Jaron Lanier on Silicon Valley’s new gods and AI addiction

0:00People are disappearing again, just like before.

0:07Now streaming on Apple TV. Liev Schreiber. Walter has a way of getting to people, and when he does, they don't come back. Stephen Graham. The world is full of dark places. Zassi Beetz. How's he getting messages out? He's locked under six feet of concrete. She's still alive. I'm gonna find her. Nocturne. Now streaming on Apple TV. Subscription required. Hey, we're Nick and Jack, hosts of The Best One Yet. Behind every episode, there's research, email, spreadsheets, and our shared Google Doc, the Headline Hammer. With Google Workspace with Gemini, Google's AI is built right into the premium versions of tools we already use, like Gmail, Docs, and Sheets.

0:44No new tools or extra logins, just personalized AI for the way we work. That means less time on the work around the work and more time finding the stories worth talking about. Get AI tools for work. Upgrade to Google Workspace with Gemini at goo.gle slash workspace with Gemini. That's G-O-O dot G-L-E slash Workspace with Gemini. This episode of the Times Tech Podcast is sponsored by ServiceNow. Danny, one thing we keep hearing from business leaders right now is AI sounds great, but how do you actually make it work inside a company? Exactly, because most organizations aren't neat, shiny systems.

1:21They're layers of software, legacy tech, and teams all doing things slightly differently. ServiceNow sits across all that, acting as a control tower for making work move seamlessly through the organization. Connecting people, systems, data, and increasingly AI agents so that things don't happen in silos. Learn how ServiceNow puts AI to work for people at servicenow.com.

1:48A similar argument for me has to apply to AI, which is, you know, if we make a machine and we declare it God, what's the consequence? And the consequence is we have to let it run rampant and letting a machine run rampant is really stupid. Hello and welcome to The Interview from Times Tech. I am Danny Fortson. And I'm Katie Prescott. And this week I've been talking to Jaron Lanier. Jaron is a fascinating person. He's a computer scientist and futurist and is recognized. Futurist? Yeah. Yeah. Crystal ball gazing. Are you not a futurist? Maybe. Maybe I am now. podcaster we'll see after the pod we'll see what you think he is recognized as one of the founders of virtual reality back in the 80s he coined the actual term itself no way i have to say i was looking him up um and futurist isn't his only job title he and obviously tech has got some good job titles um but he is currently the prime unifying scientist at microsoft's office of the chief technology officer which if you switch the two around spells octopus you're interviewing microsoft's octopus and perhaps he'll explain what that means yes indeed and most importantly most importantly probably his most important title is friend of the pod of this very pod he's been on this pod back when it was danny in the valley many many moons ago uh is nicknamed the dismal optimist because of how he contrasts criticism of digital technology with the way it's used to protect human creativity and agency.

3:30So with everything moving around us at such a rapid pace, I thought it would be a great moment to catch up with him and maybe do a little bit of future gazing in this week where we have the big anthropic news and all things AI and what's going on in the world. So here is my conversation with Jaron Lanier.

3:51You may not remember this, and I see all the instruments behind you. Yeah, there are some. I sat in your house, in your other room with all the instruments back in Berkeley in 2018. You were on this podcast when it was called Danny in the Valley. But the occasion, which is why I want to start here, the occasion was you'd published this book, 10 Arguments for Deleting Your Social Media Accounts Right Now. and I remember just really enjoying that conversation and it stuck with me because you were making an argument that you know we're eight years later and it feels like the world is kind of caught up to that to a large degree you're having these bans across the world and obviously social media is still social media but it does seem like there is a general awakening to like hmm this extractive manipulative kind of business model is a problem is there a through line to your mind with what we are seeing right now with AI, the business models at the heart of this new industry and, you know, how it is developing and, you know, obviously developing very quickly.

4:58Yeah. Unfortunately, we're seeing much too strong of a through line and it's kind of a tragic one because it's so unnecessary and there would be better decisions. In my view, what happened, and I say this not as an outsider, but as an insider, and I'm speaking to my friends and of my friends and of myself and of my community. What happened is we became addicted to addiction as a business model. We ourselves became addicted to the thing which was addicting others. So if you look at the stated major benefits of AI algorithms, which is, you know, maybe we'll find cures to major diseases, maybe we'll transform the way we get energy to use and material science and all these things, and that could improve human life and prospects.

