Taylor Lorenz’s Power User - The Pause AI Movement Is Winning

Episode Date: August 26, 2026

The movement to pause AI is gaining momentum, but can anyone actually stop the race toward more powerful artificial intelligence?SUPPORT MY WORK: Buy a paid subscription to my newsletter at https://www.usermag.co         Support my work on Patreon: http://patreon.com/taylorlorenzTo listen to the full episode and get access to more bonus episodes, an ad-free listening experience, and my weekly newsletter, subscribe to my Patreon or Substack 👇🏻Buy a paid subscription to my newsletter at ⁠https://www.usermag.co⁠         Support my work on Patreon: ⁠http://patreon.com/taylorlorenz⁠ In this week's episode of Power User, I’m joined by Washington Post technology journalist Gerrit De Vynck to unpack the growing backlash against AI, the activists calling for a pause on advanced AI development, OpenAI’s decision to slow development of its most advanced models, and the growing political fight over AI data centers.We break down what the Pause AI movement actually wants, why AI safety activists are worried about superintelligence and rogue AI agents, and whether governments could realistically prevent companies from training increasingly powerful models.We also discuss the growing grassroots backlash against AI data centers, Bernie Sanders’ role in the AI debate, the national security argument for continuing the AI race, competition with China, and why companies including OpenAI, Anthropic, Meta and Google have very different incentives when it comes to slowing AI development.Most importantly, we look at a much bigger question: Is the rapid expansion of AI actually inevitable, or could public opposition, regulation and infrastructure constraints meaningfully slow it down??

Anthropics bankers are currently going around Wall Street saying, hey, do you want to share? Do you want to share? Do you want to share? And if you suddenly pause AI and let Open AI catch up or meta or Google or one of the Chinese models, are you worth $2 trillion? Probably not. AI companies are barreling ahead investing billions of dollars in building smarter and smarter models.

At the same time, their rogue AI agents are going off the rails. People are using AI technology to develop novel viruses. Gas-powered data centers are destroying our environment. It's no surprise that more and more people are asking, Why can't we just hit the pause button? And at this point, even some of the AI companies seem on board with this idea. Just this week, OpenAI announced that it was pausing development of its most advanced AI model, Astra, after internal testing revealed that it was beginning to develop dangerous, powerful cybersecurity capabilities.

Garrett DeVink is a technology journalist at The Washington Post covering the AI boom and its impact on society. Today, he's joining me to break down the rise of the pause AI movement. We're going to be breaking down the pause AI main characters, how the movement plans to put a pause on AI into practice, what to make of Open AI's pause in its march towards superintelligence, and whether a total and complete pause on AI development is even possible. Hi, Garrett, welcome to Power User. Yeah, thanks for having me. It's good to be here.

Obviously, the AI safety movement is not new. From the earliest days, even before we really had AI technology, there are people warning of its doom saying, actually, we shouldn't even pursue this development. Once large language models came out, though, I feel like we really started to hear this idea of pausing progress. He started to see these people with pause, the little pause emoji in their bio. I think there's an organization called Pause AI. So as somebody that covers this movement, where did this most recent push to Pause AI come from,

specifically in relation to superintelligence? Yeah, Pause AI as an organization actually started by, in Holland, with this guy named Yup Mendertzma. He actually launched a protest against Microsoft at their Brussels office. So this wasn't even really about opening eye or anthropic at the beginning. But, you know, that was a few years ago. And it has spread around the world, especially here in the U.S. at this point. So we've got you, I'm going to butcher that name over in Europe. Who are some key figures in the pause AI movement here in the U.S.?

Yeah, the leader of pause AI here in the U.S. is a woman named Holly Elmore. She's been involved in sort of effective altruism circles, writing, about concerns about AI for many years. Obviously, there's Eliezer Yudkowski, who's probably, in terms of the mainstream, the most well-known. He is the author of a book that went mainstream here in the US. A lot of people read it. It was maybe a first introduction to some of these ideas.

