. And like, it only grew. And, you know, I mean, at least my personal like ordering of takeout in the pandemic like went up so much and it's probably not going to go back down. And now we can get prescriptions, we can get grocery, we there's it's expanded into so many different areas and I just don't see that changing. But it's possible that we might have to figure out that we have to pay more for it.

• 39:46Casey Newton?

  Yeah, I love living in this world where you're just like hungry at 6pm and you think, I guess I'll pay $36 for a burrito. Like, I guess that's just what it costs.

• 39:54Speaker 06

  Yeah, you know, gets back to this question where, you know, if you get to a point where you can't squeeze drivers more, either because they just refuse to work on the platform or because regulators and lawmakers forbid it, then you have to, you know, the costs have to go elsewhere. And that's where you get that $36 burrito, basically, you know, then it then it goes back on to the the consumers. And, you know, I think you can imagine a world where these services are, you know, similar services are, you know, are profitable and they treat workers fairly and, you know, customers are satisfied with them, but they're probably much smaller in scale, you know, I think for that number to work, you probably have to, you know, you can make it work in a super dense, super affluent place like

• 40:38Speaker 06

  San Francisco or, you know, Manhattan, whether you can do that in like the exurbs of Oklahoma City is is not at all.

• 40:45Kevin Roose?

  Yeah, it seems like what you're describing is really a scenario where in order for these services to continue existing and maybe even turn a profit.

• 40:56Kevin Roose?

  In a world where they have to classify, you know, some or all of their delivery workers and drivers as employees is that these things might become more like the luxuries that they once were, you know, it used to be a luxury to have a private driver pick you up at your house or to have someone bring you food to your house. It used to cost a lot more. And maybe what you're saying, what I'm hearing you say is that we should just get ready for sort of a return to that that status quo of maybe only the very wealthy can afford to get around in Ubers. Maybe only the wealthy can afford to get, you know, everything from, you know, food to marijuana to to other types of things delivered to their house on demand.

• 41:41Kevin Roose?

  Is that what you're saying? That these these sort of once affordable luxuries may just revert to becoming luxuries again?

• 41:46Speaker 05

  Maybe it's just like the stupid stuff goes away, like ordering a single banana from Instacart, you know, that that the driver has to drive 40 miles to bring it to me for $5. Maybe maybe that's not allowed on the platform anymore. But like, I mean, before like seamless even in New York, like you could get Chinese delivered by a guy who worked three blocks from you and that existed and that was free. But one other thing to add to that is just like Uber and Lyft, particularly Uber, operated for a really long time on the premise that eventually the cost will go to zero or to very low because, you know, autonomous vehicles will make this whole driver debate irrelevant. They'll have robo taxis.

• 42:31Speaker 05

  It'll be amazing. You know, I had on my calendar marked for January 1st, 2023, order a flying Uber because that's when they said their flying taxi service would be launching. I don't think that is going to hit its.

• 42:46Kevin Roose?

  You literally made like a Google Calendar invite to yourself to order a flying Uber on January 1st of next year?

• 42:54Speaker 05

  Yeah, I do that when tech execs, you know, make big pronouncements like, oh, you know, we're gonna, the other one that I had for 2023 was SpaceX was going to be doing passenger flights halfway around the world like. I could get in a SpaceX flight to Singapore for like a business trip. It would take thirty minutes. You know, they're doing great stuff, but I don't think that's that's anytime soon yet.

• 43:15Kevin Roose?

  Yeah, I wanted to ask about that because that to me is a fascinating piece of all this. I remember years ago, talking with an executive at one of these companies. You know, basically saying like, so what are you going to do when all these workers decide that they want to be paid more and they want to be classified as employees? And I think he literally said like, we won't have drivers by then. You know, like we will be, we will have kicked the drivers out of, out of the driver's seat because the cars will be driving themselves. Our three-sided market will become a two-sided market. And everything, you know, the cost of getting burritos to your door will be so low because of autonomy that we will be able to keep our current prices or even reduce them.

• 43:58Kevin Roose?

  And clearly, as you've pointed out, like that has not happened. But I'm wondering, Noam, if, do you think that the sort of regulatory push here to classify these drivers and other gig workers as employees is going to accelerate investment in autonomous driving? Like, how can we get to autonomy and self-driving cars as soon as possible?