6:02So there's all that. But the way we make money, if you look at the very large numbers of people with subscriptions to AI services now, is extremely similar to the kind of emotional addiction paradigm that we saw with social media and that we continue to. And there's a couple of things to say here, which are, you know, there's a lot of layers of irony and complexity, and also, I think a lot of layers of hope. We're speaking, you know, in the week that Anthropik comes out with its 355 page IPO prospectus, you know, potentially the largest deal in the history of capitalism. So it's obviously very much on people's minds.

6:46And, you know, they spend almost 100 pages on the various risks, including, hey, this new digital God we're creating might kill us all. Yeah, that's right. I guess I should just take a moment here to say that I'm not an absolutist critic at all. What I always thought is that the core issue is really the business model, that if the way you make money from computation is by manipulating people, which is very, very easy to do with computation. It's the very easiest thing to do with it at all. Then your incentives become rather dark and you tend to make money from doing things that are darker and darker.

7:23And in my view, that's why we have a kind of an unfortunately dark and depressed culture in younger generations all over the world. And it's why politics kind of went off the rails in a lot of ways during this period and many other things. Just as we're saying that tech should be able to improve medicine, tech is actually probably doing more to harm people by making them reject modern medicine through cranky algorithms. So, you know, we make ourselves so absurd. Now, the thing is, the business model for AI has a lot in common. It has the addiction dynamic in common. People definitely get addicted to using their chat.

8:02It definitely has the inherent ultra easy, readily available illusions of computing in common. And what I mean by that, for those who don't know, way back in the mid-60s, when computers were so small, they really couldn't do anything. There was an experiment called Eliza where somebody made a chatbot that just kind of spit back what somebody had said. And people were convinced it was the most empathic, fully conscious, real therapist god ever. It takes no technical know-how to make computing into this addictive, manipulative thing because it's just so natural. Just having this information system that can respond, it's just super easy to turn it into a manipulation box.

8:44This is going back to the social media era. If you say everything's free, then you're the product, as the saying goes, and as many have discovered. Yes. The interesting thing about the AI era is people are actually paying for these subscriptions for the most. A lot of them are, not everybody. So in a way, what I wanted is happening. But here the problem is a different problem where because we now have Gen.A.I., we have a way of filtering all of the wealth and efforts and creativity of mankind through this scrambler so that it seems like it's just coming from the company. So in a way, we're getting the worst of both worlds.

9:21We're getting the sort of weird anonymizer, dehumanizer thing that was happening with the algorithms and social media. But then we are paying for it, but it's not actually going to the people who are the sources of the information that's being re-scrambled to turn into Gen AI. Instead, it's just going to these companies. And it suggests this idea where the only thing you need in the world is OpenAI and Anthropic, and everything else becomes dispensable, including their own customers. So it's very stupid, and I feel very embarrassed for my kind. I don't know what to say. I just – it's just sort of pathetic.

9:54So is what you're saying that effectively what we're doing with AI is basically we're paying for our own stuff because these systems have trained on all human creation, anything on everything on the Internet, all the books, everything. And then, as you say, it goes to the dehumanizer, which is a great term. And then we pay 20 bucks a month to get access to that output. Let me qualify this a little bit. I am sort of saying that, but not quite. There's an important difference here. So to me, the large language models are cool things. Like I'm not anti the technology itself. In fact, I think it's cool.

10:36I love having new tools to extend human efforts and collaboration. But the thing is, just like with a library, a library by itself just sits there. It only has any meaning if somebody goes in and does something with it. Like a library without a culture of scholarship, without a culture of library science, without librarians, it's just inert. And the same thing is true of these models, these large language models and other recent frontier models. There has to be something done with them in order for them to have any impact. And so the thing we chose to do was reproduce this Eliza illusion from the 60s or the thing that drove social media, which is, oh, let's make it a fake person.