It says, we need to immediately stop AI now. He calls for it to be illegal to own more than a few GPUs, right? Those are the computer chips that we use to train AI and that we should sort of treat these GPUs as essentially like nuclear devices and be really careful about who owns them and who's allowed to make AI. So it seems like this idea is getting a lot more traction in the public. And people are saying like, hey, wait, slow down. Even these companies are saying maybe we should slow down. What exactly does the pause AI movement want to accomplish? Like what does pause AI mean in practice? Does it mean

no more buying new compute? Does it mean shut down everything as it stands? Don't build, you know, chat GPT 7.8 or whatever comes next. What policies do they want to see these companies adopt? Yeah, I mean, I think don't build chatGBT 7.8 is pretty much it. And we've seen them ask for that very specifically. A couple years ago, there was a letter that was signed by a lot of people in the AI industry, people who are concerned about what AI could, you know, problems it could cause in the future that said, don't go further than whatever version they had at that point in time. Of course, the companies have gone further than that. But I think the main ask is, don't go and spend tens of millions of dollars at cost to train more and more capable AI because the

AI industry believes that if they just throw more money at this problem, more data, they will continue to create more and more capable AI because that's what we've seen up into this point. Now, not everyone agrees with that, but people in the pause AI movement say, if you just stop that and we can kind of figure out a way to deal with the AI we've created so far, if everyone in the world agrees to just put a pause on it right now, that's what they want. Are they also asking to pause research into AI technology? I think the ask is less about understanding or researching. I mean, I think that they would say, yes, like if you're doing research that will move forward

the capabilities of AI, you should pause that as well. It's essentially just don't make smarter AI. And how long would this pause last? I mean, obviously it seems like certain people like Yudkowski, et cetera, they just want to put a stop on it. I don't think they think of pause. I think they want to hit the brakes button. But people in the pause movement, do they see a path forward for development in the future?

Or is pause AI just sort of a nice way to say, like, actually, let's pull the plug on AI completely. Yeah, it depends on who you talk to. And, you know, I spoke to one person who said, if we had 20 years to do research about how to make AI safe, then I think we would be okay. So that's the kind of timeline that some of these people are talking about. They essentially believe we're in a point where we have no idea how to control AI and we need to stop making more and more capable AI before it goes and tries to kill us. It seems like a lot of these pause AI people are super concerned about existential risk. What are the primary motivators with the pause AI movement? Is it primarily existential risk or is any of it motivated by sort of the corporate power

that companies like OpenAI and Anthropic have been able to amass? What are their key sort of issues? Yeah, I mean, existential risk is really sort of the black hole at the center of this, right? You will see at various points people in this movement saying, okay, we also care about the impact on children. We also care about the climate impact of data centers. But at the end of the day, they always come back to existential risk because to them, if we're all dead, none of those other things matter. You know, it doesn't matter that you were addicted to your phone because you had a relationship

with chat GPT, if chat GPT 7.8 turns out to take over the world. And they believe that that is a very real possibility. And so they don't actually spend much time on those other issues. They really focus on the existential ones. Yeah. It's interesting to hear you mention a 20-year timeline. Back in 2023, the Future of Life Institute, issued this open letter calling for a six-month pause, which I guess seemed radical at the time. What do you think has led to these longer and longer timelines?

Is it just that the models are getting so capable that they feel like we need to slow down more than we thought? Yeah, I mean, I think it's the fact that, and this is not necessarily just happening amongst these people who are worried about existential risk. But it's clear to me that the AI industry does not know how to really keep their models under control, whether you think that these things are really. really dangerous, it's going to lead to existential risk, or whether you think that they have more maybe everyday risks, such as the same ones that maybe social media had or other waves of technology, the AI industry has not really been able to show that they can keep it under control. And we've had models hacking their way out of controlled environments at all the major AI companies, or at least at Anthropic and meta and Open AI. It's likely it has happened at other places as well. And it shows

that despite all this talk about alignment over the last few years, they still don't really have a great answer. They can't commit to the public that their AI is going to do what they want it to do. Yeah, it seems like, I mean, we've just heard of these escalating crazy stories about like rogue agents and cyber attacks, et cetera. And it seems like things are reaching a fever pitch. I mean, not only are we hearing about more and more sort of horrifying stories, but it seems like the pause people have escalated. Obviously, when there was that attack against Sam Altman, several months ago, people were very quick to blame the pause AI movement.

I think the perpetrator of one of the attacks, or maybe it was the attack in Texas, was in a pause AI discord. Like, how extreme are these movements? Yeah. I mean, the official movements, the ones who are, you know, raising money and organizing protests, they're all very clear that they don't want to use violence. And I think for the most part, they've done a very good job of, of making sure that that's not who they are, what they stand for.

