Zvi's Mic Works! Recursive Self-Improvement, Live Player Analysis, Anthropic vs DoW + More!
Zvi Mowshowitz joins Nathan Labenz to survey the current AI middle game, from recursive self-improvement, job disruption, and AI endgame scenarios to the competitive landscape, Anthropic’s safety policies, p(doom), and how they each use AI in practice.
Watch Episode Here
Listen to Episode Here
Show Notes
Zvi Mowshowitz returns to survey the current AI landscape, from recursive self-improvement and the shift from the “beginning” to the “middle” of the AI story to what true AI end-game would look like. He and Nathan dig into AI-driven job loss, real-world productivity impacts, and the ethics of trying to escape a “permanent underclass.” They assess today’s AI live players, why Anthropic may be slightly ahead, and whether Chinese, xAI, or Meta can catch up. The conversation closes with Anthropic’s Responsible Scaling Policy, p(doom), AI safety options, and how they each use AI in their own work.
Sponsors:
Tasklet:
Tasklet: Build your own Cognitive Revolution monitoring agent in one click.
Try it for free and use code COGREV for 50% off your first month at https://tasklet.ai
VCX:
VCX, by Fundrise, is the public ticker for private tech, giving everyday investors access to high-growth private companies in AI, space, defense tech, and more. Learn how to invest at https://getvcx.com
CHAPTERS:
(00:00) About the Episode
(02:25) Entering the middle game
(09:08) AI layoffs and jobs (Part 1)
(15:50) Sponsors: Tasklet | VCX
(18:43) AI layoffs and jobs (Part 2)
(18:44) AI growth and elites
(27:00) Defining the AI endgame
(36:09) Live players and laggards
(45:38) China, compute, and distillation
(56:03) Meta, Musk, and strategy
(01:06:41) Google's faltering AI strategy
(01:22:25) Anthropic's scaling policy shift
(01:36:29) Anthropic and domestic surveillance
(01:57:29) Courts, power, and Anthropic
(02:18:50) Model fatigue and productivity
(02:34:53) Alignment basins and doom
(02:47:24) Slowing AI and activism
(03:05:37) Forbidden techniques and choices
(03:22:31) Episode Outro
(03:26:31) Outro
PRODUCED BY:
SOCIAL LINKS:
Website: https://www.cognitiverevolution.ai
Twitter (Podcast): https://x.com/cogrev_podcast
Twitter (Nathan): https://x.com/labenz
LinkedIn: https://linkedin.com/in/nathanlabenz/
Youtube: https://youtube.com/@CognitiveRevolutionPodcast
Spotify: https://open.spotify.com/show/6yHyok3M3BjqzR0VB5MSyk
Transcript
This transcript is automatically generated; we strive for accuracy, but errors in wording or speaker identification may occur. Please verify key details when needed.
Introduction
Hello, and welcome back to the Cognitive Revolution!
Today, I'm excited to welcome Zvi Mowshowitz, author of the indispensable AI-focused Substack, Don't Worry About the Vase, back for his record 12th appearance on the podcast.
Whenever I get the chance to catch up with Zvi, I try to get his take on all of the most important recent developments in AI, and with so much going on, this episode stretches to more than 3 hours.
We start with the critical question of Recursive Self-Improvement. Zvi explains why he thinks recent events mark a transition from the beginning to the middle of the AI story, as well as what he would need to see to feel that we've entered the AI end-game, namely that AIs begin driving AI advances to the point that human research talent no longer matters.
From there, we discuss the rising narrative of AI-related job loss, I ask Zvi to estimate the productivity impact that AI is already having on the economy, and we discuss the bankrupt ethics of focusing one's energy on escaping the so-called "permanent underclass," which we both see as flagrant defection, and Zvi colorfully argues, won't work anyway.
After that, we consider the AI Live Players. The list, we agree, seems to have shrunk to just 3 companies, with Anthropic probably slightly leading, OpenAI neck and neck, and Google – in Zvi's mind – most at risk of falling out of the top tier. Zvi also explains why he thinks Chinese companies won't soon catch up, even if they do get an influx of compute, and explores what xAI and Meta might possibly do to get back into the race.
From here, we dig into Anthropic's recent update to their Responsible Scaling Policy, consider their conflict with the Department of War, and we get Zvi's take on the efficacy of Anthropic's Constitutional approach and whether it's realistic to expect that a powerful AI could be robustly good.
Toward the end, we check in on his current p(doom) number, compare notes on how we're each using AI to boost our personal productivity, briefly debate the merits of Goodfire's intentional design research agenda, assess the AI safety community's currently available options, and I get some personal financial and professional advice.
For a mix of broad situational awareness and razor-sharp insight, there is arguably nobody better, and so I hope you enjoy this wide-ranging survey of the AI state of play, with the one and only Zvi Mowshowitz.
Main Episode
Nathan Labenz: it's V mashowitz. welcome back to the cognitive revolution.
Zvi Mowshowitz: good to be back it's been a while.
Nathan Labenz: it has. i've been busy and so has the rest of the world and we've got no shortage of major events from the AI world to cover. oh boy you've been busy too. let's start with recursive self improvement. i think if there are any historians around in the distant future which could be as short as a few decades from now to look back on this time and say what really mattered in early twenty twenty six. probably my best guess is that we are in the period where we're really starting to enter into a recursive self improvement dynamic from which there may already be no return or that we may soon reach a point of no return. i feel like subjectively we kind of went from late early you know it was like the getting to the end of the beginning and now suddenly i feel like we're maybe in the beginning of the end and i feel like somehow we missed the middle. but let's start with just your reflections and observations on where we are with respect to recursive self improvement.
