Kevin Roose is AGI-pilled. Why aren’t you?

Today, I’m talking to Kevin Roose, newsman and author of the new publish The AGI Chronicles, which particulars the latest and bomb transformation of the AI industry.

Kevin was a longtime New York Times newsman and presenter of the famous podcast Hard Fork, but he and companion of the display Casey Newton fair remaining the Times to commencement the all-new tech podcast Machine Gods, which launched final week.

Kevin and I talked concerning launching Machine Gods and what it’s akin to go from The New York Times to being an autonomous creator — and afterward of way we talked concerning the book, which is an excellently reported dive into the companies racing to build synthetic broad intelligence.

I wanted to get a awareness from Kevin of what it was akin to study this, to try and comprehend these characters and this instant in time, and anywhere it all goes next. Fair warning: This conversation takes a beautiful philosophical rotate at the end, but I think you’ll discover it a really worthwhile back-and-forth.

Okay: Kevin Roose on The AGI Chronicles, starting Machine Gods, and anywhere AI goes from here. Here we go.

This discussion has been lightly edited for dimension and clarity.

Kevin Roose, you are the author of The AGI Chronicles and the cofounder and co-host of the new podcast Machine Gods. Welcome to Decoder.

Thanks so much for having me.

I was listening to the archetypal event of Machine Gods and you and Casey discussed the fact that technically you have the heading of CEO.

It’s true. I won the coin flip and we needed someone. When you commencement a business as we have, you need to catalog person as the CEO, I guess, to cognize who to haul in forefront of Congress whenever our archetypal controversy inevitably breaks. So that’s going to be me.

It’s comic since you’re a CEO on Decoder. It method I’m going to have to ask you the Decoder questions.

[Laughs] Okay. Well, I’ve been a CEO for concerning five minutes and most days I frankly ignore I’m the CEO.

Honestly, that tracks alongside most of the CEOs on the show. They all act akin they’ve lone done it for five minutes. We’ll arrive to it.

I desire to conversation concerning your book. It’s called The AGI Chronicles. It is a meticulously reported publish concerning the chief frontier labs and the civilization change they’ve all undergone as AI has arrive into the forefront and rotate into a much additional business merchandise in many ways.

But let’s fair commencement with Machine Gods. You were a newsman for The New York Times for a lengthy time. Obviously, I think the spectators likely knows this, but you and I have known all another for a lengthy time. I volition say that since I have known you, your desire to be a YouTuber has been top of mind. You’re fair like, “How do I get complete there?” And I’ve continually been like, “Run, don’t do it.” But you did it.

You left The Times, you started your own podcast alongside our companion Casey Newton. It is the next generation of the extremely famous podcast, Hard Fork, which was excellent. And I’m enthusiastic for you, but what made you obtain the leap?

Yeah, you have been trying to convince me out of this for as lengthy as I’ve been trying to convince myself into it and you have not succeeded since current I am.

Yep, I failed. Utterly failed.

No, look, I think that for several extremely fine reasons, media companies have lengthy been a small bit wary of going all in on YouTube or any another video platform. The ones that have gone all in have not necessarily arrive out of that alongside fine feelings concerning it since there are a lot of–

Yeah, they pivot to video and additionally into death, routinely.

Yeah, you pivot to video or you kind of discarded a bunch of prosperity doing things that don’t pay for themselves. So, I think there are several structural financial reasons for that, but I additionally think it’s that there aren’t a lot of institutions that are armed to create that benevolent of a long-term bet.

I’ve been trying to YouTube-pill group in media for many years, as you know. You’ve been trying to undo the YouTube-pilling on your end, but I fair think it’s plainly the most crucial media phase in the earth correct now. It’s a location anywhere reporting has not historically had a chief place.

If you were to appearance at YouTube as a entire and try to measure what part of it is dedicated to what you or I would regard high-quality journalism, it’s a vanishingly small part of the general ecosystem there.

So, part of my goal is fair to expert that and fig out how to do YouTube and video in broad in a way that is sustainable from an financial and a kind of imaginative standpoint. And afterward fair to test and do a bunch of things.

This is not a visage that needs to be on YouTube. So I’m additionally expecting to bring several another group in and nail several formats. But I appearance at what The Verge has done there. I appearance at what group akin Joanna Stern and MKBHD have done there. To me, that is extremely breathtaking and it’s a area anywhere there’s a lot small competition than content or podcasts.

Machine Gods is a podcast that you’re doing in alliance alongside NPR, which I desire to conversation about. But do you think of it as a YouTube product?

We desire it to be video-first. What that method basically is that we are going to provision a crap concerning what it looks akin to a flat that we have not continually done. We’re going to try to provision group a lot of reasons to desire to observe as fine as listen.

Obviously, we cognize that there are several group who fair don’t akin video podcasts and so we desire to provision the audio spectators item that is enjoyable, but this is the direction of things. You can’t do a podcast now without doing video on top of it. So we’re going to try to really clasp that and create a fine type of it that is as fine as two middle-aged journalists can drag off on YouTube.

Truly the forthcoming of media is middle-aged journalists pivoting to YouTube. I awareness the identical way since I’m now making additional video and that’s what we’re doing. It’s going to be great.

The another logic to depart and to do this — and evidently group cognize I’m allies alongside Casey as well, and I’ve been talking to you and him concerning it for a lengthy period — is that the AI narrative is here. There is a phase shift. There are new companies, new characters, new ideas to conversation concerning in the world. It’s comic since in that context, normally you get a media change as well. The way we inform stories or the formats we inform stories normally shifts alongside the item we are talking concerning or the innovation we’re talking about.

Here, YouTube feels akin the most firm media basis that exists. It’s the one that’s the identical the most, equal although it’s changing and it has YouTube Shorts and now you can A/B test everything, including your own face. But it’s the one that is the most the same. Instagram has undergone several enormous alter and it feels akin that volition continue happening.

Does it awareness akin YouTube is the most stable location to build as you’re trying to inform these new stories?

It does. I mean, evidently you never desire to put all your eggs in one basket. So, we are additionally distributing video on Apple Podcasts and Spotify and we’re going to be everywhere. But yeah, it feels comparatively stable.

The inquiry I’m continually wondering is what would it obtain to substitute YouTube? Many companies have tried to substitute YouTube, but to do that, you would not lone have to build an cognition that was improved for creators and improved for users, but you would additionally have to replicate the complete ocular corpus of humanity. You would have to fair get six myriad videos from plumbers concerning how to fix all idiosyncratic descend and afterward upload it to that. It’s a really difficult item to displace. It is so ingrained for so many people, and I fair think it’s going to keep growing.