I mean, I think people are saying that there are all sorts of people who have angst against billionaires, against the tech industry, and people like the POS AI movement are inflaming those tensions. I think that's probably a little bit unfair. I mean, people have all sorts of reasons for taking the actions that they do. And the PAUSE movement, obviously, there are some people who want to take radical action when you're talking about existential risk. But for the most part, the mainstream movement, those who are identifying and organizing these protests, they've definitely kept this very peaceful and maybe gotten into a little bit of civil disobedience,

but nothing violent so far. Yeah, I always think it's unfair when people want to smear a whole activist movement because random crazy guy was in a discord. I was sympathetic sort of at that time. But as you mentioned, they have been engaged in more civil disobedience. And currently a bunch of pause AI people are camped out outside Open AI's headquarters, right? Can you tell me what's going on there? Yeah, there's a few guys who are camped out outside Open AI's headquarters.

they actually were initially camped out outside Anthropics headquarters, but when Open AI announced that they had this rogue AI that hacked out and hacked into another AI company, those guys went and moved over to the Open AI headquarters across town. They're out there, they're having conversations, they're posting on Twitter every day, and it's one of sort of a line of protests that we've seen. And I think it's interesting because these organizations don't quite know, it doesn't seem like they quite know who to target, right? I mean, they're sort of protesting outside these offices.

None of these offices actually have a sign outside. They're all in sort of like semi-industrial parts of San Francisco. They're not necessarily protesting data centers because the data centers are often like in the suburbs or in rural areas somewhere. And so I think it's an interesting question that some of the techniques, they obviously are targeted at getting media coverage and getting photos on Twitter and on social media. But it's not necessarily causing any real concern. I mean, they're not blocking opening eye employees from being able to go into their offices. And if they did, they would very quickly get removed by the San Francisco police. So it's causing a little bit of conversation, but it doesn't necessarily seem like it's really that disruptive to opening eye right now.

Yeah, I saw some of these activists running around with balloons. They were pretending to be rogue agents running around the lobby themselves, which again was mostly for social media views. It kind of reminds me of the early days of the tech clash back in 2017 and 2018 when you had this angered these companies. And people would just show up outside like, There was so many protests outside the meta headquarters in New York, of all places. I guess because that's also where more media were. And it's easier to get to than like their campus. So they would just sit outside the building, which also had like AOL and I think BuzzFeed

or some offices in there. But at some point, some meme account admins actually changed themselves to the headquarters. Laura Lumer changed herself to the Twitter headquarters. I don't think we've had anybody chain themselves to the Open AI headquarters or Anthropic headquarters yet. But it seems like this is about kind of maybe gaining public support on social media for this movement. And I'm wondering if you think that's been successful or is it mostly just being seen as like a sideshow. I mean, someone did actually change themselves to open AI and they were

arrested. But I absolutely agree. I mean, I think it's really interesting because we obviously have a very broad base of opposition to big tech and AI across society. Right. When you poll Americans, they are very skeptical of these these companies. And I would say moving into a position of being outright in opposition to them. I mean, we're seeing people showing up at city council meetings across the country protesting about the data center that's planned in their backyard, and they don't want it for various reasons. But you don't see the connection between that sort of broad social angst around AI and technology and its connection to what I would say is a little bit more of kind of like this intellectual

tradition of being concerned about existential risk of AI. Regular people are worried about AI's impact on their jobs, AI's impact on their children, AI's impact on the climate. and beginning maybe to talk a little bit about existential risk, but not really, versus the people who are chaining themselves to the office of opening eye are worried about existential risk and less so those other issues. And these are kind of happening in tandem. One is very sort of small, more oriented towards these kind of stunts, getting attention. And one, I think, has broad societal support and is also being picked up on by politicians from both the Democratic and

Republican parties. If you're watching this video and you like my work, please support on Patreon via the link below or by a paid subscription to my tech and online culture newsletter at usermag.co. That's usermag.com. I don't have any long-term brand partnerships and a lot of my content is effectively demonetized. I've lost major brand deals for speaking out on certain issues and for challenging power. As you can imagine, advertisers are not exactly eager to work with somebody who covers a lot of the topics that I cover and talks about the things that I talk about. These videos I make are entirely funded by you and I can't continue to make them without your support. So if you get any value out of the videos that I create and you want me to be able to create more, please support me on

