Calm Down, There’s No “AI Apocalypse”

0:00I'm Nick co-host of the best one yet. You optimize your workouts, your portfolios, and that rando startup idea that hit in the middle of the shower. But if you're not hydrated, you're already playing catch-up. So we keep a stash of liquid IV in our studio. One stick mixed into 16 ounces of water is clinically demonstrated to hydrate faster than water alone with three times the electrolytes of the leading sports drink, plus eight vitamins and nutrients. Whether you're traveling, exercising, or podcasting the best business news ever, liquid IV makes it easy to stay hydrated wherever life takes you. So shop now at liquidIV.com, liquid IV hydration that goes wherever life takes you. Think about how you describe your ideal car to a friend. Good on gas, highly reliable, nothing over 20 grand. Now describe it to car gurus the same way. With guru, you search in your own words and get real listings that match. No endless filters, just the car that fits your life. It's no wonder car gurus is the number one most visited car shopping site. According to similar webs, estimated traffic data, buy or sell your next car today with car gurus at car gurus.com.

1:03Go to car gurus.com to make sure your big deal is the best deal. That's ca-r-g-u-r-u-s.com. Car gurus.com. Hey, I'm Kate Lindsay and you're listening to ICYMI or in case you missed it. Slates podcast about internet culture. And today on the show, we are joined by an ICYMI favorite, Ed Zittron. Welcome, Ed. What's up? Ed is the host of Better Offline, writer of the newsletters where's your ed at. And is one of ICYMI listeners' favorite guests if the Spotify comments and YouTube stats from our last recording are anything to go by. They love you because, and this is a batch of honor as far as I'm concerned, you were AI's number one hater.

2:06And we are here because there have been a number of apocalyptic headlines about AI these past few weeks. We're gonna spend this episode explaining and crucially debunking most of these claims, but first I wanna talk to you about another AI horror, AI actress Tilly Norwood. Now, since you and I are both in media, we both, it sounds like received this email that had the subject line, Tilly Norwood is ready to talk. Ed, did you chat with her? Were you excited? I had to block the email. I had to block the email immediately because I wanted to respond with something that I wouldn't have regretted. Yeah, right, right. I wanted to be like, can you put me on the phone with the person who created this? I have some words for them. Yeah, so this email was interesting to me because also, so Tilly Norwood, this news broke last year around this time that this place had created a sort of an AI actress named Tilly Norwood who, as far as I'm concerned, has done nothing but just like, use up energy and data centers.

3:08But now they're saying she's finally ready to talk and people can go onto the website and choose, well, they get two minutes with her free, which I already hate talking about. I got any audition, I really be calling it her. But two minutes free and then 15 or 30 minute sessions charged at around $14 American dollars or $25 American dollars. And I think whatever they're doing is stunt. The thing that's wild to me is the number of mainstream publications that have now published a quote-unquote interview with her. I so far have variety, CBS, wired, NBC. NBC, I think NBC also put her on air. Good fucking job. It's so, and they're all the same thing. And I just wanna be abundantly clear about something. Doing a quick, chungress, make fun of it interview is just as bad. Yes, yeah. Platforming this in any way is a terrible idea. I don't care if you're like, oh, I'm making fun of it. No, you are not, you are platforming it. CNN, NBC, CBS, wired, shame, shame, shame, variety, shame.

4:11You have no soul if you take this on. And it's also just fundamentally uninteresting because they're all kind of the same thing of her just being like, it's not like she's programmed to do anything fun. It's an LLM. It's an all-ch language model. I'm going to scream. No, and it's just like, I don't wanna read your chats with an LLM and that's what this, it just feels so lazy but ultimately meaningless. And like you said, does nothing other than kind of make it be this like, who fun silly thing rather than like, what this thing is representative of is like a genuine threat to livelihoods. Oh, I must be clear. Most people took it seriously. Some of them were like, oh, I'm making fun of that. It kind of reminds me on a much lower level of when in like 2015, New York Times did an interview with Jordan Peterson. Oh, I was gonna say LLM, but that's even fun here that is sort of weird. LLM, it was also a thing, but like, I mean, it's the same thing with they were like,

5:13oh, well, we're letting them hang on their words. No, you're not. You're putting them in the newest paper. Right. And it's just, I realized, Tilly Norwood, actually, no, I find Tilly Norwood vile, but Jordan Peterson far worse. Right, right here, right here. We're gonna rank them. But it's like, it's the same thing of like, yeah, we're gonna look, look, we got this big pile of dog shit, you wanna smell? Right. It's bad, it stinks. But look at the big 4K picture we took of the dog shit. You wanna taste? You shouldn't eat it. I'm gonna eat some of it because it's bad and you need to know, no, it's just by talking even about it, actually, analyzing it's a good idea. Being like, ooh, I'm gonna talk to this thing. Look how bad the responses are. You've played into their trap expertly. You stomped heavily on the switch to make the bear trap crush your leg and then you went, gotcha. I'm just like, interview your wifi router or something. Like it's just like, if you wanna have that experience go like, you don't do something else with your useless time. Oh no, sorry, I gotta talk to Tillie Goddamn Norwood.