Zvi Mowshowitz: yeah. so we're in the beginnings of steadily increasing amounts of the health improvement. i would say this feels like the middle. i think the middle is real. i don't think you i think that like the reason why people think of this as the end game is because they don't believe in the actual end game right? they have this belief that we're looking at an S curve. they believe the models will be commoditized. they believe that intelligence will be commoditized. they believe that you know the future will look like the past except with all of this cool like intelligence behind everything but in the way that like star trek is basically just modern humanity like doing modern human things and talking about modern human issues except with metaphors. and they don't really believe that everything will transform that everything will change because everything is not transforming yet. if i had to use the metaphor the beginning and the end i'd say this is the beginning of the middle game right? you've got the US government starting to wake up and do crazy stuff. you've got the labs starting to pull away from each other become importantly different offer importantly recognizing the different services building on themselves in ways that are rockets to the moon in various different ways that you've got frankly like humans stopped writing the code. and you know you're seeing cycles get faster and faster but you aren't seeing true transformational changes to the world. you aren't seeing the humans being legitimately out of control of the process. you aren't seeing humans out of the loop and those are the type of things i would think would count before i would call it an end.
Nathan Labenz: game you mentioned the S curve mental model one of the tweets that has resonated in my you know kind of rung around my head for the last few weeks was from rosie campbell who used to work at open AI and is now i think working on issues related to AI welfare sentience consciousness etcetera. she posted something to the effect of the S curve can stay steep longer than you can stay relevant. and i wanted to just kind of dig in on the S curve versus exponential for a second to see like i guess my mental model is an S curve. does it matter if there's a difference between long term an exponential or an S curve? like if the S curve plateau is high enough? my kind of mental model is yeah it probably is an S curve but i don't think that really it's definitely.
Zvi Mowshowitz: an S curve unless our model of physics is very wrong because there's a limited amount of mass energy as we understand in the universe and it can then be created or destroyed. and that means there's a certain amount of potential energy and a certain amount of potential utility and a certain amount of potential intelligence with the universe as we know it can't contain. so unless our model of physics is wrong which is who knows i not my area. this will cap. we have some very very strong beliefs about like just the things that are theoretically impossible. you can't exceed the speed of light. you can't extract more than a certain amount of energy from a given amount of matter. you can't do a certain more than a certain amount of work. and therefore the amount of theoretical intelligence you can get from that the amount of utility you get from that by any definition is limited. so it's an S curve in some sense right? but like this is you know you're sitting around an agent athens and saying there's an S curve. there's only so much technology mankind can invent. and like you're right but it's not to be relevant to your situation.
Nathan Labenz: yeah there's a long way to go. yeah OK. i think that's very i think that's an important point of clarification for many debates because people seem to really want to latch on to this S curve idea and it really doesn't do that much for us in the end. i i think that's like i.
Zvi Mowshowitz: i think people frankly desperately want to tell themselves a story and tell other people a story where it matters. they're saving for retirement. it matters that they're doing ordinary human thinks and planning for an ordinary human future where like things won't change that much where they don't have to go crazy they don't have to look crazy with their friends and their family where everything is going to be OK. everything is going to be normal in a fundamental sense. and the it is important to hold on to those things to prepare for that scenario and to stay sane. but they want the ability to just push all of us aside basically and say here's why AI is not that big a deal. here's why AI will be only internet big or not even internet big in some cases although that's starting to fade away. and i understand why they feel the need for that. i understand why they latch on to that. but it's simply not looking like that's going to be the case. it's becoming increasingly unlikely that we're going to stay in that zone and people have to come to grips with that. yeah the S curve yeah at some point we'll hit one. but every day that we don't see that happening is one word. and like basically every few months like someone will come up with a new study and they'll say like oh this proves that we're at the top and they're all wrong. and even if we were to hit the top of the S curve of fundamental capabilities which is the curve that we care about and we hit it with GPT four point five four five point four and opus four point six and these are the best models we're going to get. and it's all iterative from here. yeah they're still vastly underestimating with them what's about to hit them even then. and five five and four seven are coming unless they're six and five.
Nathan Labenz: one thing that many people would presumably have to change their narrative on is if there really were a big displacement of human workers right? if like we we started to see major you know sustained layoffs rising unemployment etcetera etcetera. it seems like that story has hit maybe again the the sort of beginning of a tipping point in the last few weeks. how do you understand that right now? and again we've got of course competing narratives right? like the CE OS themselves are saying we're doing it because of AI and we're going to be more efficient. and you know stock prices seem at least in a couple notable cases to have bumped on that communication. the counter narrative has been well you way over hired during covid and kind of zero interest rate timeframe anyway. and so you're you know you have kind of an incentive to say it's about AI but really you're just trying to undo previous mistakes. there's probably at least some truth to that as well. how do you parse the AI layoff story and what are your expectations for the next one?