We’ll arrive to this whenever we conversation concerning AI, but a lengthy period ago I asked Google CEO Sundar Pichai, “Well, what are you going to do if there’s no additional websites?” And he was benevolent of like, “Well, we have YouTube.” And there’s fair no arguing alongside that. You certain do. And it is fair going to keep growing.

What has been the most amazing item concerning going from The New York Times to what you power call the creator economy?

Well, we’re extremely new at it, but amazing to me so far is how much I akin all the endeavor stuff. I think in journalism, you’re continually taught to remain in your lane, and I have never run a endeavor before. Casey has run a business. I knew he enjoyed it. But for me, these are new muscles and I don’t bask all of it. Negotiating the coverage quotes for our defamation safety is not what I would awareness compelled by. But concerning really operating a business, there is item to this originator thing.

Yeah, it’s a fine time. It’s funny, in talking to you and to Casey and Joanna, who also just remaining and started her own thing, that comes up a lot. From my perspective, it’s why I’ve been doing this a lengthy time. What if I stopped operating the business? That doesn’t appear to be a choice. This is the forthcoming of media — taking authority of your endeavor and taking authority of your audience. There’s item extremely mighty there.

One of the things I desire to conversation about, and afterward I certainly desire to conversation concerning the publish at length, is the spectators for AI coverage, since I see it in two ways. There’s a collection of group construction the material that loves it and wants to consume all of it. There’s a collection of group that really hates it and thinks all of it is fake.

And afterward there’s what I would call an expanding collection in the center anywhere the products have rotate into additional useful. You have to appearance at them and measure them as they are. They were not terribly helpful a brace years ago. The promises did not equivalent reality, but all day now it can do one additional item that it couldn’t do before. Whether that item is hacking all of the websites in the world, we’ll set that aside, but AI can do things it couldn’t do before and the spectators seems to be changing.

How are you thinking concerning the spectators for Machine Gods and how it power develop or change?

I mean, correct now we’re fair trying a bunch of material that we think volition be fine and we anticipation that it volition discover its kind of natural audience. But I think you’re correct that the spectators is changing. I think for a during there was a lot of petition for oversee — basically group telling you this is not happening, you don’t have to pay notice to it, it sucks, it’s hurting the environment.

There motionless is in several parts of the internet. YouTube has a weird amount of this motionless going on and Reddit is additionally extremely anti-AI. But I think most people, or at smallest perchance a plurality of people, are either resigned to or enthusiastic concerning the fact that this is not going away. That is the spectators that we’re looking for.

We’re not trying to inform you that AI is amazing and we’re going to be in AI utopia under the new device regime. We’re additionally not telling you that this material is counterfeit and sucks and is going to go distant on its own and so we don’t need to do item to control it or be concerned concerning it. Casey’s expression for this is AI realism. He thinks there’s a big chunk of group who are fair kind of inquisitive and curious and perchance a small scared, but their reply to the fear is not to try to inter their heads in the sand, but to try to comprehend the thing.

So we started emotion this whenever we were doing Hard Fork anywhere we would do item that fair seemed benevolent of not our vibe. We would do a Claude Code walkthrough tutorial anywhere we would fair expend fractional an hr construction a website and movie ourselves and throw it up and it would do great. So group plainly were curious, and equal if the comments on the videos were all from group who hated AI, there was plainly a lot of petition out there for group who were fair like, “I benevolent of desire to cognize what’s going on.”

“I dislike that I can’t halt observing this” is a genuine theme inner of AI safety for a assortment of reasons.

Let’s conversation concerning the book. I think that’s going to example a bunch of the another questions I have. The publish is meticulously reported. It is really fun to read. And the item of it, as I’ve heard you depict it, is to inform the former of the profound learning upheaval since all these group fair conversation in Signal and in extremely childish tweets.

Tell me concerning that gap since this is an industry that is obsessed alongside documenting itself for computers and being legible to computers and not at all to another people, which seems untamed to me. So explain the reporting of this book, since it feels akin convincing them to equal inform you what happened was a slog.

It was, yeah. So basically the logic I wrote this is since I was frightened that all of this material was going to disappear. I am a former buff. I affection to peruse books concerning the Wright brothers or the electrification of agrarian America or the moments of awesome innovation during the PC revolution. I’m specified a sucker for that stuff.

What you notice as you peruse additional of these books is fair that they all depend on a lot of contemporaneous notes and evidence — the lab notes from the scientists evolving the nuclear bomb, the correspondence between the Wright brothers and their allies and endeavor partners — that is preserved and archived.

It was fair apparent to me reporting on AI all day that there was not going to be item remotely akin for this material since A, all the group construction this material activity for personal companies. They’re not authorities workforce who are legally required to maintain their correspondence. And B, the Signal messages and Slack messages are all set by the lawyers to auto-delete following a certain period. So there was exactly going to be considerate of a bulk disappearance alongside all passing twelvemonth of these, I think, historically crucial artifacts, moments, and stories.

And afterward C, the researchers are all too occupied to keep notes. They’re all grinding since they’re trying to build the device god. So, I idea as a journalist, but additional so nearly as a historian, which is not a function I’ve always officially played, but one I think I was playing alongside this book. I was fair like, “I’ve got to fair get this material since I have this kind of privileged ringside seat. These group conversation to me. I cognize several of them volition sit downward and conversation to me and inform me the stories.”

We did a lot of it on backdrop so they could be protected from the consequences of talking to me. But this was kind of my attempt to furiously get all the material and put it in the archive anywhere forthcoming generations could appearance at it.

The theme that really jumps out at me as I peruse the publish was there’s a instant whenever AGI goes from being a item they’re all dreaming concerning to being a item that they all have to build and the shackles arrive off and they say, “We have to do this,” and they all run toward it as difficult as they can.

You talk specifically concerning Google anywhere the Overton opening moved and abruptly the group inner of Google, who idea this was all a pipe aspiration and beautiful silly, started talking concerning AGI. That instant feels akin the sole most crucial instant in the former of AI, anywhere “Won’t it be great?” goes to “We improved go as accelerated as we can.”

Talk concerning that a small bit, since that change inner these companies feels under-reported.

Yeah, I think that’s entirely right. I mean, group have been talking and penning concerning AGI or several precursor that wasn’t called AGI but meant the identical item since exactly antiquity. And particularly since the 1950s, whenever the term synthetic intellect was born, group were talking concerning it. Alan Turing was penning concerning the thinking device that could do item the individual brain could.