6:14I gotta talk to the customer service chatbot on JetBlue. Here's the conversation I had with the AT&T chatbot. No. It's fucking useless and it's ugly stuff. This is how financial bubbles grow. This is how grift is grift. When the media does not know how to see a grift and stop it. When the media goes, oh, I'm gonna get a lot of clicks to this and then I'm gonna frame myself as the moral pie. I'm gonna say I'm morally, I'm, I'm need to make sure people see this. Actually it's your job to make sure they don't. Yeah, yeah. And there's a particular irony to this Tillie Norwood happening right now in the fact that like press outlets are kind of eagerly participating in what is really just an AI publicity stunt because at the same time they're interviewing an anthropic, quote unquote, whistleblower who fears AI is going to lead to the end of humanity. And maybe that is because both subjects are full of shit. When we come back, why we're not entering an AI apocalypse?

7:24Have you overheard of this term sleep maxing? Because I actually wrote a piece about this a few years ago and it is the way that people have started taking supplements or wearing things like mouth tape and buying these expensive trackers all to improve their sleep. It gets overwhelming pretty quickly and the truth is if you are looking to upgrade your sleep, all you need to change is your bedding. Bolin Branch has their signature satin sheets, breathable pillows, cozy waffle bed blankets and beautifully crafted comforters all designed to work together for comfort and temperature regulation. In fact, my favorite thing to do, especially on Sundays, is to clean my house, change out my sheets into my Bolin Branch sheets and slip into bed with a book. Do you know that Facebook group that was like, I like flipping over the pillow to get to the cold side? It is that feeling. My Bolin Branch sheets are soft and comfortable but also crisp and refreshing. And in the morning when I'm struggling to get out because I'm so cozy, I just throw the Bolin Branch waffle blanket around me so I still feel tucked up in bed while I'm making my morning coffee. Upgrade your sleep with Bolin Branch.

8:25Get 15% off your first order plus free shipping at bolinbranch.com slash ICYMI with code ICYMI. That's bolin branch B-O-L-L-A-N-D branch.com slash ICYMI code ICYMI to unlock 15% off exclusion's apply. Every next move starts with being prepared because tomorrow brings opportunity and opportunity brings new risks. Meet your next project prepared with experienced professionals who know your industry. Meet new markets prepared with one of the industry's largest global networks. Meet Change prepared with a dedicated team helping businesses move forward with confidence. Meet tomorrow prepared with Zurich Insurance. Hi, I'm Jack Ravic, your Kramer host of the best one yet. At T-Boy, we're always talking about small decisions that can have an outsized impact. One of the simplest is staying hydrated. Whether I'm traveling, getting a workout in or heading into a busy day,

9:26I keep liquid IV on hand. It's clinically demonstrated to hydrate faster than water alone and contains three times the electrolytes of the leading sports drink, plus eight vitamins and nutrients. It's an easy addition to your daily routine and helps support hydration when you need it most. Shop now at liquidiv.com, liquidiv hydration that goes wherever life takes you. ICYMI 50. Those are the magic words. If you go to slate.com slash ICYMI live right now and use code ICYMI 50, you will get 50% off a year of slate plus. I get it. You were probably tired of me interrupting your listening to tell you about how slate plus members get ad-free listening across all podcasts as well as bonus episodes of our show. And there's only one way to get me to shut up. And that way just so happens to be going to slate.com slash ICYMI plus and using our code ICYMI 50 for 50% off. And if you want to get a taste of just what goes down in these bonus episodes,

10:28then stay tuned to the end of this one for a little treat. All right, that's all from me now back to the episodes. And we're back. And so before we give our thoughts on this news that you may or may not have been tuning into, I would totally understand if you saw the first word of these headlines and I wouldn't like, noped out of there. But I want to walk us through exactly what has been happening these past few weeks. And it started kind of in earnest when an ex-anthropic employee named Jacob Coxon posted a tweet. And he tweeted, I resigned from anthropic today. I spent the last three years doing pre-training research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. And this is part of a whole thread, but sort of the key part that I'll pull out that everyone's reacting to is he wrote, the people building AI earnestly believe that it could kill us all by the end of the decade. Which, if we're doing the math, is in like three years.