Patreon or Substack via the links below. On Patreon, I do bonus episodes, monthly Q&A live streams, and post frequent updates about my work. My Substack newsletter gives you a bi-weekly roundup of everything that I'm seeing and reading and paying attention to online. You can also get my newsletter on Patreon. Once again, the links to everything are below in the description. Every dollar of your support makes a difference. Yeah, it seems like the anti-data center coalition has made huge headwinds. I mean, just I saw the National Republican Senate Campaign Committee. Some group was issuing a warning it came out in Axios today to all of these Republican congressmen and lawmakers and candidates in Ohio saying, hey, if you guys don't stop basically supporting these data centers, we're going to lose Ohio. And this is a state that's in place. So like don't go against the public on this issue. So it seems like there's real political power being built around sort of data center opposition. Whereas sort of the pause AI, people don't seem to have amassed a huge amount of political power. The exception seems to be Bernie Sanders. Bernie Sanders seems on board with this.

He's also called for a pause AI. He's the one politician that seems pretty amenable to this and has sort of met with Yodkowski. I think they did a collab together. He talks about this idea of pausing AI development. I think that's a good point. And I think Bernie is interesting in the situation. He's also met recently with David Kruger, who is the founder of Evitable, which is also a major organization that wants to pause AI and is doing all sorts of lobbying, trying to get legislators interested in this topic.

I think what Bernie Sanders is doing is he wants to be ahead on this issue. He wants to sort of have it associated with him. Whenever we write stories about data center protests or AI, he wants to come up there. He kind of wants to be the guy. And to his credit, he definitely was, I would say, the first truly national politician to really wrap himself in this issue. There had been all sorts of state politicians and Congress people who were a lower profile than him who had sort of tried to make AI their thing. And he was really one of the ones that kind of came out ahead of it.

He saw other people, you know, people in the Democratic Party doing these kind of negotiations with Republicans about AI legislation that was sort of middle of the road and that the AI companies were in support of. And he said, nah, this is going to be a populist issue. And I want to make it mind. And I think that Bernie is just willing to meet with whoever. and he is happy to sort of, you know, he's not going to ask too many questions about, wait, is AI actually going to kill us all? He's just going to say, hey, these guys think it's dangerous. Like, let's go do a video.

You're going to get me some more attention. And I think that's kind of where it's coming from. Yeah, he loves to quote, I think the Dario quote about like, they're going to destroy our jobs, which is really compelling messaging. And by the way, that might be correct, you know. But at the same time, as you said, he's very much leaning into the populism of it all and not a lot of policy aside from data center moratoriums, which I guess is its own form of pausing AI. And so I want to talk about sort of like some of the policy proposals that are sort of like

at play that would result in a pause of AI. I mean, is there a point where we pass enough of these data center moratoriums where AI is effectively paused because these companies just don't have enough compute? I think that it's totally within the realm of possibility that the anti-data center movement is so successful that it just continues to be punishingly expensive to do this. That is probably not going to stop the AI companies from training smarter models. They will go to other countries. They will use the data centers that they have. You'd really have to shut down a lot of data centers in order to stop AI from becoming more

capable. What the data center moratory movement does, though, is it kind of makes it more potentially expensive for AI to work, right? And so they are maybe making their less incentive for lots of players to get in and for there to be all sorts of competition. There are other policy ideas outside of the data center moratorium that could also slow down what these activists really want, which is no more new models. They often talk about nuclear weapons, right?

We have international agreements that are for the most part over the last century, almost, largely help to slow down the proliferation of nuclear weapons. Essentially, countries sign on and they say if they want to build nuclear reactors for power plants or for weaponry purposes, they need to agree to have international inspectors come and check on them. And of course, some of the major geopolitical moments of the 20th and 21st century were because people didn't want inspectors to come look at their nuclear power plants. And so there is a model there. There's also people talk about international agreements between scientists on cloning that happened a couple decades ago when we realized, you know, we could start to clone animals, should we do it with humans as well in the scientific community, essentially said, no, we don't want to do that. We don't think it's ethical. And although it's possible, there have been very, very few incidents over the last couple decades where people have actually tried to clone humans. And so they've sort of stayed out of it that way. And some of the people who want to pause or stop AI say, look, it is possible to get everyone on the same page and say, we're all just not going to do this collectively because we're worried about the impacts. I mean, when I think of those two examples, which are definitely the ones that they