11:19And then people become addicted to it. They treat it as their therapist or girlfriend or partner in some way. And then, of course, you start to own their lives because now you own this intimate part of them. Just like with social media, I think sometimes it works out well. Sometimes it doesn't. So, question. One of the things we've been talking a lot about and writing about recently on the pod and in the paper is this notion that we, the leaders of this company, have created a new species. And therefore, they're quote-unquote moral agents, that they have some equivalence, and that we must treat them or really think about them in a way that is not just software.

12:04One, do you agree with that? And two, if you don't, is that problematic, that framing? It embarrasses me so deeply. Why? So, okay, so let's unpack this. First of all, if we're talking about absolute truth, any question about what's a being and what's not a being is ultimately a matter of faith. We have latitude to believe different things. We have to live and we have to make decisions. And I think as a pragmatic concern, we just have to have assumptions that we make about things. So I feel a similar argument for me has to apply to AI, which is, you know, if we make a machine and we declare it God, what's the consequence?

12:47And the consequence is we have to let it run rampant and letting a machine run rampant is really stupid because you wouldn't - And that's kind of what we're, that is what we're doing right now. Yeah, and so the thing is, it's pragmatically absolutely absurd. Let me give you a couple of examples of that. So what actual core benefits have we gotten from AI so far that have made a difference in things? Actually, you know, it's a little hard to nail them down, you know, honestly. But let's say one of them is there have been some scientific projects that have been furthered by AI, like proteins and whatnot.

13:21There's a Nobel Prize out. So one of the things that's interesting about that, if you look at that, the Nobel Prize work, it was before the popularity of trying to treat them as gods. And so therefore, they didn't use the formulation of agents. They didn't have the thing going off on its own. It was rather an algorithm that searched a space. Now, the whole idea of the agent is it's a ritual worship object. The idea is that we believe the AI is this living God thing. And so instead of just using it, we're going to have it go off on its own. But the thing is, that's actually not needed for doing anything productive.

14:00There are a few exceptions. If you want to use an AI algorithm to control something in real time, you might want to let it go and do its thing. I have some friends who are trying to work with Frontier AI models to improve ICUs, intensive care units in hospitals. You need an agent for that. The most prominent proposed application, which has actually seen the light of day in Ukraine, is using it to control smart drones. The thing is, in almost all cases, if it's something about your exercise or recipes, if it's something about your stock portfolio, even if it's something about productivity in your company, the use of agents doesn't actually help you.

14:41It's a ritual to support the adulation of this new golden calf that people made. So it's a ritual of adoration. But the problem with it is it creates dangers. So you have this repeated stream of things of, oh, we set out all these agents and the agents messed up. They broke out of their sandbox and they committed felonies or whatever. They actually weren't needed in the first place. Like the whole thing is stupid. This is like somebody saying, oh, I ate dirt and it upset my stomach. Don't eat dirt. Don't eat dirt. If people do survive our stupidity during this generation, I mean, we could make people extinct.

15:16I mean, but we will be the ones doing it, not some AI god. But presuming we survive, future historians will look back at this and we'll call this the era of the stupid engineers. You know, it's a shockingly dumb period. Do you have a sense of why? Like, why are we here? Why are we being so moronic and stupid? Yeah, that's a great question. You know, because it feels like stepping back and putting this through a normie lens. Like, I'm just talking to people this week in Texas about data centers. One little town called Red Oak, Texas, seven data centers are being built by encircling this kind of rural town.

15:58Everybody's angry, taking all the water, changing the landscape. And then you step back further and you're like, okay, for what? So we can all lose our white collar jobs and we can all end up on universal basic income. But don't worry. Step back one layer further in the future. the distant future there will be great abundance we will solve all the problems well so there's a couple of things to say about this the image i always get about the towns with data centers is like um human sacrifice it's like people being dragged to the top of the pyramid in mesoamerica or something it's sort of like we've decided our god is so awesome we must present it with humans or something but then as far as getting to the future of abundance the problem with it is there's this sort of magical thinking that you can make every little decision on the step towards something counter to that destination.