11:31Two things just to remember about Jacob as we move forward. He is 27. And according to Business Insider, only joined Anthropic earlier this year. And I also think something that you pointed out at on a previous podcast is this is like one of his first tweets. So lots of things to keep in mind about Jacob. But he tweets that and then a current anthropic employee named Evan Hubinger. Yes, Joe. One of those. Well, sure. He currently works at Anthropic as the alignment science lead. And this the term alignment will come up a lot. Alignment just means making sure AI agents are acting as they're told and in accordance with like human values and attention. Make sure they're making sure they don't go rogue. So that's what this word makes sure they give good responses and do the things you'd expect them to. Right. And so Evan then quote tweets Jacob's tweet and isn't like, oh, no, no, no, this is crazy. He goes Jacob is cracked here. We really do earnestly believe AI could kill all humans. I personally think it is more than 10% chance within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.

12:37And so I will admit when this first dropped, I'm not. It did exactly what it was in trying to do in terms of what it like my emotions like it got dropped in slack. And even if no matter how much I go into stuff like this being like, I'm a day I hate her. I'm a real skeptic of these people and the more we'll talk, you'll see there's so many reasons to take this with like not even a grain of salt or whole rock of it. But it did cause me anxiety for about a day. Even if I didn't realize what I was anxious about, it did hang over me because I mean, quite frankly, I have enough. I feel like I've gone my whole life hearing apocalypse predictions and being told that like I am either at the in a dying industry or you know on a dying planet. I'm like, I do not have room for one more, but I also like don't super understand a lot of the things they're talking about. And so it's really a much easier for me to be like, that sounds scary. I guess, I guess they're right. But us, the public not understanding it, I would say is a crucial reason that it that this narrative is getting as much traction as it is because it just took a few days for me to start seeing what other people were seeing, which is like, all right, some of this stuff isn't adding up.

13:55And so I think I want to talk about the one concrete incident that these people making apocalypse predictions are pointing to, which is something called hugging face. And hugging face is a tech company that has sort of open source sharing of AI software. It's not super important. But what this incident is is open AI employed a bunch of their agents to conduct a series of experiments and in pursuit of these experiments. These agents, quote unquote, went rogue and ended up hacking into hugging face in order to achieve the assignments they were given. And that is what has been extrapolated into these apocalyptic what ifs like if an chap is going rogue and it will stop it. I got to stop you. I got to stop you. They're not going. Yeah, they're not going to rogue. That's actually the best place to start. They're not going. Going rogue, going rogue suggests intention and consciousness and all these other things. They were to be clear. It was like, I think it was 10,000 of them each one of them trying to work out a way in.

14:55To solve a problem. And then when they worked out that they couldn't say they they the models generated responses and plans and basically came to a consensus that they could not solve it through traditional means so they would. Try and find a way to cheat at the test. They found a vulnerability in hugging face and they exploited it like a regular cyber attack. Now when I say they I mean large language models connected to infrastructure, so by the largest companies in the world, not people. They do not act on their own accord or go rogue. They took this they made decisions based on a tree of this basically generating a plan when we'll do this. And then they check with something called a coding harness, which is a program that said, you've got to finish this exploit gym evaluation. And the way to do that is you got to do it right. And these more because these are large language models. They are probabilistic, even though they've got more intricate and they can do more things. They are still an open AI themselves has admitted this. They are mathematically certain to hallucinate, which means.

15:56When they are told they are aligned by which is again an effective altruist term. They are aligned to not cheat, but guess what? They can also ignore that. They regularly do they ignore instructions all the time or in trying to follow something. The interpretation of that is completely different to the one you had, which is, okay, well, I won't cheat, but this isn't really cheating if I. I do this. Now I must also begin with something very simple about this whole thing. You were scared. I'm sorry you were scared. Oh, thank you. What were you scared of? He unknown because they didn't actually name anything. Yeah. So two ways that scam artists scam. Number one is rushing you. If anyone ever tells you you have to do this immediately, you must do this now. We must must must must slow down every single time unless the person is like the building is on fire. There's a person with a gun outside. If it's a vague threat, which you must take immediate action on, even a positive one like you, huge opportunity, slow down.

17:02Classic grifter thing. Two, never feared the unknown. Get specifics. They say they are scared of this stuff, killing you. What will kill me? What are you afraid of? Where? Emily Fallini from Fortune had a really good point about this. He said he was scared of everything and the topic. He didn't say what it was. He said that they didn't have the right safety processes. He didn't say what those were or what could be improved. He said there should be slow down. He didn't say what a slow down. And because we have a I would say a not corrupt, but corrupted media ecosystem. Yeah. Where the largest media outlets in the world do not have object permanence. Do not have the ability to catch a scammer is scamming. We had this big scare campaign, the actually kind of got it out of hand because I think the Jacob Cox and being a 27 year old, which to me is ordering on a child. When he worked to open AI and open AI. So actually, let me take a step back. There are these things called the rationalists. I swear I'm going some with this rationalist effect your altruist.