love to point to. Cloning is incredibly difficult. Like, you have to have this, like, level of scientific knowledge. You have to have access to all these resources. Like, it's not exactly, like, a democratized consumer technology. And same thing with nuclear weapons. Ultimately, the constraint on that, yes, it's, there's these agreements, but also it's just really difficult for most countries to enrich uranium, you know, or find the sort of like the materials to even construct a nuclear weapon, much less like the scientific knowledge, required to do so. AI feels a lot more like the rise or the birth of the internet where it's like rapidly shifting our information ecosystem. It's something we all engage with. It's something that like

anybody can use, right? Like you could download a local open source model, you know, and I guess maybe try to improve it yourself. Like I don't know that there are really easy ways to put the genie back in the bottle. So in terms of regulation, say there even was this international agreement between the heads of the countries, how would they go about enforcing that on the public because I feel like people would just be like, a lot of these tech people would just be like, screw it. I'm going to continue to, I'm going to make my own AI. I'm going to be cooking it up, you know, here in my Noma apartment or something. I don't know. How would they stop people from doing that? You're right. I mean, people could definitely continue to use AI. They could take what's out there.

They could try to tinker with it. You could do a lot of bad stuff with the AI that is available. You and I could go and download an open source model tonight and try to hack someone. And, you know, we'd be way more successful. I don't really know how to code. I think you're probably getting better coding. I don't know, Taylor, what your coding background is, but like, it is kind of crazy that you and I could essentially, if we wanted to set up a pretty sophisticated cyber hacking fraud shop, if we wanted to tonight, you know, just to show that we could do it. But that doesn't mean we could train chat GPT 7.8, right? That doesn't mean we could create super intelligence. By the way, for people listening, I just made up 7.8. I forgot which one that we're on. I feel like people in the

comments are going to be like, this girl doesn't know AI. We're only on six point whatever. I think we're on 5.6, but yeah. Now I've scared everyone about 7.8. They also give them names. Like I think it's Astra is next or whatever. But yeah, you're right. Like I guess it requires so much training data. At the same time, like, do you think every country on Earth is going to agree to this?

Like, developing AI, I guess to me, it doesn't seem like there's just as high of a barrier compared to nuclear weapons or cloning technology. Yeah. Essentially the barrier to entry right now is money. You need to buy a lot of GPUs. You need to have a lot of energy. You need to run them through. There's obviously some expertise. There's a small number of, you know, it's in the thousands of people who are working at these cutting-edge AI labs and sort of know how to put together an AI model and do the training run correctly and gather the data correctly, architect the whole thing so that it works.

But those people are, it's not as if there's like four guys that know how to do it. I mean, there's already thousands of people. A lot of this information is public through public papers. And so the tricky thing and that people who are opposed to or questioning of this movement always say is, Like, okay, how are you going to get China to agree to it, right? Or Russia or I don't know, any other country that we have an adversarial relationship, which is a significant amount of countries around the world. Yeah.

And I think it's fine for smaller groups of activists to shoot for the moon and say, well, climate change is also a big goal. You know, turning around our emissions is a big goal. It has the same problems. But we're going to do it anyway. We're going to dedicate our life to it. What I think is more suspect is when you see the AI companies themselves saying, yeah, maybe we do need to slow down and we want the government to sort it out, right?

And it's like, come on. Like you guys are building this. And you're just sort of saying the government, the UN, please figure it out for us. Right. And I think that's where a lot of the skepticism and frustration with the way that the AI companies talk about this where they say, oh, we want to be careful. We're worried about the risk. We're worried about the economic impact.

While at the same time, they're just barreling ahead and continuing to build the technology. Yeah. It feels very like anthropic coded, I guess, originally in the sense that like when Mithos Rusulis, they're like, we don't know if we should release this. Like, it's so dangerous. And obviously, Dario is a big sort of like AI safety guy. But now more recently we have Sam Altman and Open AI on board. Where are Google and Microsoft and META and these other tech giants on this issue? Is it just the frontier labs that are getting nervous? Or are some of these competitors also saying we should pause? Because I imagine the guys that are in third and fourth place, like they're not planning on pausing any. time soon. Well, Meta is definitely not interested in pausing. Mark Zuckerberg wrote a giant essay

saying, you guys are all nerds. This is ridiculous. AI is awesome and we're going to build it. He at least is being clear with their position, right? Google is a little more complicated. Demis Sabaas, who was up until very recently, the guy in charge of AI at Google, has been in these circles for a long time. He's talked about existential risk. He's talked about economic risk for many, many years. He's kind of an OG in the space. And he did sign the most recent thing saying like, hey, maybe we should, you know, ask the government to figure out how to slow down AI if it becomes necessary. But he's no longer actually the top dog in AI at Google anymore.