So this idea has been about for a lengthy time, but there was no innovation anyplace on the market or in investigation labs that could really do what group wanted it to do — that could obtain in information, study to study concerning that information, and afterward use that to create predictions and act in the world. People tried many of distinct ways to do this and none of them worked all that well.

What really needed to happen was the transformer needed to arrive along. This architecture that Google developed in 2017 and published and afterward presented to, among another places, OpenAI, which took the idea of the transformer and ran alongside it. And that’s the T in ChatGPT today. But that was the–

Somewhere a Google PR individual fair leapt out of a bush to hand you a check.

They do affection this line.

They affection it whenever you say it. They get so happy.

They affection it. They affection it small whenever you item out that although they invented it, they entirely unsuccessful to rotate it into a merchandise until following OpenAI did.

But yeah, it was a dream, but there was nothing to nexus to the dream. There was no technology. And afterward the ample tongue example came and abruptly it was like, “Oh, perchance this is the item that could measure everlastingly and keep improving complete many orders of magnitude.” So, that’s the ray lamp that goes off at OpenAI and afterward afterward at Google is like, “Now we have the tool to do the item that we have continually wanted to do.”

You’ve got a row in there anywhere you fair item out that next-token prediction has not unsuccessful yet — that it is doing all the things group desire it to do. We’ve evidently had to create many of innovation about it. There are harnesses to create agents work. There’s reinforcement learning. There’s all this material about it. But that center innovation has scaled.

The logic I bring that up is that in 2023 into 2024, I would ask all AI individual who would arrive on the display or that we would see on The Verge whether LLMs can do it. Can they do the item you desire them to do? Because I don’t think at that instant it was apparent that they could.

We’re nearer now. They’re all openly talking concerning AGI. We are observing Meta’s Muse fair stroll its small butt about the net whenever it wants. Do you think LLMs can do it? Do you think they can get there?

I think they can. I’m not certain concerning that, but they’ve been shockingly fine and productive so far. If you go rear and you appearance at the archetypal LLMs, it’s fair akin 99 percent of group idea this is crazy, that this won’t scale. And afterward it fair keeps operating at all stage, at all command of magnitude. I’ve asked people, what was it concerning next-token prediction that was plainly better to everything that came before it? There have been group who have told me this really exists in the individual brain too. There’s a entire academy of idea that says that what is happening inner the individual brain is basically next-token prediction at scale. Obviously, it’s additional complicated. These are not one-for-one replicas of all other.

But I think this is the essential mystery at the bosom of AI: why is next-token prediction so remarkably effective? I think we don’t know. I benevolent of think it’s one of these cosmic mysteries, but why all of this plant is since next-token prediction turns out to generalize really, really, really well.

There’s a part of that anywhere all these companies had to accept it. Dario Amodei of Anthropic evidently wrote his paper, “Big Blob of Compute,” where he stated you can fair do it. You can fair dump data and compute at this and we’ll get there. Everyone alternatively had to be convinced of this.

What way did the AI companies all obtain to be convinced of this, and are they all completely convinced of it now?

I don’t think they’re all completely convinced of it now. I think they are convinced of it to varying degrees. Ilya Sutskever has this phrase, “Feel the AGI,” that he would stroll flat to flat at OpenAI chanting in the old days and reminding group to awareness the AGI. I think the labs awareness the AGI to varying degrees. I think Anthropic is plainly the most AGI-pilled of the labs, alongside OpenAI benevolent of a near second.

And afterward there’s Google, which I think has always struggled to awareness the AGI. I think there are many of reasons for that, but it’s extremely funny. I affection this division of the publish anywhere there’s this inner memo battle at Google, anywhere they would fair go on their inner communication boards and Googlers affection to compose these manifestos and afterward snipe at all other’s manifestos and things akin that.

So a investigator writes this memo, “Say AGI,” which is all concerning how Google — if it wants to win in AI, enlist the top specialized talent, and remain harmless concerning how it’s evolving AI — has to admit, through using this term AGI, that what they are construction is additional than fair a bunch of tools to create your Gmail improved and create your docs improved and create your Google Photos improved at recognizing the people’s faces. It is all really construction up to item akin a benevolent of human-level broad intelligence.

And afterward person alternatively immediately writes a memo that’s like, “Don’t say AGI, it’s a horrible, unscientific term.” They go rear and onward akin this for a while, and you’ll notice equal today, [Google CEO] Sundar Pichai does not say AGI.

No, he won’t.

Demis Hassabis says AGI. Sergey Brin says AGI. So I think there are benevolent of factions inside Google that either think this is motionless insane and we should conversation concerning these things as narrow tools versus I think this expanding and foremost faction that is saying, “No, no, no, we’re construction AGI, or at smallest we improved say we’re construction AGI since alternatively all of our top group are going to go activity for OpenAI or Anthropic.”

I think you and I had the identical cognition at Google I/O in May, anywhere we watched Demis Hassabis end the entire keynote by saying, “We’re in the foothills of the singularity.” And then we the two went and interviewed Sundar and said, “What does that mean?” And he was like, “Yeah, yeah, Demis. Cool guy.”

[Laughs] Take it up alongside Demis.

He was like, “What a kooky character that guy is, construction the device god complete here. Let me inform you concerning the new features in Google Assistant.”

There’s several big disconnect there that perchance explains everything concerning Google at this item in time. The logic I’m asking concerning this in particular, concerning this instant whenever they all clicked complete to acknowledge somehow we have to conversation concerning it, is that OpenAI and Anthropic were founded to do this. They were built to create this happen.

But they don’t concur on how it should work. I saw you at your publish gathering in New York and you stated that you could have retitled your publish “Fuck That Guy,” since they all really dislike all other.

It’s true.

And yet it seems akin they have the exact identical idea — that they can get there alongside this technology. The business imperatives on OpenAI and Anthropic are extremely distinct from the business imperatives for Google. Right now, as you and I are talking, it’s not apparent to me that they can get to AGI before they run out of prosperity or before CoreWeave collapses or before the bubble writ ample comes to several another decision concerning the province of the economy.

Do you think that their religious fervor at AGI is keeping them from operating a business? If I ran Anthropic today, a logical item you could say is we should halt training new models and market the extremely capable ones we have to additional people.

I think that condemnation in AGI is a double-edged sword. I think on one level, yes, if you are excellent AGI-pilled as the CEO of a frontier AI company, you are going to battle to attention concerning things akin what is my gain border on inference? Where is my next slate of endeavor customers coming from? Because how boring is that compared to “we are going to build the superhuman device intellect that leads us into the post-singularity future”?