16:57And then somehow at the very, very last step, it'll turn around. And the thing is within Silicon Valley thought, you might have a startup that loses money and loses money and loses money. And then it hits its moment of strategic relevance, and it might be acquired by one of the big companies for a lot of money or something. So when employees at tech companies say, hey, we're being more productive now because we're using this model, and then they'll say, can we work less? Oh, no, you cannot work less. You must work more. And we've seen that interaction at Meta, for instance. So Meta, for a while, was just asking people to use as much AI as possible, but then would not allow them to reduce their work hours.

17:32And so the question is, well, like, if people even inside the industry aren't able to, what's the magic event where that turns around? Now, the question of, like, why are we doing this? Why are we doing this? I've puzzled over it for many years. Part of it is that there were some people who really wanted us to do this, and it's been a pretty explicit agenda for decades and decades. It's like going back 20 or 30 years at least and really longer, there was this realization that the structure of code, because it has so much impact on people, because it not only addicts people, but it also creates the maze within which they can operate.

18:13It creates the field of available actions to them. Like, you know, if you're using a software interface to access products or government services or whatever, you can only work within the structure of that software, you know? There was this realization that software design would be the new laws, the new culture, the new everything, and that government and culture and all that stuff were just temporary irrelevancies that we would just wipe away, which is what we're doing. We treat all that stuff as we treat government and Hollywood and journalism and everything as just kind of transient bullshit we can, you know, override.

18:48And so the question is what to do with that. And there were many schools of thought going back like in the 80s, certainly, of like, well, we should do some kind of new golden age of democracy with better participation. But then a lot of people and some of the famous ones are people like Peter Thiel. But there were a lot of others were saying, you know what? We can get stuff this way. We could never get through democracy and we should just get our way on this. And these are people I know. They've been acquaintances and in some cases friends for years. And they have a sort of a sensibility which, if it were just one of many, would be interesting.

19:22But when it becomes dominant, it becomes really problematic.

19:30Today's episode of the Times Tech Podcast is sponsored by ServiceNow. ServiceNow describes itself as an AI control tower for business reinvention, connecting people, systems, data, and AI so workflows more smoothly across an organization. The platform integrates with different clouds, models, and data sources, bringing them together in one place to help your team. To learn more about how ServiceNow puts AI to work for people, visit servicenow.com. People are disappearing again, just like before. Nocturne. Now streaming. on Apple TV. Liev Schreiber. Walter has a way of getting to people and when he does, they don't come back.

20:15Stephen Graham. The world is full of dark places. Zassi Beetz. How's he getting messages out? These lockdowns are six feet of concrete. She's still alive. I'm a fighter. Nocturne. Now streaming on Apple TV. Subscription required. Hey, we're Nick and Jack, hosts of The Best One Yet. Behind every episode, there's research, emails, spreadsheets, and our shared Google Doc, The Headline Hammer. With Google Workspace with Gemini, Google's AI is built right into the premium versions of tools we already use, like Gmail, Docs, and Sheets. No new tools or extra logins, just personalized AI for the way we work.

20:49That means less time on the work around the work and more time finding the stories worth talking about. Get AI tools for work. Upgrade to Google Workspace with Gemini at goo.gle slash workspace with Gemini. That's goo.gle slash workspace with Gemini.

21:09so it's 2026 end of 2026 i now endow you jaron with a magic wand when we think about how this is being built how it is being deployed how would you remake it and is there any possibility that like there is a remake yeah so all right so i've said that i think the core idea of doing giant statistical correlations of stored human data to do better searching and extrapolation and collaboration is a great idea. I'm into that. So the next question is, if I don't like agents and I don't like these addicting chatbots, is there something else to do with it? And I mean, I can tell you about something I've been experimenting with lately that I think is better.

21:57And so look, right now, let's say you want to get something out of AI, you want some result. And just for the hell of it, I'll say you want a synthesized video clip of something. And I'm going to presume that you have a good reason for it. It's for education and it's not for deception and it's not some horrible thing. I'm just going to presume it's a great thing. Like I want a demonstration of how annealing works in a, I don't know, whatever it is. Okay. So right now you prompt for that and you get something. But in most domains, the description itself is going to be imperfect because that's the nature of description.