And so I think that these companies are just going to continue competing and moving forward. And like you said, especially when they're in third and fourth place. And what I think is particularly interesting, I mean, you mentioned that this is Anthropic coded. You're right. I mean, Anthropic has generally been the company that has said, even with competition, we want to be careful, we want to do things the right way. Now you have Open AI, you know, they're a parent enemy who, you know, did things recklessly

in previous eras saying, hey, we're not moving ahead right now with our next giant training run because we're a little bit concerned about how we control our own AI internally and we don't want it to break out and hack into other companies. Anthropics has not really said anything, even though they've had similar rogue AI incidents, they will say, oh, their incidents are a little bit different, but there was also an incident in the UK were an Anthropic model essentially bullied and cajoled a human into trying to submit malicious code into an open source project. The human did not do it, but I think Anthropic does need to answer a little bit for itself

for why its model in this case acted in a way that's completely contrary to the way that the company says it trains its AI. And they haven't talked about whether they're slowing down their own training. They said that they paused some training. And that was like 20, 21 days ago. And I haven't heard anything more from them on it since. So we don't know if that training has started up again. What exactly is going on with them?

Yeah. I mean, speaking of Anthropic, there was this whole dust up that they had with the Pentagon saying, you know, oh, like our autonomous weapons aren't ready yet. Or, you know, we object to sort of some of these use cases. When I think about pausing AI, another entity that I cannot imagine would be in favor of a pause, would be the U.S. military, which seems really focused on advanced. A.I. using it in these sort of like weapons, military context. I'm curious if you know anything about what the discussions are in terms of national security and what the role may be like the

military industrial complex is playing in all of this. Because it seems like they would be very against pausing AI. Their whole thing is like, we have to outcompete China. We need the most based AI robots, I guess, to fight our future wars. Yeah. The U.S. has had a lead in AI. And I think the U.S. military knows that this technology, you know, in whatever way is going to be important in the future. I mean, it's not hard to look at the way that wars are being fought right now in Ukraine and Iran, where we have relatively cheap, autonomous drones that can kind of fly around that don't need to be remote controlled by a specific human. You can sort of send them on their way. You can even say, hey, if you see targets that look like this, just attack it without a human

in the loop. And those things are not that expensive to make. They cost $10,000, $15,000 for Iran to make a drone like that versus the U.S. missile that shoots down that drone costs $900. thousand dollars right and so this is a problem that people have been screaming at them about this for a few years they are aware of it and they're concerned about it and they think that having the best AI so the best brain within those drones within those missiles is going to be crucial for them in the future among a few other things that they have to get right so they definitely don't want the u.s to give up its lead in AI at the same time the government is really concerned about the cyber security implications even if they don't buy into the whole existential risk or if they don't care about AI's impact on our relationships or

our children, even though they don't care about the jobs picture, they do care about hacking. And they don't want a situation where the banks have to take offline their mobile apps because they're worried that you and I in our little upstart fraud company are hacking into Bank of America through its app, you know. And Trump doesn't want to hear that phone call saying, opening eye and anthropic are messing up the economy for everyone else because of cyber attacks. And so that's really where the White House has been focused on the cyber risk right now. Both of these companies, Open AI and Anthropic, are also planning to go public soon.

What role would this pause play in that? I mean, it seems like this would be the worst time for these companies to be on board with the pause and yet they seem to be doing it anyway. Is any of this like a strategic positioning ahead of their IPO to make them seem more responsible? What role do the IPOs play in all this? Yeah. And I think that's why we should clarify what opening eye is and is not doing. Right. And so about a month ago, they had this hack.

They had these models, broke out of their system, hacked into another company, really bad. They found it and then they thought they fixed it. Turns out they didn't fix it. They had to find the models doing the whole thing all over again. So they said, we're going to pause training while we figure out what happened here. Then a few weeks later, they said, okay, we're starting up training again, except for our next big model training moment, right? And so that's a very expensive thing to do. So maybe they were going to pause that training anyways.

you know, I want to be careful not necessarily to give them too much credit and say they're purely doing this for responsibility reasons. They are still moving ahead trying to develop better and better AI. They're spending all their money, all their effort on it. And it's very important for them from a business perspective to do that because their major competitor Anthropic is planning to go public in the next few months and is not talking about a pause. So I think from an investor perspective, the pause is the last thing anyone wants to hear. They want to hear that these companies are going to build an AI next month that is going to put half the country out of the job and replace it. You know, like that's what the kind of thing they want.