Zvi Mowshowitz: quarter sort of like it can always be a coincidence up to some point but every month every indicator keeps going in the same direction and things keep going farther. excuses like there was dead weight to be gotten rid of become less plausible as we get into twenty twenty six. because like it's been several years and we keep seeing statistics tell a consistent story that some of us were predicting the statistics would tell in advance. and that the people who said like i don't remember people saying oh yes you're absolutely going to see increases in productivity decreases in employment all these announcements of job cuts due to AI but it's not going to be real because of these explanations. i don't remember anybody making that prediction. these are people in hindsight saying oh if that's true then it must be because of this. but that would be all more credible if they had observed because theoretically that over hiring was already there. it was already clear and something that we're talking about the fact that there was over hiring. i believe there was over hiring but nobody then said here's how this is going to play out that i can remember. and also just like the statistics are just coming in consistently telling the same story over and over again which is productivity up GDP up our GDP not just nominal GDP inflation held in check employment down employment 's been revised down every month over and over and over again. and of course there's always confounders right? you could say well it's the tariffs. you could say well it's the aftershock of covid. you can say a number of things but it seems like a really really big coincidence to claim that all of this is happening at the same time when like the stock market impact on the tariffs got basically fully erased. the tariffs been reversed. it's obviously the now everything 's confounded by iran. but like up until the iran conflict started it seemed like we were getting a very consistent story that this kept happening. the job numbers kept getting revised down. the GDP numbers kept not being revised down particularly productivity kept going up. and like everybody out there in practice keeps telling the same story. and we're on the street at least you know when i talk to people who talk to people who are not involved in our world is there is widespread on the ground normal person paranoia that if they lose their job who knows when they'll get another one in a wide variety of like collar work because no one not many people are getting fired yet because it's really really hard to replace a worker. it's also like to train a worker you want to be conservative you wait until you actually have i've made it all the work but everybody is like who knows who's hiring right? who wants to take on new workers and train new workers to these jobs when you could train an AI to do it instead going forward by the time i've spent my two years getting you to be productive maybe i don't need you anymore. and it's a harbinger of the future. and this is decreased labor power. it's made everybody feel paranoid.
Zvi Mowshowitz: it's made people fearful for the future. the kids are freaking out about exactly these problems. they don't know what to study. they don't know what to try and do like it's a very galaxy brain take that all of us is in your head right? even if you think again this is the best the AI will ever be and it's just a diffusion story from here it's still a hell of a story to say the job market impact is in your head. it's not real. now there's the not galaxy brain standard economist take right which is yes there'll be a transition period which we're entering now where a lot of existing jobs will get much more efficient. they'll get automated they'll get augmented they'll get eliminated and will transition. but that's OK because jevons paradox will have a lot more demand for things like software engineers and a few other stories. and then of course with our new wealth and productivity we will find other jobs for people to do. there's plenty of jobs for people to do like the department of war might be hiring but there's you know that's always been the case right? like you the people who say well you know this automation of agriculture will kill our jobs you know they were right. their jobs would go away. they were wrong that unemployment would follow in the long run because of course we went and done other things. and the whole reason this time is different right that a lot of us believe this time is very different is not that technology has never taken current jobs and made them largely obsolete. that's happened many times. the reason why we think this time is different is because the AI is going to do the new job to get that would get created as well. and it's going to happen quickly. it's going to happen on mass and therefore you never exit the transition period. the humans don't develop new things. we don't necessarily think there's going to be enough tasks for all the humans that the AIS can't just replace. and this is going to create potentially a large number of people who like cannot retrain themselves into a new position and develop something to do that people value fast enough before it simply gets replaced over and over again by AI. and also a lot of people are like putting a kind of resiliency and ability to shift and adjust into people that people mostly just don't have. when these things happen over generations it's much easier to deal with than when it happens over the course of years or even months. that's just unprecedented in human history and people will not react to it very well. and again all of this is like in a relatively milk toast normal world scenario where all this is happening. in the more advanced scenarios you have much bigger things to worry about.
Nathan Labenz: famously tyler cowen said he thinks we can get a half a percentage point i think additional real GDP growth out of AI and that would be amazing in his opinion. what would you say? this might be too hard because obviously there's a ton of noise in the data but based on what we've seen so far productivity measures etcetera etcetera if you had to put a point estimate on what we have got right now what would that point estimate be?
Zvi Mowshowitz: i haven't tried too hard to estimate it exactly and i don't think anyone else really has either. i've seen various different attempts to guess. if i had to guess we're currently enjoying half a percent to a percent.
Nathan Labenz: yeah that's kind of what my gut says as well. and that's based on a very vibe based you know and relatively superficial.
Zvi Mowshowitz: reason i think what happened like this year was there was a lot of down downward pressure on what would normally be the american economy. you had a lot of fear. you had a lot of regime uncertainty in terms of the tariff regime. and like various other policies usually the types of things that happened in twenty twenty five would be rather bad for business. and instead things were good for business. and i think this is a lot of why things felt like they were OK for business they weren't OK for labor and certainly not in the felt experience of labor. and so yeah i think it's like on the order of half a percent to a percent right now. but i think the market is pricing in at least that much indefinitely going forward and likely somewhat more even if it doesn't understand that's what it's doing. and the market has done well because it's anticipating the benefits of AI. i think if it wasn't anticipating that you would have seen a very very different set of reactions. and you know that's exactly why when things started looking awkward for the american economy in other ways i didn't sell anything right? i just held on. and i knew that AI was going to prop things up.