• 44:20Speaker 06

  Certainly in other industries, we've seen automation as a response to increased regulation. I think the classic example is you raise the minimum wage and, you know, retailers and fast food restaurants turn to self-serving kiosks. And I think academics would debate how much of a correlation there is and how much you can replace humans with machines. But clearly, like, there is some acceleration in investment in automation when these things happen. That said, you know, I had similar conversations with Uber folks and Lyft folks beginning, you know, in twenty sixteen, twenty seventeen, certainly. And whenever I would raise these questions, these similar questions that you raised about drivers.

• 45:01Speaker 06

  Drivers were always regarded as this kind of intermediate step. And I guess the the answer is, will this accelerate the push for automation as well? They've been pushing pretty hard on the automation piece. And if they could have accelerated it, they probably would have because it wasn't like a scenario that has surprised them, you know, this idea that working with humans is messy and they get upset and they want to be, you know, taken care of and they want higher wages. So I think, yes, there will be an effort to double down on this. But that idea that can see has been there from the very beginning. And as Aaron said, just around the corner always proves to be like, you know, around a few more corners than you anticipated.

• 45:43Casey Newton?

  You know, and just to say real quick, in terms of, you know, how automation is going, last week, Amazon, which was making this scout home delivery robot, just pulled the plug on the entire project. And you got to think, if any company in the world was incentivized to spend whatever it took to just have robot deliveries, it's Amazon. And they just threw their hands up last week.

• 46:03Speaker 05

  Wait, do you guys remember the the delivery drones? Whatever happened to that? That was like the big 60 Minutes story. Are we coming up on the 10 year anniversary of that yet?

• 46:12Kevin Roose?

  Did your Google Calendar alert go off yet?

• 46:14Speaker 06

  We defer to you, Aaron. You're the keeper of the dates.

• 46:17Speaker 05

  I might have to skip skip ahead or back a few years to to find that one.

• 46:21Kevin Roose?

  All right. I think that's a good place to leave it. Thank you both so much for coming on and hopefully next time, maybe in two years, we will be having this conversation inside an autonomous flying Uber.

• 46:32Speaker 06

  Yeah, great to do it. Thanks for thanks for having us.

• 46:34Speaker 05

  Thanks for having us. Kyle is still sleeping.

• 46:37Speaker 07

  Aw, good dog.

• 46:50Kevin Roose?

  We're gonna take a quick break and then when we come back, I'm gonna try to blow Casey's mind with some developments from the world of AI.

• 47:08Speaker 00

  Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the Internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T dot com. AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model,

• 47:53Speaker 00

  helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 48:12Speaker 07

  Today's AI builders need infrastructure that scales with their ambitions. Iron is a vertically-integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Iron's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Iron, AI cloud for the next big thing.

• 48:46Kevin Roose?

  Casey, as you know, I'm obsessed with AI. And I think we'll be talking about AI a lot on this show just because it's like a lot is happening in AI right now. It's a period of extremely fast progress and development and just new things coming out every week. And we could probably do an entire show every week just about what's going on in AI, but then we'd never get to cover any other fun things like legs in the metaverse. So I thought we could just do a little segment called This Week in AI, where I tell you about some things that happened this week. That's great.

• 49:29Kevin Roose?

  All right. So number one, Google, which has already this year alone released text to image AI and text to video AI more recently, is now going into our field, audio. The company's researchers just released a paper about what they're calling Audio LM, which is a new system that can generate hyper-realistic audio clips, basically the way that GPT-3 or other language models learn to finish text. This model, Audio LM, can learn to finish audio. So you feed it a snippet of someone talking, it will finish the snippet. It even works on music. There are some amazing demos,

• 50:08Kevin Roose?

  and one of which I can play for you right now. So you'll hear it start to play some speech, and then at some point it switches over to just fully auto-generated AI speech, or piano, or whatever they fed it.

• 50:23Casey Newton?

  Okay.

• 50:24Kevin Roose?

  So this is the prompt.

• 50:26Speaker 08

  Nay, nay, lording, answered Wolf.

• 50:28Kevin Roose?