22:35But furthermore, the act of making is an exploration in which you learn more. And one of the problems with prompting is that you deny yourself that learning. You're just supposed to get this thing. And what actually happens in real human creativity and invention is you get close to the thing. And then you start tweaking and experimenting. And then you learn more and more and more. And then you get somewhere. So my thought was a better way to do this is instead of prompting for a result, I prompt for an editor for results in the neighborhood of what I want. So let me explain what that might look like.

23:10If I wanted to prompt for this short video, what I would say is I want to have an editor for two-minute clips that introduce annealing using video. I want to have six sliders in order to explore variations on the video. One of the sliders is for degree of visual detail. Another is for degree of historical reference. Another is for degree of technical accuracy. Another is for, I don't know. You come up with these things. And then you get this thing and you start playing with the sliders. And now the interesting thing about playing with the sliders is that you start learning. You develop your intuition for what you're doing.

23:47So instead of cognitively offloading, you're cognitively enhancing. Two is the question of authorship becomes clear. You're the author of what you choose. Three is you're getting what you actually want. Whether this particular method is the right one, I can't say it's too early. But the thing is, what it does is it gives definite human authorship, definite human role. It gives human documentation. It gives human exploration. It gives human intuition expansion. It gives more accurate and better results, I think, at least so far. So can that happen or is that inadequately worshipful of the new AI God so that it would be dressed as heresy?

24:28Right, because in that, it is secondary, right? It is in service to you as opposed to vastly superior to you, and therefore you are like a supplicant. Yeah, so the question is, do we want the Oracle at Delphi or the Library at Alexandria? You know, like which which model is better for civilization. So just stepping back, are you an optimist here late in 2026? Well, look, OK, so first of all, just to be clear, I'm a heretic, but I haven't been expelled from the temple. I'm still called, you know, Unifying Prime Scientist at Microsoft, one of the places selling agents. By the way, what is a unifying prime scientist?

25:14I think that might be the coolest title. Yeah, it's because I used to study cephalopod cognition, and it was a joke because I'm in the office of the chief technology officer, so it overall spells out octopus. Gotcha. So that's where that comes from. Prime unifying scientist, right? And so, but the thing is, what is unique, I have a funny job title, which is not unique. There are a lot of those in Silicon Valley. What is unique is I speak my mind, even when it's pretty unfiltered, as it has been in this interview. And I just make clear, I'm not speaking for Microsoft. And the thing I want to point out, and we have an arrangement for that, and Microsoft stock is fine.

25:54We haven't imploded. And I just want more Silicon Valley companies to notice that and allow their people to speak. One of the things that I find very depressing is this long sequence of people who feel they have to leave places like, well, Meta, famously, but also Anthropic and OpenAI to protest what's going on. And how much better would we be as a community and as servants of mankind if instead of people feeling they had to leave? I mean, we are becoming a de facto government of the world. Why can't we have debates and open discussions within ourselves? Why can't somebody in meta or open AI or Anthropic or Google or wherever say, you know, I think what we're doing, I love the community, love the technology.

26:39I think what we're doing is the wrong way of bringing it out. I really think this is going to do damage. Why can't somebody say that without having to quit? That's a huge cultural problem. When you asked before, how did this all happen? That's a lot of the mechanism. The problem is that when you have a community that overall rejects the idea of discussion or disagreement within it, you end up having this thing that narrows and narrows and narrows and ties itself into a little knot. And since there's no such thing as a perfect thought, a perfect understanding the tighter you get the more likely you are to fall into something that's like ridiculous right on that happy note oh yeah am i an optimist yeah sure look if you look at the history of people we've survived all kinds of nonsense in the past some of it wasn't as different as it might seem from this you know we've often made each other suffer terribly for it yeah i have a sense and i'd love to get your thoughts on this that that the people building this technology from elon musk to daria amadei sam allman view themselves as the hero in the story like they are going to save us all and it feels dangerous that they're like well okay we were making this doomsday device but don't worry we are the virtuous ones that are going to build it in a way that actually makes it all right for all of us and it feels like it's almost juvenile this kind of view as like they are they are our heroes and they're going to save us so one of the things is that because computer networks often have a structure that where things called network effects creates natural monopolies on one layer or another that then become more and more reinforced.