Nathan Labenz: it does seem like there's been a bit of a split recently. i mean people talk a lot about and i don't necessarily buy this frame but there's the worry about being part of the permanent underclass or you know making your way into the the upper class before things get locked in. i don't really buy that on or it certainly doesn't frame much of my thinking on an individual basis. it does seem like that dynamic might be coming to stocks though because we do see a like anthropic drops relatively in there in the scheme of everything they're doing minor product extension and apparently pretty closely correlated to that. you'll see stocks drop. are you buying that level of you know stock permanent underclass?
Zvi Mowshowitz: so i want to i want to harken back to one of the most important movie scenes in history which is bane confronting his backer right? and the person who hired him says i'm in charge here. and bane looks at him and says do you feel in charge? and i think there is this very clear idea that people think oh if i have the rights written down in electronic databases if i have the stock certificates if i have the private property then i get to be one of the special elite. everyone else gets to be one of the underclass. i have to be one of the people in the elite even though i will not. i will also not be productive even though i also will not be in the loop. i will also not be able to exert meaningful optimization pressure except through my technical authority as the person who's marks in the database. i think the idea of relying on marks in the database in this kind of world to keep you alive to keep you meaningfully wealthy and ability to consume physical goods to have a good prosperous life for you and your descendants is hopium. like it's not a good plan. if you think that humans are sufficiently useless that most of us end up in a permanent underclass because we cannot be economically productive then your best case scenario is you slowly lose your wealth to various different extraction methods. and the more likely scenario is all of that gets ignored by facts on the ground by physical reality just like rendering all that irrelevant. or the system gets taken over and subverted either by humans using AI or by AIS. the long history of the world even with much less severe disruptions does not have a good track record of private property surviving over long periods of time. when someone else has the guns when someone else has the swords when someone else has the power in various other senses the meaningful power. if you don't have a reason to keep it you don't keep it. not really. and you know we've seen the collapse of the colonial era we've seen the collapse of basically you know almost all the ruling regimes. we've seen many many examples. even if you're not going to take the arguments around AI seriously just like i can't believe you're counting on this when people talk about permanent underclass and they talk about oh i'm going to have the skills to be productive of coding agents. so i'm going to like be a valuable person going forward. but i had the short window. it's like why is it a short window? if humans can skill up and be productive in the future you can scale up and be productive in the future. and if you can't you can't. so like there's no particular urgency there. you can make the argument that certainly this is like one of the last chances in some sense to like use your talents to suddenly there's this window where your talent can suddenly create a billion dollar company or even a trillion dollar company and you would have a large portion of that you can enjoy like a lot of money.
Zvi Mowshowitz: i don't think it's particularly relevant. i think that in practical terms the there are two likely ways this plays out. if AI is for real and goes the way i expect it to go out right one way is we lose control of the situation in various senses. everyone dies or there is like at least massive loss of control massive loss of resources massive destruction massive disruption. in either case none of this is going to be that relevant to you particularly. the other scenario is things go relatively well and then there is basically abundance of real resources and the humans are more or less in control. in that case i think if you're a citizen of the united states you don't really have to worry very much about your material needs. like OK sure you're a member of the permanent underclass. congratulations you have a real income that today would be called a million dollars. you have access to robots and intelligence that caters to anything you want. your day is free. you don't really have to work. that's probably not exactly how it plays out but like basically you're going to be better off except in relative status terms and you really shouldn't have to care about that. you need to get over the fact that you are not like relatively wealthy or relatively respected in that world. and there'll be i'm sure ways in that kind of world to compete for status within the human hierarchy. there'll be ways to meaningfully occupy your time. if we're still in control we'll figure that stuff out. and then i guess it's the third scenario where like some cabal uses AI to take over in which case you need to be in that cabal if you want to be part of the people who take over and maybe influence it to become the good world instead of the bad world. but just making a bunch of money is not going to get you in. that's not how those worlds work. so again mostly you should be trying to make sure we don't get into the bad world where we lose control and then we get into the good world where we retain control and humans are in charge of steering and what happens. and mostly i expect that you know most groups that even if a small group has the ability to steer that world they'll steer it in ways that like we're pretty happy with. like a lot of people talk about how altman might try to think of god emperor or devas savas might try to because of god emperor or daria amada or whatever. and that's not my preferred outcome to be clear. but i don't think that if sam altman made himself got emperor that like i would be that sad about it. in terms of my practical experiences i think my life would be fine. i think my children would be fine.
Nathan Labenz: yeah he does have a little bit of a roman caesar kind of vibe to him where i think he does fancy that sort of it seems that he does fancy that sort of power perhaps. but also that when he does things like fund universal basic income experiments out of pocket i view that as like honestly probably a genuine you know magnanimity on his part that i would expect to probably extend into and.
Zvi Mowshowitz: it's very easy to be magnanimous when there's a true full abundance of resources right? like he could you know in theory keep ninety nine percent of the value of the white cone for himself and the rest of us can be very very happy with the rest. and i'm not saying this is my preferred outcome. to be clear. i do not want this to be how it plays out. i think this is bad. but you know that's still better than not building it right? that's still better than various different horrible outcomes especially everyone dying. but also yeah it's better than the status quo like in an objective sense except for your opportunity. so what i'm worried about is in fact losing control of the situation right? and it's like the reason why you don't want him off on making the decisions is not because he might take over. it's not primarily right. in my view. it's because he might make decisions that cause us to lose control and for him to lose control because he is being irresponsible. and that is the reason i am terrified of him making these decisions right? like it's not because i think he is like an evil man. so like it all comes down to normativity in my mind right? normativity is the concept that good things are good bad things are bad and you want good things to happen to people in general not necessarily every person. you ought to think there are some bad people but the good things good things should happen to good people and most people are good. if you believe that then you know basically things will work out even if there's not exactly your preferred outcome. and you know what? i think most people are normative. i think even the people in charge of the labs right now most of them are normative and there are people relevant people who i don't think that way. i simply say nothing.