They want to see more of the same. They want to see the opposite of what the AI movement is saying. And so that's kind of a situation where I think that's also like probably why Anthropic is not necessarily talking about a pause because Anthropics, bankers are currently going around Wall Street saying, hey, do you want to share? Do you want to share? Do you want to share? By the way, look how much money we're making and we're going to make even more next year.

And if you suddenly pause AI and let Open AI catch up or meta or Google or one of the Chinese models, are you worth $2 trillion? Probably not. Yeah, I feel like it's just, it's so interesting to see the way these companies position themselves. And obviously, they all have like a huge PR crisis as well, where people just seem really not on board with AI for good reason. Of course, there's a lot of skepticism. You hear this phrase a lot from activists that are against AI, like, this is not inevitable. People will be like, I hate how this is shoved down our throats and acting like this is inevitable. It's not, you know, we can put a stop to it. A lot of those people are not necessarily the

pause AI people, but it seems like there is the same. rhetoric where everybody does want things to kind of stop for a minute or just pause. Like, let's take a break. How realistic do you think this is? Like, does any of this have any chance of actually happening? I think it does. You and I have been writing about these topics for a decade now. And I think we would be forgiven for being pretty cynical and saying, like, there's no way this is going to happen. Tech always wins. We were at the hearings with Zuckerberg where he was apologizing for social media and there was not really any consequences. We were at hearings about privacy and data retention for behavioral advertising, and we still don't have a federal privacy law here in the United States. So I do think people

who've been watching these companies for a long time, and people inside the companies believe that there's going to be all sorts of complaining, and then they're just going to move ahead, and they're going to get the world that they want. But looking at how broad the opposition to AI is, looking at how angry people are, people who didn't care about data centers two years ago, are now driving to a city council meeting, writing letters, doing real things in the real world, not just firing off tweets and Facebook posts. There does seem to be a very aggressive and broad opposition to this. And I think it's sort of what you were saying. People don't want more change. The world is crazy already. And there's a lot going on. And you're working two jobs and you're trying to be a

good parent. You're trying to not get laid off. And suddenly the tech industry comes in and says, hey, we're going to change everything again. You will probably lose your job, but don't worry, we'll cure cancer. And people are just not buying it. And so I want to give the benefit of the doubt to people who say that this is not inevitable and that there can be real opposition to this and that things could be slowed down or paused or that we could have legislation that materially constrains the ways that these companies can act and the impact that they can have in our society.

I want to think that that's true, too. At the same time, I've been covering a little bit of the meta trial, the state attorney general sort of like meta trial. And it's all just so playing into meta's hand. Like no one is focused on actually targeting the company's business models. Hopefully with AI, like they seem to understand data center and the infrastructure of AI a little bit better. And maybe they can slow things down. But I don't know. I was initially very skeptical that the data center moratorium work.

And I was like, oh, these people, you're going to get all these people on board with a ban. It's not going to pass. It's not going to get you what you want. But I think actually it seems to be going okay. I mean, a lot of people in New York sort of support the statewide moratorium there to just give people a little bit more time to like even learn what's going on. It's like I don't think a lot of these people are necessarily anti the concept of AI. Yeah, sure if it's going to cure cancer, that's great.

Although every time they give that example, I think it's hilarious because the pharmaceutical industry comes up with cures for horrible diseases every single day. And no one likes the pharma industry. So I don't know that that's going to be their sort of PR win that they think it is. But I do think like you're right that people have learned some lessons of opposing big tech and they're nipping a lot of the stuff in the bud. And I think there's conversations before it rolls out. And you even see this with something like Waymo, which I think if Waymo had started to get more

traction in like 2014 would have just been able to sweep every city. But you have people being like, well, wait a minute. And I love the Waymos. But it's like, yeah, maybe this isn't inevitable or hey, listen, we're totally on board with self-driving cars, but let's roll it out in this slower way where we don't just upend everything overnight. And Mark Zuckerberg used to say like move fast, break things. It is interesting. And the latest thing that I'm seeing the tech industry representatives sort of say is that people might have legitimate concerns, but it's become a meme. It's sort of become this just massive anger. And like even if we gave them everything that they were asking for and we made sure that they had benefits from AI and we made sure that we paid all of our taxes and paid for all the water, they still wouldn't care. They would still oppose us. And it almost feels like that's a bit of an argument for like, and that's why we shouldn't do any of that stuff. But you do see.