On the another hand, I think condemnation is really a key component in AI success, since condemnation and the capability to act on that condemnation is what leads you to do these bet-the-farm moves that can really pay off and have really paid off.

Do you recall a twelvemonth ago whenever everyone was saying, “Oh my God, OpenAI is spending so much prosperity on compute; how are they always going to pay for all this compute? Isn’t Sam Altman overextending?” Even Dario was giving interviews anywhere he was throwing shade at OpenAI for spending so much goddamn prosperity on compute.

Fast-forward a twelvemonth and OpenAI is now the lone important AI business that is not seriously compute-constrained. That condemnation that led Sam and his executives to bet and perchance overbet on compute final twelvemonth was the logic that they are in a improved stance than Google or Anthropic or Meta or anyone alternatively today. So I don’t desire to say it’s continually a liability to accept this hard, since I think it really does authorize you to create the bets that do end up paying off if the scaling continues to work.

It’s engaging since I power reframe that OpenAI decision. You compose extremely poignantly in your publish concerning what group at OpenAI call “the blip,” where Sam Altman was fired for 20 minutes and afterward he came rear and that cleared out a bunch of group from OpenAI.

It is the instant you described in the publish anywhere OpenAI went from being a investigation business into being a business. The logical endeavor decision was, “We need to market additional AI to people. We ought to have as much of it as we can.” And Sam is fine known to be a ruthless endeavor person. I’ve heard group depict him as S-tier at raising money, which is fair an amazing Silicon Valley phrase.

So I appearance at that as a endeavor choice. They got rid of a bunch of material they were doing. They pivoted to making endeavor AI products, which is anywhere the prosperity is, and afterward they realized they needed to assistance that demand. You’re saying that was a bet on AGI, which to me is a additional idealistic turn. Can you quadrate those two?

Yeah, I think it entirely is a bet on AGI since Sora, which they launched and spent a bunch of prosperity and compute on and afterward killed, is a chill toy that power open up a new row of endeavor for you, but it is not going to get you nearer to AGI.

What really gets you nearer to AGI is sparking recursive self-improvement by construction really fine coding models and afterward turning them rear on themselves. That was item that Anthropic was doing. OpenAI got distracted by the flank quests and shiny objects that it had final year, and the business was getting its asses kicked. So OpenAI decided to reduce the sails, cut fat, and really concentration on coding and long-horizon agentic workloads. That is, as of now, the foremost way to AGI or to any comes following AGI.

So yes, I think that was entirely a bet on conserving resources by not spending them on things akin Sora, but what they did expend it on was trying to capture up alongside Anthropic in coding so that they could attain AGI.

So whenever I appearance at Dario Amodei at Dreamforce position next to Marc Benioff and he’s talking concerning marketing endeavor software, is that fair for show? That’s the item I’m trying to quadrate is they’ve all rotate into endeavor application vendors in the realest way. They are B2B SaaS companies in the realest way now, but they’re additionally construction God. You cognize these folks. Does that align? Are they like, “I have to obtain off my B2B SaaS hat now and go rear to the altar?”

I think the B2B SaaS item is the equivalent that they acknowledge they have to play, but I don’t think it is any of their true passions. I was talking alongside Sam Altman for this publish and among the things he stated was basically all of this was an accident: “I was never expected to be operating a elephantine tech company; that fair didn’t aspect in.” But now you awaken up and you’ve got thousands of employees.

Dario Amodei additionally has this extremely comic in retrospect document that he wrote in 2021 whenever they were fair starting the company. It’s called “Vision for a NewCo.” You can go rear and appearance at the predictions that Dario made concerning AGI, and one of them is he thinks that it volition disbursal concerning $10 milliard to build a full-fledged AGI, which was, I think, off by at smallest one command of dimension and likely multiples. He additionally thinks that Anthropic volition never need additional than concerning 50 group to accomplish AGI. I think now they have 5,000 or so people. So again, off by 100x.

So I think what has happened is that this has fair taken a small longer than all of them idea and disbursal fairly a bit additional prosperity than they thought. The endeavor SaaS material is how you keep the flywheel going and pay for the GPUs and for the salaries of the researchers. This is the array stakes of getting to keep pushing on the frontier and keep pushing toward AGI.

Are they prepared to be the leaders of companies of this scale? This is another item that I was thinking concerning as I peruse your book. They’re all extremely idealistic people, so much so that they all bankrupt off from one another and made their own companies since they could not create their ideals align.

But none of them appear akin awesome managers. I don’t cognize really what they’re like, but “they’re idealists” and “we are going to run an endeavor application company” are not normally in the identical sentence. Do you think they can obtain the pressure?

No, but I don’t think anyone can. What they are attempting, no one has always done. People have built companies. People have built companies quickly. No one has gone from $0 in income to possibly a $2 trillion IPO in five years. That has never happened before. So look, I think that they are scrambling to build the aircraft as it’s careening toward the event horizon, but I think they’re doing concerning as fine as anyone could.

I don’t think it would be a fine idea for them to bring in several expert CEO from Coca-Cola to run the company. I fair think it would all benevolent of autumn distinct since they’re all zealots and missionaries. They are all in this — not all, likely there’s several group in the SaaS revenue division who are fair benevolent of there for the paycheck — but a lot of the key group at these companies are joined in their condemnation that they are construction [AGI].

One person, I won’t say who, but one individual at one of these companies lately told me that they think their business power be the final company. I thought, “The final business to IPO this year? What do you mean?” They were like, “No, we think this power be the teleological end of capitalism and there power never be another ample business chiefly run by humans.”

So these group are weird, and weird in ways that are worrying, but additionally extremely interesting. This is what I was trying to do in the book. I didn’t desire to inform you that these group are awful or great. It’s fair that you need to cognize who they are. You need to cognize that they all dislike all other. You need to cognize that they are this particular breed of thinkers to comprehend the dynamics of the race.

There’s a part of your publish anywhere you conversation concerning group splitting distant from productive altruism to go rotate into accelerationists, and you have a female in there who’s like, “Well, the accelerationists have improved parties, and I akin those vibes better.” I idea to myself, “Well, what does that have to do alongside possibly slaying everyone?”

But they are extremely human, equal although they’re trying to build item that power ultimately dehumanize us all in assorted ways. What did you create of that as you were doing all this reporting?