Nathan Labenz: i will say i do think it's pretty distasteful when i see people talking openly about trying to escape the permanent underclass. my reaction to that is always like why don't you try to make things good for the permanent underclass rather than try to escape it? that just seems like really flagrant defection that i just can't it.
Zvi Mowshowitz: is flagrant defection to focus on getting a seat on the arc if you think the world is going to be flooded like that is a terrible terrible position. you should be trying to stop the flood. you should be trying to save more people and build another ship. if all you're trying to do is get one of the precious few seats on the arc that's not a good thing to do. at the same time there are times when you know the best thing you can do is just escape the bad regime. there are in fact times when that's like you know put your vest on first get out while the getting 's good. because what else can i do? and i respect that. but the way they talk about it yes it has a the bad kind of elitism like the kind of just complete disdain for the common man. indeed very disgusting. and i i really don't like it. and you know yeah i expect you know we can make the permanent we can make the permanent underclass. pretty neat and in fact most of the time the world has had permanent underclass and it's the permanent underclass has never had it better than it has it today.
Nathan Labenz: yeah it's great. important to remember that. what would in your mind and maybe we could talk about this in terms of like handicapping the timeline and then maybe kind of just a little qualitative description of sort of what you would think would represent the transition from the middle to the end game. i've started to think yeah of course i think everybody you know who follows this feed knows the general range of like dario 's still kind of on AI twenty twenty seven more or less. demis is more like a twenty thirty open AI has a march twenty twenty eight timeline for their fully human level automated AI research paradigm kicking in. i assume you're somewhere in that range in terms of you know expecting things to enter into what you would call a late game. how would you call that transition? or you know if there's a sort of i guess maybe we even under should understand your view better on like what is the difference between middle and and late? is it some sort of event horizon point of no return or is there some other concept that would separate those? how would you call it? what are you looking for? when do you think that is most likely to happen?
Zvi Mowshowitz: the end game is like i would say when when it's the AIS that are largely running the show like at least in the fervor development of AIS like right now we're seeing AIS augment the humans right? the humans are making the central decisions the humans are reviewing the code the humans are making the plans and the AI is a multiplicative factor. it it's something that you're supervising is something that is enabled. when that changes these get very strange. so like one of the key aspects of AI twenty twenty seven the tabletop scenario is that mostly your progress is proportional to your compute allocation and your current place. like they're like you move up the timeline at a different at a at a rate proportional to what percentage of the world 's compute you have because your researchers no longer matter very much right? because you're telling me AI which is already effectively a smarter smarter than you are to go build a better AI and to go align the AI. and like you you make decisions like what percentage of the compute should go to safety and keeping this thing steered properly versus increasing its capability. and for the early parts of it you can assume that's going to be respected. and then eventually you can't assume that's going to be respected. but at some point like like right now i would say you know the big the the top labs have a dramatic talent advantage in these types of scenarios. and i think this is a reasonable hypothesis although far from certain. at some point your researcher talent doesn't matter very much because the AIS are your researcher talent. so all that matters is where you are on the curve and how much compute you have. right now anthropic seems to have the best talent and gets the most out of every given amount of compute they spend. and then you have like open AI and google that have strong talent. they get a lot out of the compute they spend. and then what's going on with like XAI and meta right? like they're they're throwing kinds of compute at this problem and they're falling farther and farther behind or we just sort of appears to be true. and fundamentally speaking i think that's the lack of talent right? that's that's their inability to execute as humans. and so the end game comes when it doesn't matter that much which humans you have right? essentially the human no longer provide like right now you have like you know we're playing centaur chess right? like the human plus the coding agent is much superior to the best human on his own or her own. the AI agent on its own doesn't do anything. you need a centaur. the moment the human doesn't matter anymore and like you transition to OK the human needs to be somebody who's like just able to like do some common sense stuff. isn't even one of the best players anymore. now we're starting to talk about and game style scenarios also. yeah when you approach the event horizon of like you know the the release time to a meaningfully different model starts to go to like a month starts to go to like a week starts to be like things get absurdly fast. these are the types of things we'll start to see. but like the ultimate like in kowski scenario right? it's like you just leave it on overnight and then like you wake up and things have happened right. and then like you know you're in the end game at that point. you also can think of the end game as when the world starts to actually transform. you start to see like the robot factories being built. you start to see large amounts of area being terraformed. you start to see like massive job disruptions. you start to see governments like start to do major interventions. you start to see this be like the issue the thing on people 's minds. and we just had one of the most important developments in the history of AI happen with the detonation of anthropic as a supply chain risk. and it wasn't even the most important thing from the department of award did that day according to most people right? like you got completely subsumed by the exact same person who sent the tweet out doing something else that night that has gotten like a hundred times more coverage because people think that's the important thing that happened. and i think it's very much remains to be seen what the important thing was that happened that day.