  And now it's generating speech.

• 50:29Speaker 08

  We know not how to call you lord or lady. We have lived too long in the forest.

• 50:35Kevin Roose?

  This is the prompt.

• 50:36Speaker 09

  Our first impressions of people are, in nine cases out of 10,

• 50:39Kevin Roose?

  And now this is the AI.

• 50:40Speaker 09

  mere skeptical reflections of the actuality of things. But they are impressions of something different.

• 50:46Kevin Roose?

  Here's piano.

• 50:52Kevin Roose?

  And now this is the AI.

• 51:07Casey Newton?

  So if I were to give this machine the prompt, throw your hands way up in the air, what are the odds it would come back with wave them all around like you just don't care?

• 51:16Kevin Roose?

  I would say, based on statistical analysis, we're not quite at 100% there, but we're getting there.

• 51:23Casey Newton?

  I didn't really, like, I was sort of trying to follow along. The only one that really landed for me was the piano one, because you can sort of see in the image, like, where the prompt ends and the AI begins. And the prompt is actually maybe, what, like, two seconds? I mean, it's a very short amount of prompt, but then, you know, it sort of goes on for maybe four or five times longer.

• 51:44Kevin Roose?

  Yeah, it picks up the original key and the tempo, and it just continues the music. And what I find very exciting about this for us is that, you know, right now, we spend all this time recording a podcast, we could just record the first 30 seconds and let the AI do the rest.

• 51:58Casey Newton?

  First of all, we would never, because on this podcast, we believe in the power of the human mind.

• 52:04Kevin Roose?

  This is artisanal audio production, baby.

• 52:06Casey Newton?

  That's right. There's gonna be a lot of synthetic podcasts out here, but on Hard Fork, you're always guaranteed real people and real talk.

• 52:13Kevin Roose?

  The real deal. So, okay, that's AudioLM. And now we're gonna go off to Microsoft, which announced this week that it is integrating AI-generated art into its suite of office software with a new app called Microsoft Designer. According to The Verge, "The app uses Dolly to let users type prompts and instantly generate a variety of designs with very minimal effort." Microsoft says that Designer can be used to create everything from greeting cards and social media posts to illustrations for PowerPoint presentations and logos for businesses.

• 52:47Casey Newton?

  Okay, so this is cool. This is legitimately cool. Like, the first thing that I did with computers was, like, desktop publishing. Like, there was a program called Microsoft Publisher, and I used to just design brochures. And I would, I wanted to, like, use every single font on the computer. Like, I was obsessed with this stuff as a kid. And the idea that, like, you know, today's kids will just be able to be like, whip me up a cool AI greeting card and just mess around, it seems super fun.

• 53:13Kevin Roose?

  I agree, genuinely cool. It's basically replacing clip art, which, you know, you used to buy on, like, CDs at Best Buy, and then it all went online, but like... I sometimes have to make PowerPoint presentations to give talks and things like that, and it's always my least favorite part of the process because I am so bad at design. And it seems like that's a problem that AI is already solving and is going to solve is just like, everyone is going to have the world's most beautiful slide designer right there in their PowerPoint, which is pretty cool.

• 53:45Casey Newton?

  I love that.

• 53:46Kevin Roose?

  So, okay. And then here is the real mind blower, which is that researchers have figured out how to use AI to do something called silicon sampling. Does that ring a bell?

• 53:59Casey Newton?

  Um, I saw the phrase, but I don't know what it means.

• 54:03Kevin Roose?

  Okay, so basically, researchers from Brigham Young University put out this paper in September that I just learned about this week. And it showed that GPT-3, which is this sort of large language model from OpenAI that we've talked about that can, you know, finish your text or transform it or write a poem for you, can also be used as a proxy for humans in social science experiments.

• 54:26Casey Newton?

  Oh, come on.

• 54:27Kevin Roose?

  So this is what they're calling silicon sampling. And basically, it means that you can use GPT-3 to survey simulated humans and ask those humans questions as opposed to interviewing the actual human.

• 54:41Casey Newton?

  But how is this real? That like, that doesn't make any sense.

• 54:44Kevin Roose?