18:08When they say rational, they do not mean rational thought. They mean they have come to the rational conclusion that we will build super intelligence and it will kill us. They have misled people with that term to think, oh, they mean thinking rationally. So these people want everyone panicky so that because just to be clear, they said we'll do a slow down. That means putting an organization called METR, which is an anthropic back to supported organization. I don't know if anthropics funded them directly, but the same people fund by them and a lot of one of the this guy I think is called Joe Benton who quit anthropic like a day after Jacob Cox and was like, yeah, and I went to or said he left two weeks before and he was like, oh, I'm just saying I've done this now. And I've gone to work at METR. Yeah. So yeah, I've really depend on. Yeah. Exactly. I've got I haven't even had to change my my slack room. But this is the thing. We need to look at anthropic and a chunk of the AI industry is if they are run by a cult because they are the cult of super intelligence.

19:11We have no proof that super intelligence is possible. They talk about recursive self improvement AI that teaches itself. We have no proof that that's possible. We have some evidence that it might be maybe one day, but we really don't have a route to where. And we have no proof about super intelligence, but the rationalists will say, well, because we have no proof because we have no proof, but we have some signs that maybe it's happened, but the hugging face attack that's proof that this is on the way. And because we've made some predictions along the way that vaguely sound like this. Well, that's proof. And because everyone loves connecting dots and we have we are in a society built on a cult like worship of the wealthy and endless credentialism people go. These are the smart people and they work at the organizations. They'd never mislead me. And indeed, maybe they don't think they're misleading. That doesn't mean they're right. And we do not have fundamental evidence to back up any of this. So I think what happened was you have the rationalist types to once the worker and the topic. And just to be clear, Jacob worked open AI where they are just they have some rationalist, but it's mostly a regular company work there for a couple years has equity in the company.

20:20And probably went there and everyone's like, I love my job at computer. You doing any hacking? Yes. Great. But like regular job and he goes to anthropics like day one like, Hey, how's it going? Welcome to the murder factory. We are working on killing everyone. But if we don't do it, the Chinese will do it. So just to be clear, we're all going to have lunch later. We're going to talk about how this is going to kill everyone. Hope are you free? Oh, you're working on killing everyone. Okay. Well, hit me back. Hit me back when you're done on that. And they would claim, Oh, we're aligning it to stop all this happened. Yeah. I can think of a real alignment on the server on the server. Audio part. Bam, bam, bam. Like on the servers and so the long short of it is none of this really means anything. But because we as a society go, well, the smart person said it. A smart person said it. I must believe them. We just left in this mess of world and this mess of a mediary ecosystem.

21:24Yeah. Cause I think once you sort of understand what really happened with the hugging face, sort of controversy, the rest of this does start to fall. Because like you kind of mentioned, it was not going rogue. It was doing pretty much exactly what it was told to do, which was. It was part of sort of a series of experiments they do in which some of these ages were given purposefully impossible problems to solve. And so they are going to use every method possible and open AI. I guess didn't or hugging phase didn't realize that there was they had left these openings and work around for these ages to again do what they were told to do, which is figure out the answer. At all costs. And I know, you know, I listened to what next TV these episode on this with Paul Ford and he really said, this is a human failure. It is not an AI failure. And I think the fear comes from what you pointed out this anthropomorphism of me being like they're going rogue. They're they're literally intentionally going like he and running off. And I think that is another reason why this can feel scary at first blush is like we as humans can't help but do that to them.

22:28But that is like lending a lot more credence like lending a motivation to like a series is like basically a series of math problems and it just doesn't exist. And I want to then if we sort of accept that the sort of the premise of all of this isn't the thing that these doom sayers are claiming it is. Then it sort of gets to okay. So what are they? What is their motivation? Because you know Jacob Cox and I believe in a previous interview you referred to him as a waste of skin, which was I think beautiful. I really stuck with me. And he has been you know we mentioned some of this up top he worked at Anthropical and they for a couple of months. This is like pretty much his first tweet and he is now on this media blitz Jacob Cox and has left anthropic. You've probably seen former anthropic employee Jacob Cox and viral posts on social media. Jacob stepped down from his position. We are joined now by the man who started this debate. Jacob Cox and today we talked to the researcher who got us to this milestone moment. Jacob Cox and and I think something you pointed out is like you you've worked in PR like there is an angle to what he's doing. What do you think that angle is.