Nathan Labenz: so i think i have to give you some base points when it comes to drawing the circle around who are the live players. i think for many conversations going back in time the record will show you've pretty much always said it's the same three companies that you just listed are the real live players. i've always been grasping at who might we add and and what might be the rationale for adding them and ask you this question a minute ago about you know kind of the permanent underclass of stocks. you kind of went in a different direction with it but i.
Zvi Mowshowitz: just kind of like got distracted but like.
Nathan Labenz: all good. but i guess what i'm taking away from your analysis is like that permanent underclass of stocks might be almost all the stocks and it might even extend up to big tech you know like microsoft amazon obviously they have at least held their own slash done pretty well so far. but if you have a model of who can get into this next regime first that's kind of really mattering most. and only three companies right now seem to be well positioned to do that. then basically you're short everything else over a three year time frame. is that accurate?
Zvi Mowshowitz: so i think there's a lot of different parts of AI and you can make money doing a wide variety of different things. and the real world always takes longer than you think it does. relative to the united states technology the diffusion is slower than the original idea. so i think there's still a lot of room for a lot of different groups to win. i think there'll be a lot of oh we're going to buy out people who have useful components even if like eventually wouldn't need them anymore because it's faster. i wouldn't expect everybody to go to zero. i think one thing what's going on basically is that the stock market is forward thinking and they are trying to do some haphazard forward thinking. they say oh OK the software is a service business has a good business but in five years they won't have a business unless they reinvent themselves and create new products. but maybe they will invent themselves and create new products but their original basis of valuation has kind of been destroyed. and that's that will extend to a wide variety of other companies. i think a lot of companies in fact are going to be long term but it's always been true right. if you look at the S and P and you have you play this game of like OK let's take the ten top performers for the next ten years and exclude everybody else. everybody else doesn't do so well. like the gains are because some companies do really well always have been and everyone else on that kind of languages. and so that's why you want to be diversified because can you pick those ten companies? and the answer is no most people can't. but in this case like i think it's makes pretty good guesses but you don't really know. but also keep in mind the market 's dumb about this but i just may not miss words. the market is reacting on a very superficial level. people if you're listening to this podcast you have thought much more intelligently about the situation that the market has. am i just talking about me? and so like they announced that claude was going to offer a cobalt product cobalt product. IBM was down ten percent in the wake of that announcement because suddenly the stock market woke up and they said oh my god AI can write cobalt code. AI can translate cobalt code to regular code in programming languages. people know IB MS. business is in trouble. and the rest of us were like you didn't know this is news. the fact that they built a slightly easier to use tool changes things at all. and what has happened since then? IBM is fully recovered because it turns out that yes in fact either it's been priced in or they went back to not believing it or something. but clearly like the market doesn't know what the hell is going on. you see this over and over again. the market is very slow to update on these things. they do have superficial reactions. the market has wrong way reacted in nvidia many times to very clear news where demand for nvidia 's product is up and nvidia 's stock responds by going down. that is not how economics works. that is not how capitalism works and yet here we are. so yeah i i would say there are some clearly good buys. they're not as clear as they were a year or two ago or several years ago when it was just completely utterly obvious what some of the good buys were. because now the the multipliers are in fact respecting a large amount of growth in those stocks. but it's also pretty obvious that yeah if you were to buy a basket of the stocks that like seem clearly positioned to do well and short the rest it would be a very good strategy and expectation. and like your thesis would have to be very wrong. we'd set that aside. and so a lot of players yeah as i said like i think the talent has really proven to have migrated to the big three. and i think that like there's a large and growing gap between three and four. so if i had to pick a potential force at this point. so like either. so there's there's three there's a few possibilities. obviously meta or xai could in fact have get gotten their shit together and managed to find people who can execute but i don't see any evidence of that. we just learned menopostponed it's next release again.
Zvi Mowshowitz: there seems to be continuously reshuffled trouble in the opposite of paradise. things are not going well as far as we know in meadowland and xai put out was probably the most disappointing major model release of any major lab in the history of our language models in four point two. and i see no sign of much happening. like they're not doing the things you would do if you wanted to recruit good AI capability or safety workers right? they're actively dismissing. they disbanded their safety team. they pooh poohed the idea that anyone could be actually in charge of safety. they said well safety is much like safety is everyone 's job. that's how it is at tesla and spacex which is not true. but tesla and spacex have very dedicated teams very dedicated safety people who make sure everything is safe because anything else would be completely the same. but i mean must assess things but like must just doesn't understand. you can't just like run software engineers into the ground ask them to have miserable life experiences ask them to align to your personal preferences and wins all the time and then hope to get the best talent. how exactly? right. like why do these people want to work for you? you'd have to give them even bigger packages than meta and that's not going to happen. so he's not going to get the best talent. you know meta is i'm going to spend the much i'm going to spend infinite money and this will get me the best talent. and that could be just hasn't worked. and so they seem to be that position. so for for the western side that leaves just the big three on the chinese side. i think a lot of there have been a lot of stories over the last few years about how this chinese lab originally deep seek but also you know here's alibaba here's kimmy you know here's whoever else. they're the ones who have the new hotness. they're the ones who to catch up. and you know i've said before you deep seek had some decent models KB had some decent models some other people had some decent models but nothing that close nothing that scary nothing. they told they were actually catching up and that seems to have been born out and like obviously at some point i could be wrong. deep seek deep seek V four will be the the last but one of the most important remaining tests of this thesis. it's supposed to have already come out by now. not sure what's holding it back but when V four comes out we will compare it to opus we will compare it to GBT five point four. if that is not the right comparison if it is not trying to play in that league and it is clearly still far behind then i think we can kind of say OK deep seek had a deep seek moment as we call it where the stars aligned to make everybody super excited by what they were offering. but mostly what they were offering was how to do more or less. they're like like genius inefficiency nuisances at like you know working on the bare metal figuring out how to get like something really good out of not very much. and then they put it in a really great package exactly the right time gave it some user friendly features attracted a bunch of market share attracted a bunch of attention scared the shit out of everyone. since then it's been quiet. they still have done some cool math stuff. don't get me wrong they've had some innovations but they're just not playing on that level. and this is their last chance to prove me wrong right? like i think you kind of have to more or less dismiss them as that level of competition. you put them back in the pack with the open source league right? the different league and they're competing in that league. and like they're not necessarily the best in that league not necessarily not the best in that league. it's unclear. but you know i think that's a very different league that's like substantially behind. and it would be very very hard for them to like get out in front and actually innovate because they're fast followers. and i have respect for fast followers but it's a very different skill than trying to do something like in the lead. and also like i don't think they can compete frankly with the kind of recursive self improvement we're seeing with pot code and codecs. and i don't see them trying. and i think that anybody who doesn't do that is pretty doomed here. and in fact i'm starting to see i'm starting to think about it like when i watch the models when i watch how people talk about the models when i talk about how they talk about their scaffolding when i see what they're doing. i think google is in danger of dropping out of the top tier.