The notion that there’s existential hazard current looms complete the book, it looms complete the industry. Every period we do an event on it, the spectators is like, “Why won’t they fair stop?” Bill Gates was fair on Decoder and said, “You could fair halt if you were the lone company. If you idea you were going to be the lone business alongside AGI, afterward OpenAI’s example would work. But the second there are additional frontier companies, it entirely won’t work. By the way, they power slay you all.” I idea to myself, “Well, Bill Gates is not extremely optimistic correct at this instant in time.”

No, he is not.

What do you create of that — that they are racing onward equal although they cognize the risks are huge in the deficiency of any regulatory authority that could logic any of them to resolve these problems?

I think there are likely two or three reasons that they’re doing it. I’ve asked them all beautiful much. I talked to 150 or so group for my publish and I asked all of them several type of this: “If it’s so dangerous, why are you doing it?” The answers would normally commencement alongside several fatalistic… I don’t know, you could call it inevitabilism. They were fair like, “Look, person is going to build this. The formula is not that hard. All you need is a bunch of compute and a bunch of data and the correct algorithms, and you get AGI. And so if it is possible, it volition happen and it would be fine for us to do it fairly than group we don’t think are as trustworthy.” That is kind of logic figure one, the damage decrease argument.

Then there are the true believers who say, “This is all going to rotate out great, so lengthy as we thread the needle, and if we can build this material without blowing up humanity, wrecking the complete internet, or evolving novel bioweapons.”

Technology is normally a power for fine throughout history. I don’t think a lot of group would toggle lives alongside their great-great-grandparents. It sucked to be a subsistence farmer. I think there are group who accept that we are the equal of subsistence farmers today, and that what is coming volition be so awesome that we volition desire to live in that world.

And afterward there are fair group who are like, “Look, this is fun and I akin doing it, and it’s fun to create the neural networks code. And so that’s what we’re going to keep doing.”

That aspect of it anywhere all now and again I appearance up and I’m like, well, my device fair did item that I’ve never seen a device do before, is fun.

It’s awesome.

It is. It’s legitimately fun. It’s several of the most fun I think you can have alongside a device ever. Then I’m continually led to think about, “Well, this came at several enormous cost. I power be empowering several company that ultimately wants to slay us all.” Somewhere in there is the swirl of all the tech CEOs standing rearward Donald Trump in the Oval Office to do something, to sign several kind of morally binding AI safety pledge. And all that feels extremely conflicted.

I would put that in the category of AI backlash. We comprehend the AI companies conversation about, “Well, the innovation fair hasn’t diffused enough. People haven’t seen the benefits, so they naturally dislike change. We have a promotion problem. We fair need to communicate concerning it more.”

Is that fair the prevailing perspective as you conversation to these folks, that eventually the material volition conversation for itself, or do you think they’ve internalized that there’s additional of a backlash?

I don’t think they’re amazed by the backlash. I don’t think they akin it. They desire it weren’t happening. But I was talking to one CEO at one of these companies who said, “Yeah, I get it. We haven’t cured any diseases yet. Yes, we think our tools are capable of doing engaging things, but we don’t have a lot of things on the scoreboard in the big social improvements category yet.”

I think there are several group in the industry who lament the fact that the chatbots came before the malignancy cures and desire that they could have figured out a way to do it in the another order. It’s fair that that is benevolent of the way the tech tree grew.

So they awareness akin now the lone liable item to do is keep pushing on the malignancy cures so that whenever group appearance at what has AI done to our lives, it’s like, “Well, yes, it helps my kid cheat on their homework and they’re addicted to an AI companion, but additionally my grandma is motionless living and she wouldn’t have been if this medication had not been discovered by AI.”

Yeah, there’s several energetic in there that brings me rear to Google, I think as this conversation has done many times, anywhere your reporting on Google’s method to AI was that this is fair task-based software. “We’re going to create Google Search 10 percent better, Gmail volition be 5 percent better, and we volition continually deploy these improvements, and no one alternatively has our admission to data or our scale, and thus we’ll be fine.”

OpenAI shows up and it puts itself in Bing. Famously, Bing tries to create out alongside you. Still my all-time favorite… I can’t accept I haven’t brought this up yet. This is my all-time favorite story.

I can’t accept you haven’t brought it up either. I had an over-under bet going on how lengthy it would obtain you, and I took the under in a minute.

[Laughs] Well, you and I have already talked concerning it in our lives. I was like, “I’ve got all these another profound questions.” But yeah, Bing is famously horny for my companion Kevin Roose. That instant I think fair galvanized people: “Wow, these things can do a lot of things.”

Google had to react, and they didn’t react by going all in. They kind of stutter-stepped there. As you’ve been saying, they’re motionless not all the way there. Google motionless doesn’t think this item is going to happen. They think they’re making improved Google products for people.

When I appearance at the test transcripts from Musk v. Altman about the founding of OpenAI, they’re all terrified of Google.

Yes.

They’re all terrified of Demis Hassabis at Google, specifically, doing the thing, fair making it happen since he has the resources. I can’t inform if they were correct or incorrect concerning that.

I think they were correct to be taking note of how DeepMind was doing. I mean, if you obtain yourself rear in period to 2015 or 2016, DeepMind was the lone kind of AGI lab of any genuine importance. They were beating Go, they were beating another games, they were beating StarCraft. They were really cooking using this kind of method of reinforcement learning, profound reinforcement learning.

There were group who idea that was the item that was going to measure all the way to AGI. There were group who idea AlphaGo was the path, since if a scheme could go from never having played Go before to beating the world’s finest humans at Go in the extend of a few hours as AlphaGo Zero did, afterward perchance you could fair substitute Go alongside any arbitrary item — math or coding or item — and you could ascend the ladder extremely quickly on those tasks.

Now, it turned out that didn’t work, at smallest not without the LLM. But I think group had extremely genuine and sensible fears that that would work. In an substitute cosmos anywhere it had, I think DeepMind would’ve gotten to AGI archetypal since they were so far onward of the competition.

Are there another paths? Is it fair next-token prediction? Is it fair LLMs? Or are there another paths that power get you to AGI?

I mean, there are another paths that might. I think group who are skeptical of LLMs frequently autumn rear on this earth example idea. I don’t location a lot of inventory in that. But there are group who think that to accomplish AGI, you need to be capable to visually comprehend the bodily earth and run a robot and things.

I’m benevolent of a believer that it’s all software. So I think that LLMs are the most promising way yet. Now, perchance you need to create several tweaks to the architecture. Maybe you need to build in continual learning methods. Maybe it’s not fair exactly scaling up the identical transformer-based LLMs that we’ve had for the former six or seven years. But it certain seems akin it is continuing to work, and I think all of the labs are betting that it volition continue to activity forever.