23:33Okay, so I'm a whistleblower right I'm I'm blowing the whistle. First of all, you know what a whistleblower does they blow the whistle on some right there's a thing to point to. They go this is wrong. Second of all, they have stuff to show. So think of Francis Hogan who is the brave woman who went to journalist Jeff Horwitz and George Wells, I think of the Wall Street Journal and some others forgive me for not remembering the name and revealed the Facebook files. Not that that like that was a big story, but not that it really led to anything other than a bad sauerkraut. There was some good coverage out of it. It's just that nothing really changed with the companies because we don't regulate tech. Then there was of course Edward Snowden feel however you do about Edward Snowden. He had the documents on prism. First of all, you don't get an exclusive in the Wall Street Journal. That's not how this works. Like that's something I don't do PR anymore. I don't know for myself, I guess. But you don't get an exclusive like the journal doesn't refer to it as exclusive guy leaves company. Yeah.

24:34Like they do in stories like the Facebook files, but they do so with coverage. There's like a whole thing about it. This is like exclusive guy scared. Yeah. And that just immediately maybe go and ask something up here. Then there's the fact that he was on TV so often and this is a bit a bit tisomy of me, but his background. The background was interviews. That's a nice house. That's a nice house. Maybe the venture capitalist office doesn't look like a podcast studio to me. Looks like and if he can afford a house like that, I got more questions for him. Then there's the other thing of he had so many interviews. He learned them up back to back to back to back to back to back. How do you do that? Well, he's learned some. No, there's someone doing this for him. Someone doing this with him. Also, if you look at the people who reposted his post, his one post, he'd never posted on Twitter before and it all came very quickly. Okay. Why were the first people that reposted it like nine minutes after effective altruist types like that prick Daniel who wrote AI 2027.

25:36Why that's weird. But again, people look at this and they go, well, he worked at the company, he talked to people at the company. Okay, but he didn't tell you anything. The funniest response that got to that point, but I was saying, um, never hear of an NDA. Then you're not a whistleblower. Right. Right. If you're sticking to your NDA, you're just following rules. Point of the whistleblower is you break the NDA. You go, hey, I am taking a risk here saying I'm going to do this and I'm going to hope that I don't get sued to ribbons by using the media as a shield and saying this is in the public good. Instead, he's like, I'm so scared of this. I'm so scared. It's in the bottom of the wide interview, by the way. He was like, yeah, I'm probably going to do something like AI 2027 or getting AI safety job. He said he was leaving the AI industry. No, he's not. He's moving down the block again. Yeah. Yeah. Is he even leaving the same office building like right. And the thing is, I am just a bloke. I may have a big newsletter, but I saw this and went, no, that's really obvious.

26:36I see him on the same thing as CBS and Fox. So I'm on Fox being like, actually, yeah, climb it was a hoax. I agree, Mr Trump. Like just like fucking hell. And the thing is, the media is not doing the job here. There are some are like for the most part, the media is just like, hmm, slot me up. Yeah. Oh, I love skit. I love being serious. And people are like, oh, words to get clicks. No, no, it's actually not to get clicks. It's because they believe. And in some cases, I think it's because they want to believe. People want to believe that this is this because if you strip all of this away, paraphrase more wonderful girlfriend on this one who will be list them. If you strip away all of this, Tom Fulery and all of this scary stuff and all of this stuff about super intelligence, what's actually been happening is thousands of AI researchers have been working on cloud software that's really unprofitable, but mostly sends the money to the largest and most profitable companies in the world that actually this is not a software that is democratizing anything or creating value for the many. It is centralizing value in Microsoft Google Oracle Amazon.

27:52It is sending the money right back where it always goes. If it if this is not working on super intelligence, if LLM's are not RSI, if they are not the future. Then everyone has been working to send venture capital dollars to Microsoft Google and Amazon for the most part. That is all this is and its mediocre cloud software at that LLM's have improved. I'll give them that. However, it's cost half a trillion dollars to get there or more at that price. I am underwhelmed at that price. It better kill me. Exactly. No, exactly. It's this like I'd like a best drop in an envelope. Yeah. That's the thing. And if you take away all this pomp and circumstance, if you really remove all of the rhetoric, what have we got here? We have coding models, which for very specific functions are useful and for software engineers who are proficient in can read code. It can speed stuff up. It is good at generating charts. I have used it to fix my son's mind. Yeah.

28:55And that's because you know what? The weather storm doesn't work sometimes. I'm just going to fix it. Point is, even that took an entire day and it broke halfway. Point of making is when you remove all the rhetoric, it's kind of fucking sad. It's we've we've all hyper obsessed over this one thing and all collectively pissed our pants about what this might do. And if it's not super intelligence, it's just software. And if it's just software, we wasted our time, our energy. And most of the media got freaked out by nothing like a chihuahua seeing its reflection. We'll be back with more of my conversation with Ed after this quick break. Support comes from wise, the smart way to manage the currencies you need around the globe, fed up with losing out to hidden fees when you send money abroad with your everyday bank, choose the smart way. Wise, you can count on the exchange rate you'd usually find on Google. No unwelcome surprises. Plus, ditch that wears my money feeling. Most transfers arrive in under 20 seconds.