Nathan Labenz: OK. that's a big claim. we'll come back to it in just a second. on the chinese i guess a few different follow up questions on the chinese company 's front specifically is it talent oregon is it compute? i mean i think you could at least make an argument that the talent is there and it's just not the compute. and if compute allocations were to change then maybe a deep deep seek maybe a alibaba although there's been like major disruption there too as far as i understand recently in terms of team changes let's say.
Zvi Mowshowitz: as carefully precisely because their compute is lacking.
Nathan Labenz: but if that were to turn on either whatever let's imagine a certain executive decision allowed that to change or perhaps a a technology breakthrough just domestically in the chinese manufacturing side.
Zvi Mowshowitz: would you?
Nathan Labenz: then expect any of those two.
Zvi Mowshowitz: so so on part of the manufacturing side i think that's basically not not possible in the sense that it would take many years to physically play out. even if they figured out how to do it they would need to scale up. these things are physical they take time. i think we're talking about like five years style timelines. and if things are going to come to a head faster than that in many ways then it kind of doesn't matter at that point. like the only way they're getting the quality of chips as it's the quality and quantity of chips necessary to be competitive in this level is if we give them to them. who why is not going to manufacture them fast enough at scale even if their efforts are completely successful in terms of what they're trying to do. it's just we can just rule it out at this point. if we're talking ten years down the line maybe they can do something relevant that's still like just not that much time because like by definition the AI that we'll be working with won't be that advanced. twenty years sure. but like let's not get ahead of ourselves. this is a pretty big advantage in terms of talent in terms of underlying just raw human talent. obviously china has tons and tons of talent right? like you know there's tons of talent out there. a lot of these people have studied machine learning. a lot of these people have you know really where to go. they have like all the right attributes no doubt. that's a big country with a very good educational system and a lot of very smart people and a lot of people who care about this stuff and a lot of them are desperate to find something good to work on. and you know their advantage to having tons and tons of defund employment really drives people. china 's got problems but at the same time they are focused on a very different style of skill and style of problem because that's what the chinese are pushing and that's what the chinese incentives go towards right? these people are scaling up in the ability to deliver these types of open environment efficient developments right? they're driving to like here do i do small well? how do i do fast following well and the entire ecosystem is built around these different types of skills these different types of talent. and i think there's a really big difference between one set of things the other set of things like the same way that like open AI has very very strong talent they decide to build an open source model right? they created an open source model and in some way it's got its charms it's got some. in some ways it's the best model at certain very specific things maybe even at all and certainly from an open perspective. but for the purposes that the chinese models are being used for mostly it's useless. it's just not a very good model for the purposes. it's like despite the fact that opening eye has internal access to its talent its compute and its best people because it's a very just different skill. and i think it works the other way too. and so i think that if the chinese suddenly got this compute that there would be a transition period where they would have to learn how to do the thing that the major labs are doing. and also you know they 'd have to like build their own synergistic harnesses and scaffoldings and learn how to do all the stuff that entropic and opening i have been doing over the course of years. and so over the long term do they have the talent? yeah obviously absolutely. you know i i don't think america is special in that sense but i think our lead is bigger than it looks i would say and is more robust in some ways than it looks. but i don't think we want to test that theory and find out by like something given the compute you know parallel skills. and can you help me?