  Okay, so basically, there's a whole, you know, there's tons of social science experiments that involve, you know, calling up people and asking them what they think about certain things that, you know, the whole field of like, political polling and focus groups, you know, billions of dollars a year are spent trying to figure out, what does some group of people, some statistically significant group of people think about something? And these researchers showed that you can get essentially the same results if you just train an AI like GPT-3.

• 55:18Kevin Roose?

  In a variety of personality types. So what they did, the study is very cool. They did actually three studies to prove that this silicon sampling concept worked. So they would create essentially artificial personas with GPT-3. So they would say, okay, you are a family values conservative. And they would give it some sort of training data to make it similar to what, you know, things we know that family values conservatives believe. And then when the model was trained, they would basically interview it in the same way that you would interview someone as a pollster. You know, ask them what, you know, who they're going to vote for, what they think about certain issues. And you could do the same with progressives. You could do the same with moderates.

• 56:00Kevin Roose?

  And essentially, if you do this the right way, you get very, very similar results to if you actually call up thousands of progressives and moderates and family values conservatives and ask them the exact same questions.

• 56:13Casey Newton?

  You know, so that's fascinating. And, you know, it comes to mind that, you know, generally speaking, pollsters ask people about relatively few things. And so maybe it's actually not that hard to like arrive at, you know, the same, you know, 10 conclusions that that people typically, you know, reach when you ask them about politics. The one thing I would say, though, you know, if these models are not dynamic, which is to say if they're not being fed new training data over time, it just seems like they would get old in a hurry, just because like the the opinions of the body politic are changing every day as new events happen and new controversies emerge, right? So I like, it might be sort of interesting as proof of concept, but if I'm a political pollster, like, there's no way in heck that I would want to rely on some sort of AI model

• 56:58Casey Newton?

  that's who knows how old.

• 57:01Kevin Roose?

  Right. And I think the way this is most likely to be used is in sort of a like a study design phase. So if you're a sociologist and you're trying to, you know, do this big survey that's going to involve, you know, sending out thousands of mailers or calling thousands of people, you might actually test it first on a sort of simulated. Audience to make sure that the study you designed is is realistic, that it's going to give you the kinds of answers that will be helpful to you in your research. So that might sort of be the first step toward doing some sort of large scale social science experiment. Just test it on the robots.

• 57:37Casey Newton?

  And if that's true, then that that just becomes another example of how AIs like the first way that we feel it is it is just going to be a tool everyone uses at their job to do their job a little bit better.

• 57:46Kevin Roose?

  Right, but it's I learned about this from Jack Clark, the AI researcher and former journalist. He has a wonderful newsletter called Import AI. So he wrote about this as basically being transformative for all social science experiments, but also having some potential for trouble. So one thing that you can imagine people doing with this knowledge that you can essentially simulate people at scale through these large language models is, for example, to to test out propaganda campaigns. If you are a government that's going to do some large scale, you know, manipulation of public opinion, you might test it on a million virtual citizens before you actually put it out into the world and see which one is the most likely to work.

• 58:31Kevin Roose?

  You might also use this if you are, for example, a fraudster who is trying to, you know, scam people out of giving you, you know, their social security numbers or their credit card numbers. You could actually test the scam on simulated humans, figure out how to make it more convincing and compelling, get a sense of how it's going to work on real people, and then go out into the world and do it on real people.

• 58:56Casey Newton?

  Man, that that really is terrifying. Although, of course, as soon as you say that, I'm like, well, yes, but if it works for those purposes, then like, won't every advertiser and marketer also want to run a simulated campaign on the entire population of the United States to learn what is the most effective way to convert them?

• 59:13Kevin Roose?

  Absolutely, and that's going to be fascinating as it works its way into targeted advertising because you're right. I mean, right now, there's sort of testing that you can do, you know, you show 50% of the audience one, you know, one. Marketing slogan and 50% of the audience, the other marketing slogan. But in the future, that's all just going to be possible to do through simulation on these AI-generated personas.

• 59:38Casey Newton?

  You know how in the rebooted Battlestar Galactica, like, the only reason Battlestar Galactica survives is because it wasn't connected to the internet? I'm just gonna say, you have me thinking about that right now.

• 59:48Kevin Roose?

  Yeah, you're ready to move to the mountains.