30:05Join millions, saving billions on hidden fees. Be smart, get wise, download the wise app today. Tease and seize apply. One second, you're enjoying your day off. The next, your phone buzzes. A window has been broken at home, your miles away. In one second, everything can change. That's why ADT security systems can monitor your home 24-7. They're backed by the most company operated monitoring centers in the industry. So you're always supported during an emergency. When every second counts, count on ADT. Visit ADT.com to learn more. It's a certain point. This needs to be this huge, scary thing. Otherwise, what has all the money and attention been about? It's like if it's not this, then people are going to realize we're full of shit. The second thing, and I think this is something that I know our coworker, Nathish Pawel, pointed out to me when this first was surfacing, was just like, we cannot underestimate the bubble that these people who work in AI are living in.

31:15The ways that they basically all have AI psychosis about how great and powerful AI is that is removed from reality. It is such a bubble. We are getting the thoughts of someone who has slept in four days and has been on their phone the entire time. Like you said, we're getting this news from what could be described as a cult. So yeah, it really is. Is it a cultist or is it an asshole? Yeah. Because that's pretty much to if you believe super intelligence is going to happen, you are either so far down the whole. Like, you know what? There are probably three or four people, advanced mathematicians who could really speak with authority about the likelihood and they would probably say no. You've got various vacuous software engineers, scientists, types who were saying like, oh well, with this progress, we can tell it raw from here, which is not really how that works doesn't tell you the future. So you've, I just made this comparison elsewhere. So forgive me, but it's kind of like a Cohen Brothers movie. People are like, oh, this isn't off.

32:18That's because it kind of seemed like one because the Jacob Cox and thing. What's more likely is in my view is the Cox and thing comes out. They didn't expect there and everyone's on the back for Altman was on an interview with Forteunes. Fucking, let's shunt out, I believe it was and was on there and went, yeah, actually, it's ill time to go public right now. And what actually happened was they already delayed their IPO because their economics look like crap. I know I've seen their books. I've seen their financials. I'm the one that published them a week and a half before they, they, the time said that they were delaying it. But now he's like, oh, it'd be ill timed because I got to go public. We can't go public right now because of safety, which is weird because open a eyes trying to raise a 1.5 trillion dollar valuation privately. Anthropic, the same day is the slowdown announced that the Nasdaq was going to be where they do their IPO. No one's actually slowing down. But it really is all of these people reacting and while they're not moving together, they're moving in concerts. So they're they all like AI and they want AI to stick by. But some of them are like, yeah, AI safety. We love that.

33:23Some of them are like, this narrative's awesome for me. I've been an AI Doomer forever. I can really play with this. People are falling for this. The morons are out in session. I'm going to have fun with this. There are people like Jensen Huang, CEO of Nvidia who are weak beforehand and said, AGI is here, which is meant to mean super intelligence by the way. And I feel like I've heard that a bunch. But the point is everyone's jumping on different bits than narrative. Everyone's trying to find their piece and take advantage of it. Very few people are trying to actually speak to the truth, which is no one's slowing down. No one wants to do anything. But everyone, all of the deeply cynical people are all being like, oh, well, here's my version of safety. Actually, I love this. Alex Kant from Palantir was just like, yeah, Daria Amadei, he's a great engineer and he's one of the most ethical people in the world. Two months ago, he was saying that Anthropic steals everyone's IP. No one believes anything. And to the nutritious point, it's right. Yeah. This is when you're surrounded by rich people or when you're surrounded by people who are all insane start-up people who all believe the same thing in acting the same way, you build an insane monoculture.

34:29You can see it in some coverage of tech. You ever see almost famous? Yes. Yeah. You remember the scene where Philip Seymour Hoffman says, don't make friends with the rock stars. You know, and you got it on his face and they're going to tell you everything. You cannot make friends with the rock stars. That is exactly what's happened with like, Hoff, the tech media. They've become friends with AI scientists and they're scientists because think about it like this. Think about any job you've had that kind of sucks when you're trying to impress someone. You're going to be like, yeah, you know, go, got to work and I'll hit a few bars and do a training run, spend like half a million dollars on compute. Maybe it'll get somewhere. I've been working 12 hour days for six years, but I may be I'm worth four million dollars on paper. I mean, it's LLMs. They're kind of useful. I use them for this. No, they're going to go, I am working on the frontier. I am working on the frontier of this. And when you're around people who enable that and they're saying like, wow, yeah, everyone around me, they're the smartest people I've ever met.