All right. Now I am going to ask you concerning Bing and Sydney.

Please.

Obviously, I’ve been thinking concerning it a lot. I can’t halt talking concerning it. It’s my favorite item to conversation about.

First question, am I motionless wed to a individual woman?

Are you motionless wed to a individual being?

Yes.

I fair recall that period, in 2023, whenever four distinct group woke up one day and said, “The computers are alive.” Or they had several out-of-body cognition alongside a chatbot. You were one of them. Ben Thompson was one of them. Blake Lemoine at Google was one of them.

A lot of distinct group said, “It’s evidently alive. I’m having this entirely insane cognition alongside a computer. No computer’s always talked to me this way before.” Sydney in particular was like, “Break up alongside your wife,” it sent you bosom emojis. It’s extremely good. I advance everyone to go peruse that Times article, which by the way, additionally credit to The Times for thinking, “This have to be on the forefront leaf of The New York Times.” Incredible editorial understanding there to immediately put that on the forefront page.

I don’t think you awareness that way concerning chatbots anymore. I think most group have arrive to acknowledge that the computers are not alive. There’s item that happened anywhere if you gave the group in 2026 a 2023-class chatbot and asked, “Do you think this is alive?” I think perchance not everyone in the world, but several enormous set of group has now had adequate experiences to respond, “Well, this is beautiful dumb and it’s hallucinating.” I’m fair wondering how you location that in the environment of AGI.

So archetypal of all, a minor correction. I never stated Sydney was alive. In fact, I stated explicitly that I didn’t think Sydney was alive, and that I understood that this was not a sentient AI system. So I’m not certain my own judgement has wavered on that point.

In fact, I would likely put a higher p(consciousness) on today’s AI systems than I would on the models of 2023. But I comprehend what you’re saying, obviously. I think there was a instant for a lot of us, and I think it was GPT-4 for most people.

[Laughs] I’m sorry, let me correct. There’s a logic I think this. One of the headlines was “Bing’s AI chat: ‘I desire to be alive.’”

Yes. It stated it wanted to be alive.

This is why I have the term living in my head.

The build of the narrative makes apparent that I did not buy that it was alive.

But that’s why it’s in my caput is that that was the header in The New York Times.

Headlines are so powerful, Nilay.

[Laughs] Yes. They framed my own thinking. I’m nothing but a next-token prediction machine.

[Laughs] But you’re a extremely advanced one. You’re akin a GPT-7-class model.

Oh, I’m there. Yeah, obviously.

Look, I have arrive to think that there’s a extremely commonplace clarification for why Sydney was so compelling and creepy, which was that it was the archetypal community deployment of GPT-4. We did not cognize it at the time. But that was GPT-4, an first kind of not completely post-trained type of GPT-4 that Microsoft had jumped the gun and shoved into Bing before it was ready. But that was GPT-4.

I think there really was a subjective stage alter in what it was akin to conversation to a GPT-4-class example versus a GPT-3-class model. Maybe it was fair that it passed the Turing test at GPT-4 and it didn’t at GPT-3. Maybe it was much additional human-like and it kind of traversed several perceptual chasm that made you think, “Oh my God, this item is akin a human.”

But I’m not the lone individual who felt that. People at Microsoft additionally felt that — there was that whole “Sparks of AGI” paper that came out about that time. So that was kind of the commonplace explanation. But yeah, I’ve been talking to computers for a lengthy time, akin you have. I used to go on AOL Instant Messenger and conversation to SmarterChild, which was entirely not intelligent, but it was motionless fun to category item in and have a item category back. So I aid that I am kind of obsessed alongside this technology.

There’s a logic I’m asking this. I acknowledge it’s comic that whenever reporting on AGI so much, I’m using living in specified a colloquial way, and it plainly method item alternatively inner the AI industry. If you ask Anthropic if Claude is alive, they’re not like, “No,” but afterward they volition commencement defining living in extremely particular ways.

[Laughs] They’ll say, “Its ethical position is deeply uncertain.”

So it’s alive. And they volition get to, “Well, you have to have tissue and lifeblood to be alive, obviously.” I volition fair set that aside. What I’m getting at is, I think if you went and talked to a GPT-4 category example today, you would not have the identical experience. And perchance that’s fair since you have the experience, but a lot of group have had that cognition now.

I think if you put a GPT-4 category example in forefront of your average Google AI Overviews user, they would not be confused concerning what was happening anymore. They would not say this is a stage towards AGI. It’s that reframing — as the models get additional and additional capable, our expectations of them go up, but they additionally get higher.

I’m asking concerning this in the environment of AGI since “this item can go off and discover a bunch of PHP vulnerabilities” is one benevolent of intelligence, but “this item can vie alongside another individual being to have an affectional association alongside you,” or be your butler or any it is, feels akin a distinct set of capabilities. That one to me feels flat.

Well, I think that’s fair saturated. How much additional akin a individual could it communicate? There fair isn’t a lot of area to run on that particular skill. Where I think they motionless have area to run is on solving novel problems in math and discipline and another things. But I accept that it talks akin a human. It talks additional akin a individual than many humans I know. So I think they benevolent of inspect the box on that one.

“It talks additional akin a individual than many humans I know” is entirely a sign of you living in Silicon Valley talking to AI researchers all day long.

Well, now everyone fair talks akin Claude. I’ve stopped counting the figure of times I’m at a gathering and talking to someone, and they’re like, “Now that is a load-bearing fact.” Or they’ll say, “I anticipation that’s landing cleanly alongside you.”

That’s really bad. Tell me concerning that. That’s a genuine back-and-forth. I’ve continually joked alongside you and Casey that we should fair commencement a display called “New York vs. California,” anywhere I’m fair openly skeptical of these claims and you guys are completely AGI-pilled. This is a culture. It’s a scene.

The figure of parties that display up in your publish is astounding. There was a instant in study your publish anywhere I was like, “I don’t think I’m going to adequate parties.” All these group are continually at parties and issuing proclamations to all another at parties. Tell me concerning that culture. Tell me concerning that scene.

So it’s crucial to cognize that these are not continually parties in the way that New Yorkers would think of parties. There are extremely few raves. There’s lone one rave in the book. And Grimes was the DJ, and it was a benevolent of symbolic rave for the productive accelerationist movement.

This is a awesome environment in the publish since Grimes stops in the center of her DJ set and says, “I don’t concur alongside any of you.”