• 59:50Casey Newton?

  I mean, I don't know. When you talk about being able to, like, simulate how the entire population of Earth will respond to campaigns and just keep running those simulations until you've reached the maximally effective persuasive techniques, I don't know what defense mechanisms we are supposed to muster, like, in defense of that, right?

• 1:00:13Kevin Roose?

  Yeah, I mean, I think it also points to the fact that a lot of us are pretty predictable.

• 1:00:17Casey Newton?

  Yeah.

• 1:00:17Kevin Roose?

  I mean, if you can simulate the behavior of large groups of humans using AI, it kind of reflects poorly on us. Like, what does it say about humans if we can, if what our actual opinions are and beliefs are, are basically identical to what you would predict using just a set of demographic inferences about us?

• 1:00:40Casey Newton?

  Right, but most of us don't think we're predictable, which perversely makes us easier to manipulate, right? It's like the whole thing where people thought that Facebook was listening to their phones to sell them ads, when in reality, their behavior was just really easy to predict. And everybody's like, "Well, you never could have known that I wanted to buy a toothbrush, right? I just happened to say toothbrush, and like, you picked up on that." I was like, "No, like, like, Facebook knows who needs a toothbrush." Or at least they used to before Apple introduced app-tracking transparency.

• 1:01:06Kevin Roose?

  Now all they know is that we want legs.

• 1:01:09Casey Newton?

  That's right.

• 1:01:14Kevin Roose?

  We'll be right back.

• 1:01:23Speaker 00

  Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the Internet of Cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T dot com. AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model,

• 1:02:08Speaker 00

  helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 1:02:27Speaker 07

  Today's AI builders need infrastructure that scales with their ambitions. Iron is a vertically-integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Iron's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Iron, AI cloud for the next big thing.

• 1:03:03Casey Newton?

  All right, Kevin, let's do an Elon Musk vibe check. Last week, we talked about the saga of Twitter and Elon. And Kevin, you said the deal could close as soon as two weeks from when we taped. It has now been a week?

• 1:03:19Casey Newton?

  How are you feeling? Do you still think this deal is going to close?

• 1:03:21Kevin Roose?

  I'm feeling good about my prediction that this will close. It could be a week more, two weeks more, but I remain convinced that by Halloween, we will have an Elon Musk-controlled Twitter.

• 1:03:32Casey Newton?

  Well, that remains a bold take. I would just note that there does not seem to have been much news over the past week about Elon lining up those funding commitments from his banks. There have been really no positive or supportive tweets about why he intends to close this deal or affirming that he's really going to do it. He has instead just insisted that Twitter is refusing to accept his deal. So I remain on the highest of alert and look forward to checking in with you next week to see whether the vibe has shifted.

• 1:04:07Kevin Roose?

  Yeah, I'm, I'm pretty convinced I'm going to win this one, but as always, I could be wrong.

• 1:04:17Kevin Roose?

  This has been episode two of Hard Fork. As always, if you have questions, comments, reactions, guest suggestions, things we should talk about, you can email us at [email protected].

• 1:04:29Casey Newton?

  Yeah. Also, some people told us last week that you were having trouble distinguishing our voices, so I just wanted to say that if you hear a smart thing on the show, that was Casey. Hard Fork is produced by Davis Land. We're edited by Paula Schuman and Hannah Ingber. This episode was fact-checked by Caitlin Love. Original music by Dan Powell, Alicia Bat Etoop, Marian Lozano, and Sophia Lammon. Today's show was engineered by Rowan Nemestow. Special thanks to Shannon Busta, Kate LaPreste, Nell Galogly, Mahima Shabloni, and Jeffrey Miranda. You can email us at [email protected]. That's all for this week. See you next time.

• 1:05:44Speaker 00

  AI isn't just opening new business opportunities, it's changing the way businesses get work done. The questions they ask, the decisions they make, and the routine tasks they can hand off. And nobody understands that better than NetSuite. That's why they put together a free expert guide you need to grab today. Aligning for the agentic era, how AI is changing everyday work. Download it and you'll see exactly how AI is reshaping work across every part of your organization, and how to make sure you come out ahead. Get your free guide at netsuite.com/NYT. netsuite.com/NYT.