35:31You know, I like the smartest guys in the room. Look, Enron, the smartest people in the world. These are the smartest people working on hardest problems at the frontier of AI. That all sounds really powerful. And also means nothing. It means they go to work. Yeah, there's like a bigger note I want to end on and you did talk about this. I just want to make sure I understand it correctly mostly because I know this weekend I'm going to be answering these questions from people in real life. The final key to this hysteria after, you know, the tweet from the quote unquote whistleblower and then the reaction of people who are like AI enthusiasts is then the people who are in charge of these companies like Sam Altman, you know, are saying these things like, oh, yes, we must slow down. Like, please regulate us. Like, is there a reason for saying that like their motivation for that? It sounds like it is because this is a convenient perhaps cover for not actually being as successful or making as much money as they need to be. When there is a theory, I don't know if I buy it that they're going to use this to slow down and not have to pay their $1.3 trillion worth of compute bills.

36:39I think what it is sadly is they know that we don't regulate tech. Oh, so it means nothing to ask for it. Yeah. It means absolutely nothing. What they are trying to do is self-regulate. Self-regulation just means jerking their shit from first principles. It just means doing whatever they want when they want in a way that you'll notice that none of these AI safety things. Bernie, come on, mate. Jesus fucking Christ. Bernie Sanders guy, we got to regulate this, but none of the actual specifics will be clear. Everyone they are having the conversation, kind of like the conversation we've been having where it's like they're saying the right things. But not actually saying them until nothing happens. Hakeem Jeffries is on him. Oh, no worries there. Hakeem Jeffries has said he is going to push through AI regulation. Anything on there about the people killing themselves because of chat GAPT or the murder suicides or the environmental damage.

37:41No, no, no, no, it's about super. We're going to ban super intelligence. We're banning Grinches folks. We're going to shoot the lower acts. The Ninja Turtles will not be allowed in New York. We will be saved. These companies want to suggest they care so that they don't get regulated out of existence knowing they're in the scourge of neoliberalism that we're not going to change anything. So they are doing this so that they get to choose the regulation. If we actually want to regulate this, we could do it tomorrow. It's not real easy. Or you'd have to do real simple. No more mental health stuff. No more conversation. Remove anthropomorphization. You can do natural language, but it has to respond kind of like a computer. No emotion. No medical advice of any kind. I honestly think no legal advice people be like, well, it's useful for that. It hallucinates you more on. So it gets I can spot when it gets it wrong. Can you? Well, human beings make mistakes. Human beings can love from they can actually learn from experience.

38:41LMS don't do that. How about legal copability for these cybercrimes that happened? We have put people in jail for committing cybercrimes before using GPUs. We should die row. I'm a dance. I want to be in handcuffs. I don't know if they should be held held. You need to have a conversation at station. You need to get the FBI. I'm not not being pro copper anything, but we do have laws. And we should at least use them. We have Lena Khan made a great point. We actually have regulation for unsafe products. If a pill kills someone, if a car kills someone, if a toy hurts or kills someone, these things get pulled. We should pause all of this. We should take out of the M's offline completely. Oh, we can't do that because of the economy. Well, that's not my god damn problem. Do you care about safety or do you not? Well, yeah. And what you just said is actually exactly what I wanted to land on, which is that I think another benefit of positioning a eyes main harm as this sort of unknown doomsday distracts from like the very real harm it's doing right now. So you've mentioned that this the mental health issues that people interacting with these LLMS have had, including suicide.

39:49I mean, there's obviously the environmental concerns. There's like the price of utility bills that normal people foot because of these data centers job loss. I mean, I just put this is for some reason the thing erosion of independent thought because you're just relying on these systems. It's like if we make the big bad and apocalypse that no one can actually define or name or point to it means these things seem like small potatoes when actually like those are the material things that are going to affect us and already are. And you also don't have to do anything. It's great for the journalists who mentioned AI safety. They can say well look my reporting led to AI safety as a politician. You can say you care about AI safety. You ban super intelligence. So you slowed down the process of super intelligence. Mark Zuckerberg said yeah, we are now going to dedicate less of our compute trying to reach recursive self improvement and focus more on customers. And it's like well, first of all, I've used Facebook and Instagram. Fuck off. I know that you don't do that. Second of all, wow, how noble of you to focus most of your compute on the thing that exists rather than working on the thing that doesn't good stuff. Mark and everyone's just like well, we best trust Mark Zuckerberg. Why? Because he has money.