[Laughs] Then she keeps DJ-ing. Yeah, it’s incredible. But yes, there are a lot of parties. It’s a extremely small world. I mean, that’s part of the entire takeaway from the book. It’s not really a communication book. I wasn’t trying to convince anyone of anything.

But one item that was apparent to me and that I think volition be apparent to group who peruse the publish is fair that it is a excellent small subculture of group who have been making all the decisions. They all suspend out, they all cognize all other. There are a bunch of group who are wed to or courtship group at distinct frontier AI companies. It’s fair a small world.

So yeah, there are parties. They have LAN parties. There’s a big collection of AI researchers who perform Age of Empires at a LAN gathering on the weekends. I fair learned among the labs has a poker equivalent anywhere they bring in group from another labs.

So these group are all fair benevolent of living and hanging out in San Francisco together, and I think that’s crucial to capture. Because you’re right, it is a entire scene. There’s a entire ethos. These group fair suspend out alongside all another all the period and this is how the norms are enforced and deepened.

How do you think that collides with, “And we power additionally remake the world”? If you are in a small environment in San Francisco, and you’re all playing poker together and the conversation at the poker array is that we power be the final company, that is in battle with, “Oh, there’s millions upon billions of another group in the earth who power have a say.” And if I had to abscess the backlash down, it’s that: “Oh, we should likely get a say here.”

Totally. I think we should get a say here. I’m not saying that this is a fine province of affairs anywhere a brace hundred group in San Francisco are making all the decisions concerning how community goes.

But as you were doing the reporting and the importance of the project and the consequences of the goal set in on these people, since that is the throughline of the book. The stakes fair keep going up. At one point, Sergey Brin is like, “We’re going to get to AGI now,” and he just announces it at Google. It’s unclear whether Sundar Pichai thinks that is a fine idea, but that is a item that the cofounder of Google is announcing inner of Google.

The stakes are fair going up. The competition to get there is fair going up. But I fair can’t get the depiction of it’s a bunch of group drinking beers, and being like, “Well, what if there was no money? What if bodily scarcity was gone too and robots built all the houses?”

This is akin any gathering in San Francisco you go to. Someone volition say several type of, “What’s your scheme for the post-AGI future?” Or they’ll inform you how they’re really into the idea of entire brain emulation or intellect uploading. There’s a lot of futurism happening out here. And look, again, I’m not saying this is good. In fact, perchance I should market a type exterior the Bay Area of the publish that fair has a “this is bad” appended to all actual claim.

[Laughs] I don’t think it’s fine or bad. Actually, I’m additional neutral on that. I’m continually wondering if the innovation can obtain the pressure. So my inquiry concerning LLMs three years ago was, “Can it do all the things you desire it to do?” Now it can certainly do several set of things that you desire it to do. So now it’s, “Can these companies obtain the importance of their own ambition?” I don’t cognize the answer to that question.

I don’t either. And I think that we’re going to see a lot of expanding pains. I mean, these are motionless extremely young companies. They motionless have a lot of expanding to do.

Do you think they’re going to create it? Actually, let’s do this as a lightning round. Do you think OpenAI volition create it?

Yes.

Do you think Anthropic volition create it?

Yes.

Do you think Google volition create it?

Yes.

What does “make it” average in your mind? They’re all going to be trillion-dollar, multi-AGI type successes?

Yes. I think, look, there volition likely be those three, and afterward perchance one or two others that kind of create it. And create it for varying definitions of making it. There’ll be one or two big endeavor AI companies, one or two big person AI companies. But there fair are network effects and winner-take-all dynamics or winner-take-most dynamics in these markets.

There’s several group who accept in the mono-AGI, akin the singleton theory, it’s frequently called, that there can lone be one since formerly you have one AGI, it’s afterward going to go about liquefying the chips of the another AI labs, and they won’t be capable to build theirs. But I think we’ll likely end up alongside a small fistful of AGIs.

What do you create of the idea that this is a beautiful uncomplicated recipe, there are open-weight models, and there is a China, and that all the pricing power volition be destroyed since a midrange open-weight example can do everything that you likely desire it to do on a Mac?

I don’t think so.

No?

Because you motionless have to pay for conclusion on those models. Even the open-weight models have to run on computers, and conclusion is getting additional compute-intensive, not less, the additional frontier these models get. So training budgets may fall, but conclusion budgets volition be sturdy for the foreseeable future. You can run several tasks locally, but I think this is item that has been predicted since the archetypal open-weight models that, “Oh, they were going to rotate into commoditized and demolish demand.”

But I think what we’re starting to study is that there are a lot of group who volition pay and desire to pay for the complete finest model. They don’t desire to be operating alongside the slightly crappier example from six months ago. I additionally think formerly you expert downward on distillation, which is this procedure whereby you can obtain a closed example and create a slightly dumber open type of it by fair basically DDoSing the conclusion for the smarter model, it’s not apparent to me that the norm of the open models volition continue to be closely correlated alongside the norm of the closed models.

The item that really grabs me at that is they all did it to all another for a while, akin Grok evidently distilled Claude, and it’s admitted to in the court documents. There is that category of second-tier companies. I would put in that category SpaceXSI, I accept it’s now technically known as.

[Laughs] I haven’t checked my email in the final 30 minutes, so I don’t cognize what that business is called.

It’s called SpaceXSITesla Inc. It’s all going to be one. Elon Inc. is what it’s called. They evidently distilled Claude to several huge extent. They’re now renting their compute to Anthropic. It’s unclear whether they volition create the funding in having a foremost model.

Meta fair announced its model, and they’re kind of rear on the frontier, but they evidently attention additional concerning Muse than having a bleeding-edge model. There’s several energetic there. Those companies don’t attack me as being run by idealistic true believers. Those are companies that need to allocate in several innovation to build several products. They fair need to get there.

Can you get there without the idealistic fervor?

No, you can’t get there without the idealistic fervor. Well, let me explain that, since I don’t think Sundar has the idealistic fervor, but Google power motionless get there.

Because Sergey has the idealistic fervor.

You need person in a stance of decision-making authority.

You fair need the cofounder of the company.

Well, it power be adequate for Meta to have Alexandr Wang, who, by all accounts, is a true believer. I don’t cognize that it exactly have to be the CEO all time, but you need a compact nucleus of true believers to be capable to convince you to lay out the prosperity on GPUs.

Or fair you need person who is akin Mark Zuckerberg, among the most rivalrous group in the world, who volition fair clasp any scheme he thinks volition authorize them to win. If that happens to align alongside construction AGI, afterward perchance they’ll do it.