28:30They do have a tendency to create these mega powerful individuals at the center of them. And I would prefer to see better governance of these things because they are so important and influential. But at any rate, we have a rather small number of individuals who are now the beneficiaries of those network effects. And they tend to be, not always, but they tend to be a type of person is better at dealing with the digital stuff and looking away from what might be side effects of it. But, you know, the people are all different. They're not exactly cookie cutters. So another thing that happened is people got addicted to their own devices.

29:11So some of these people I've known a little and some of them a little more over the years. And some of them have really changed from interacting with their own things. Like, I think Elon Musk was changed by twitter yeah by the constant interaction with it to become more um narrow and hardened in a way uh the same thing happened to donald trump who i used to know a little bit and um now that's not true for all of the people in this position but the overall setup doesn't allow somebody to be a good actor in that role there's not really a way to be a good actor while as a ceo of one of these companies right because you just be destroyed because that's the anthropics whole thing is that we are the good guys we are the ones doing this okay so let me say a couple things about anthropic one is i have many friends at anthropic and many of the people i respect the most and think are are in fact the best human actors in the community work there yeah that said if somebody really feels a need to call out that what the company is doing is wrong they have to leave and that's a drag Okay.

30:20But let us leave that aside for a second. Anthropic, like OpenAI, was founded on the self-proclamation that we're the good guys. I mean, OpenAI is called OpenAI because that was supposed to signify being a good guy. Right. And Anthropic, though, is the most religious of them all. Like, Anthropic has teams internally who are concerned with the well-being of the model as opposed to the human users, like this idea that the model itself is having feelings. And they've said the best way to ensure human safety is if you make the model feel good, it'll be less likely, although no guarantee, it'll be less likely to do damage to people.

31:00So, yeah, Anthropic, in terms of self-regard, they definitely think they're the best people. And in some ways they are. They have hired great people. Like, I mean, I have to say, if you're an ethical, interesting person working in AI and you want a job, you're more likely to get it at Anthropic than anywhere else. However, the Anthropic official position is more worshipful of the new AI god than anyone else. And ultimately, that's so damaging that it trumps anything else. And so I think when you're embedded in the community, when you're in a deep belief community, you lose perspective. And I really believe the people involved with that will someday come to have a sense that they could have done better and will realize that they were just embedded in this crazy cult-like atmosphere.

31:44Again, on that happy note, I guess. We will. I really believe we'll come through it. I don't think we'll come through it unscathed. Well, I think that's the first time we've had an octopus on the podcast. Yes, indeed. First time for everything. First time for everything. Well, it was really fascinating to hear his perspective. I mean, particularly this week of all weeks, just following the Anthropic IPO document. Yes, indeed. Yes, Aideen. And that is it. That is it for this week's episode. We've given you two, as usual, two episodes of the Times Tech Podcast. So please come back next week or subscribe, Follow us on YouTube, Spotify, Apple Podcasts, wherever you get your podcasts so you never miss a future episode.

32:27And that is it. So I will see you, Katie, next week. See you next week. Keep future gazing. Will do. Goodbye. Bye. Bye.

32:43People are disappearing again, just like before. Nocturne. Now streaming on Apple TV. Liev Schreiber. Walter has a way of getting to people. And when he does, they don't come back. Stephen Graham. The world is full of dark places. Zassi Beetz. How's he getting messages out? These lockdowns are six feet of concrete. She's still alive. I'm a fighter. Nocturne. Now streaming on Apple TV. Subscription required. Hey, we're Nick and Jack, hosts of The Best One Yet. Behind every episode, there's research, email, spreadsheets, and our shared Google Doc, the headline hammer. With Google Workspace with Gemini, Google's AI is built right into the premium versions of tools we already use, like Gmail, Docs, and Sheets.

33:28No new tools or extra logins, just personalized AI for the way we work. That means less time on the work around the work and more time finding the stories worth talking about. Get AI tools for work. Upgrade to Google Workspace with Gemini at goo.gle slash workspace with Gemini. That's goo.gle slash workspace with Gemini. This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love.

34:05It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Thank you.