40:56There is no rational explanation for that. By the way, other than I love big and rich people. I love big rich people. I want to be their friends or at least I want to be an or or fear of them. Because doing that allows you to exist within the current system. It means that you don't have to reconcile with the this is quote, we got in curses hyper normalization. You don't have to reconcile we're getting a bit big brain here with the complexities of the real world. There are so many things that are bad. But if you want the simple to view if you just go AI safety good, we're prepared for super intelligence now. You get to avoid dealing with all of that. If you take Jacob Cox and seriously, you don't have to deal with the actual harms caused by Daria. I'm at that. This week, the financial times reported a story that was shockingly bad. They said the anthropic has been telling investors that they have 80% gross margins. If you remove the cost of training and slot based compensation. No, I mean, that even the article was just like saying it like just zero source. So a tick-tax through the through the brain there.

42:01That is an unethical way to do business. Yeah. When you normalize this thing, when you take super intelligence seriously, when you take AI seriously, you get to live in a very simple world where everyone buy and video chip make more money. And throw pick open AI using video chip make more money AI make more money for everyone even though in the productivity data isn't present even though we have tons of harms. But you can just ignore all those heart thoughts and just say number go up when I say, well, you can't afford it in the future. They go, well, number go up now. Again, simple, simple, simple because if you don't deal with any of the complex that you can leave this wonderful life with a perfectly smooth brain, even really smart people can do that. They can say, where do I get my information from? All the people who are most likely to have an agenda. Those are my source of information. Other sources. If I listen to those, I have to break from the consensus. And that might be tough. Every time we repeat these cycles is the same thing. We listen to authority.

43:02But not really authority. We kind of let arise hurt. We just like kind of squint and we go, yeah, there are there are an authority. It seems plausible enough. I'll keep living my life and every time this happens every single time this happens regular people get hurt. Regular people lose money and regular people dying. And until we break the cycle until the media actually has responsibility, it will keep happening again and again. Okay, that's the show. Thank you so much to Ed for joining us. We'll be back in your feed on Wednesday. So definitely subscribe. That way you never miss an episode. Leave us a rating and a review and Apple or Spotify and tell your friends about us to see all the visuals, reference in today's episode. You can follow us on Instagram at I see why my underscore pod. And you can always drop us a note at I see why my at slate.com. I see why my is produced by Vickwit Lee Berry and meet Kate Lindsay. Daisy Rosario is our senior supervising producer. Mia Lobelle is late executive producer of podcasts and Hillary Fry is late editor in chief. See you online and not in the AI apocalypse.

44:25Hey there. If you stuck around until after the credits, I'm going to guess you are a fan of our show. So first off, thank you. And second, I want to make an offer. I hope you can't refuse even more. I see why my you've probably heard me say that if you subscribed I see why my you get a bonus episode every other week. Sometimes more as well as ad free listening to all of slate podcasts. But because you're still here. I wanted to give you a taste of what are off the wall. I see why my after dark bonus episodes. I actually sound like. Now we can really talk about the tea. It's like now that all everyone else is gone. Let's get down to it. I know that we've talked about this. I don't know if it's ever come up or needed to come up in an episode. Drop the lore. I'm going to drop the lore. Before the press starts reaching out, I was at the event last night where now Caroline Callaway and the good noticing podcast are in a feud. I'm trying to remember what the fancy grocery stores are there. I mean those whole foods isn't one that I know that I loved one is like head book or something. He cheebie. He just alienated. Full disclosure. I was on an edible at this concert.

45:37Hell on noodle. Yeah. Oh, that's just a little rain. Okay. Can you hear her? Yeah. You're anything to say? This week. Here come to the mic. The name of the AI story they're writing is called. What? So looks like a chicken is involved. The more we try to figure it out, the less fun it is for me. If I'm going to be. You're like, you know what I actually, this is your gym. I like the know. You know what you should be brave enough to say it, even though everyone on Instagram is so mad at me right now. So if you want to hear producer Vick and I pop off in your pod feed, we are offering a special for new slate plus subscribers. If you sign up now using the code ICYMY50 at checkout, you'll get 50% off a year of slate plus that and all the other perks that you won't want to miss out on. Go to slate.com slash ICYMYplus to sign up. Thanks. And I hope to see you in the bonus feed.

46:42The idea is a major driver. Kids are spending up to nine hours a day on screens, often unsupervised and studies show a direct link to anxiety, depression, and even suicidal thoughts. That's where Gab comes in. Gab offers safer phones and watches for kids with no social media tailored to every age, offering the right tech at the right time or tech in steps. GPS tracking enabled watches for younger kids to phones with parent enabled apps for teens each device allows kids to more safely grow their independence. Visit gab.com slash get gab and use code get gab for an exclusive offer. That's g a b b dot com forward slash get gab gab tech in steps independence for them peace of mind for parents. Every next move starts with being prepared because tomorrow brings opportunity and opportunity brings new risks. Meet your next project prepared with experienced professionals who know your industry. Meet new markets prepared with one of the industry's largest global networks. Meet change prepared with a dedicated team helping businesses move forward with confidence. Meet tomorrow prepared with Zurich insurance.