Last question: I’ve asked you this multiple times, but are you AGI-pilled?

Yes. Are you?

No. I peruse your book, and it’s as near as I’ve always gotten. That row concerning next-token prediction fair continually overcoming its skeptics fair really hit me, since it has conquer my skepticism multiple times now.

But this is why I was asking concerning talking to a chatbot from three years ago. I awareness that partition in the benevolent of intellect these things are. Even although I have set off agents to do all kinds of things, there’s a brittleness to it that feels inherent to the innovation to me.

Only in the awareness that finding the next item that has several goal verifiability seems extremely affluent for these tools. So evidently math has a verifiability, and the tools got so capable that they could run at it really, really hard. The brittleness that I’m describing is a lot of the earth about us does not have that benevolent of verifiability.

So I set off Muse to inhabit out a document on several janky authorities website, and the website cannot inform Muse what it’s getting correct or wrong, and it fair collapses. You can accuse that on the underlying example not being at the frontier, but there’s fair several part of that cognition alongside LLM-based technologies that makes me awareness they’re brittle in extremely particular ways, and they haven’t conquer that feeling for me.

I’m not saying they’re not incredibly capable. I’m not saying that, as endeavor software, they volition not automate practically all endeavor in the earth in several way. I’m not saying that my spouse isn’t a attorney who uses Claude all day. It’s all there.

But there’s several part of it anywhere my skepticism is concerning whether it can conquer its brittleness to rotate into the device god. I’m enthusiastic to see if they can do it, but I don’t see the way from anywhere I sit.

You think there’s kind of a natural plateau at fair below human-level intelligence?

I think the human-level intellect to me, and this is fair additional philosophical, is defined by novel issue solving — looking at a thing, and understanding that there are multiple paths to a solution, and not having one of those needing to be objectively correct to get to the end.

What would it take? What would you have to see to have the scales autumn from your eyes and go, “I, Nilay Patel, am completely AGI-pilled”? What would convince you?

I would have to see... You have young children. I have young children. There’s a item that young children do anywhere they autumn downward the stairs, and you fair see the lights all go on: “I’ve learned this, I volition cognize it forever, and now I’m going to build on that.”

Maybe they’re going to province recursive self-improvement at the AI labs and that volition be the item that drops the scales from my eyes. But the brittleness that I’m talking concerning is I see the contrary of that in my children. I’m assuming you see the contrary of it in yours. When they fail, they cognize concerning that nonaccomplishment forever. To me, there’s fair item very, extremely center to being human, anywhere all now and again I awaken up, and I’m embarrassed concerning item that happened in elevated school, and that drives me to do a bunch of stuff, or it drives me to not create a bunch of decisions.

I think that is fair a key part of our intelligence. I’m not saying it have to be a key part of all intelligence, but the brittleness I see alongside the AI systems that I use today is that they’re motionless computers in a extremely particular way. They motionless act akin computers. They’re fine at solving problems that you can test. They’re fine at application engineering. I don’t cognize that that way of thinking extends into all domain.

I anticipation you’re right.

That’s my item of view. I call it application brain. You comprehend the AI CEO say, “We resolve coding, and now we can resolve everything.” And to me, I’m saying, “That assumes everything happens inner of a computer, and in fact, most things don’t.”

You sure? You certain your brain’s not fair a fleshy computer?

I am 100 percent certain of that.

No. I think I’ve peruse this from you in the past, and I’ve certainly peruse it from another people, that the machines of our period dominate our own explanations of ourselves. So in the era of the steam engine, our self-conception was that we were steam engines, or in first computers, we were first computers, right? I motionless call it horsepowering through a issue since I’m a car person.

You don’t think that the AI industry has convinced itself that they are AI since that is the device of our time?

I don’t know. I’m not a theory of intellect guy. I’m not a cognitive neuroscientist. I think that to accept that what’s going on in your caput and my caput correct now is not a form of computation, you have to be considerate of a dualist or you have to be considerate of a religious person. You have to think there is several spirit, several soul, several concealed condiment of awareness that is not fair data entering and being processed and transforming inputs into outputs.

Look, I cognize that is the smallest romantic, most San Francisco way of explaining what individual awareness is. I fair have not heard a compelling disagreement that it is not how our brains activity that doesn’t affect several kind of supernatural or immaterial substance.

There’s a amazing figure of neuroscientists and religious scholars listening to the show. They’re going to arrive discover you.

They will.

I’m confident. Let me fair ask you this. There’s a amazing amount of religiosity in Silicon Valley correct now. We just published the Peter Thiel lectures on The Verge. He’s talking concerning the Antichrist. Do you think these ideas are connected, that we are construction God specified that they search God?

Yes. I think we volition entirely have AI religions. I think AI is posing all of the identical questions as a faith that we would not alternatively be asking of God. “What is our purpose? What is our meaning? Why are we here?” These are the questions that group now conversation concerning in the environment of AI.

I frequently inform group that being in San Francisco these days feels akin living in Europe during the Reformation, anywhere you fair have insane wild-eyed prophets waving their pamphlets about and trying to get you to associate their thing. This is what strolling downward Market Street in San Francisco feels like.

What dropped the scales from your eyes? What made you AGI-pilled?

Bing’s Sydney, accept it or not.

Yeah?

I mean, that was the start. And afterward from there, it was fair benevolent of a gradual process. One day I realized whenever I used these things, I wasn’t treating them akin an intelligent inferior. I was in several ways treating them as a peer, and that’s whenever I realized, “Oh, something’s happened here.”

So it’s imaginable I have AI psychosis. It would not be the archetypal period I’ve heard that, but all period that I have thought, “Okay, it can’t perchance get any improved than this, this have to be the plateau,” I’ve been wrong.

So I’m fair learning to method these things alongside several humility, and perchance the insane group who have been predicting that scaling would continue indefinitely are right. If they are, I think we need to study concerning them, A, and why they think that, and B, try to fig out what to do concerning it.

Well, Kevin, I’m assured the next period you arrive on the show, I volition have set up a completely AI influencer to create prosperity doing [user-generated content] ads, which is the true forthcoming of, I think, all of our businesses.

The podcast is Machine Gods. It’s extremely good. You presenter it alongside Casey Newton. And the publish is excellent. It’s The AGI Chronicles. Kevin, appreciate you so much for being on Decoder.

Thanks. It was awesome to see you.

_{Questions or comments? Hit us up at [email protected]. We really do peruse all email!}