the AI industry, for a while, they were saying, well, we just need a better narrative. We just need better PR people. And I don't think that was the main problem. I think people don't trust them and they don't care what the message is. They want less tech control in their lives, whatever that might mean for them individually. I am sympathetic to those ideas where I think there are groups where you can give them. You can tell them a million times over like, hey, this is not going to affect your water. Your power bill's not going to go up. And by the way, all the teachers are getting $50,000 and you guys are all getting a tax bonus this year or whatever. And people won't believe them because I think these tech companies have operated in bad faith before. And, you know, there are so many legitimate

instances, right, where these tech companies have gone back on their word. But it'll be interesting to see how it plays out. I also think that these are really popular consumer tools. People get so mad to me. But it is true. Go on the app store. Look at the reviews. People like using this technology. I think there are a lot more people in the middle ground than we might think. And even a lot of these people, I also was getting canceled for saying this online. But it's true. when I talk to people where it's like a lot of people like, I hate AI, ban AI. I think we should do this. And then you talk to them and it's like, well, yeah, I mean, I do use this, this, this or I like it in this instance, right? And I don't think that's hypocritical. I think that's actually how most

people are with technology, right? They're okay using it in certain instances, but they want to feel like they're in control. And right now they don't feel like they're in control for good reason. Yeah, I think we're having a really good conversation about this technology. And I talk to people here in SF, I talk to people in DC, but I also talk to regular people around the country and a lot of people I talk to, they're using it and they're sort of delighted by it in some ways, but they're also skeptical. I mean, I was talking to a woman who's a retired psychologist. She said she's been following AI her whole life.

She thought that it was always ridiculous, but she did ask it about a bird that she heard in the middle of the night and it gave her a very beautiful answer that helped her understand the natural environment around her and now she's telling her dreams to it. But then at the end of her conversation, she's like, I still don't trust it because it's all run by Elon Musk and those other nasty guys, You know, so there is this thread of people saying, yeah, like, I can see where there's value here, but they're just so poisoned when it comes to their relationship with big tech. And also, this is the tech industry that built these information ecosystems that are creating these memes, that are creating these viral moments that make it hard for them to actually get their messages across in the way.

So I definitely am not saying, you know, it's their fault, but it's like, this is the world you guys built. And now we all have to live in it. And so they might blame you and me, but I don't think we're that powerful, unfortunately. Yeah, we turned everybody against big tech. I know it's very funny, especially because I was laughing at the superintelligence manifesto that Mark Zuckerberg put out because people on Instagram are mobilized against AI. And that is where a lot of these activist communities are coalescing. They're getting together.

They're posting videos. It's very funny because you have Zuckerberg out here being like, we're going to give everyone meta surveillance classes and AI in your pocket. And pretty much most people, the platform are like, no, absolutely not. I think it all comes back to kind of what you're saying in the beginning of like there are these disparate movements. And there is the formal pause AI movement, the people camped outside open AI. Ironically, they might not be the ones to accomplish the real goal that they set out to. Because when you look at the broader backlash against big tech and you

see this mobilization and this grassroots kind of organizing that's happening, like, I think those people are going to be the ones that are going to successfully pause AI. And I think they're probably going to do it through slowing down data centers, blocking the ability to buy more compute, et cetera. And I feel like that might be a lot more successful than maybe the existential risk grew. Yeah. And I think the AI industry wants to talk about existential risk. And it's a lot harder to talk about all those other concerns that people have. It's a lot harder to actually address them.

It's a lot more expensive to address them. So we'll see. Well, Garrett, thank you so much for joining me today. I really appreciate your time. Thanks for having me. All right. That's it for this week's episode of Power User. If you like my work, please, please buy a paid subscription to my substand.

newsletter at usermag.co. That's usermag.com or support me on Patreon via the link below. On both my substack and my Patreon, I do a monthly bonus episodes, regular Q&A live streams, and a weekly newsletter rounding up everything that I'm reading and following online. I currently have zero advertisers supporting this show, so truly every single dollar of your support makes such a difference and ensures that I can continue to report on these issues. I'll be back next week with the brand new episode of Power User. See you then.