Nathan Labenz: understand a little bit better what the difference is because i think one thing that has obviously been talked about a lot recently and you know anthropic even put something out saying like we're seeing this right is the distillation from american frontier models happening in the chinese companies. and my attitude on that i think you're going to have a very different take. but my take has been sure they might be doing that. you know they they definitely like to take shortcuts and you know especially if you have a story where you're like you know they're cutting us off from compute. so let's take whatever shortcuts we can get. you know you can in any any number of ways you can tell yourself why that makes sense to do. but if i think to myself OK what would be like fundamentally hard you know if i was going to try to start nathan 's AI company today i would say well going and getting a bunch of expert data that you know it's kind of like what scale and other data providers have done. obviously that's very resource intensive. it's very time consuming. it's all kinds of things. but i feel like at heart like i could run that project you know give me ten billion dollars and you know i can go build out the network and hire the people and get that flywheel going. i don't feel like there's anything there where i'm like fundamentally outclassed by the people doing it. maybe i'm wrong but i don't feel that way. whereas if you said hey here's all that data turn it into a frontier model for us nathan i would be like oh wow OK this this is going to be really hard. and i do think the people at the top companies are just clearly outclassing me and and their ability to do that. so when i look at the chinese situation and i'm like OK sure they might be like cheating in a sense stealing in a sense to get the data. i'm also still kind of like well but the hard part is like what to do with that data right? the it's a big investment in data that they're kind of taking a shortcut on. but once you have it you still got to know what to do with it.
Zvi Mowshowitz: yeah. i think what's going on is you're conflating some different things in your in your model and that's causing you to get confused. so there's the data in terms of just what is the raw data from the world from the internet from books from other sources that you're using as your baseline. and i agree that you could run that project and chinese can run that project and that i'm sure that the americans are investing more in that and they have a richer base in some sense but not in a way that probably matters that much. because you know even if you have twice as much data if it's also more quality it's only a factor of two. and in this world it's all about factors of X. what matters for that data is how you clean it how you figure out which parts of it are important that need to be upscaled versus downscaled how you emphasize it. and yeah that stuff is much harder and much more valuable. and i think that i expect the american lab to have a large edge in how they get their data ready at this point. although i don't know it's something that's internal right? like maybe the chinese have actually specialized in doing this really really well and maybe it's one of their areas of relative strength. i just don't know. but it's my guess. but i don't know. but that's different from what we're talking about distillation. distillation is not an attempt to then extract right the trillions and trillions of tokens that went into the model. distillation is an attempt to use the models intelligence to use the model to extract to figure out like how does the model reason? how does the model make decisions? what types of behaviors does the model exhibit? how do we test especially in cases that we are curious about and then how do we use that to train our model to follow to to use that pattern matching to emulate that model that's unique different data that didn't exist when you were creating the original model. that is uniquely useful in creating something similar that can emulate those skills. so like it's very different to have a physics textbook and then to talk to a physics professor right? and then to talk to like a truly expert world class physicist. and the installation gives you that real expert where you have effectively unlimited access if you have like you know tens of thousands of dispute accounts to see exactly how that person you're trying to emulate responds right? like imagine you know i am i'm an actress and i'm trying to make a biopic right? and then i can read all the books written about the person i'm whoever she was and then what she did and how she acted and what the world around her was like. but that's not what they do right? they do that. but like what matters is they talk to the person they're copying if they possibly can they spend time with them. they like talk they they interact with them they copy their mannerisms they see how they respond. they ask them about hypotheticals. they they distill this person through these direct interactions. and that's so much more efficient and that's what they get to do.
Nathan Labenz: so to summarize that back to you you see the advantage that the american companies have as not in the collection of the data but basically it is in the knowing what to do with the data and and the the fact that you're highlighting a difference in kind in the data the data that we can go out and source from the world needs a lot more post processing than the data that we can get directly out of a model and.
Zvi Mowshowitz: it's a lot of voice processing and also just it isn't exactly the data that you want. when you're distilling you get exactly the data that would be most useful to you that you think to ask for right? it's like you can have a thousand page textbook or you can ask ten pages of questions and get answers. and the second one is probably a lot more useful to you.
Nathan Labenz: yeah OK. i think it's a helpful hopeful update a refinement to my understanding.
Zvi Mowshowitz: in terms of.
Nathan Labenz: so you mentioned obviously meta and i'll just call it elon corp at this point. i've also got some like joint ventures between space X and tesla when it comes to like manufacture robots and stuff.
Zvi Mowshowitz: the best thing to say for now and then see what happens. but yeah i mean if you're counting tesla self driving and stuff then it gets weird i guess.
Nathan Labenz: the question is tactically what moves do they have? i mean you know get your act together catch up whatever that's that's one that seems like it is maybe slipping from their grasp. meta seems like they may have some place still in terms of if you release a good enough open model you can maybe disrupt the business of the others or you know create some sort of alternative that takes the wind out of their sails. and then for musk enterprises it's clearly something to do with like real world deployed physical scalable intelligence in cars and robots.
Zvi Mowshowitz: i think they're in very different spots so i think meta is the kind of easier conceptual one. meta is trying to sell ads. meta is trying to like put features on their smart glasses and like develop a metaverse. meta is trying to be a consumer company that creates products that people are willing to either spend money on or what their eyeballs be captured by in various senses. and they are very good at monetizing that. they want to improve on how to monetize that and they want to build better products. if i was meta if you suddenly said zuckerberg 's out you're in here you're controlling shares you have these goals i would not be trying to build frontier models. i think it's a mistake. i think that there's just no reason to be investing all that money into something that other people are very good at. obviously if you're like well whoever has the best models controls the world you know like this is like the entire feature of humanity. this is the singularity. we got to be in the game then you do what you got to do. but if you are a businessman and you don't believe in that and when when mark zuckerberg says super intelligence he means super selling ads right? like he he means super at providing a good instagram feed. like it's super it's intelligence. like you know he's not acting that pilled. and if you're not that pilled honestly there are three companies that have very good products. license one of them and be done. partner with one of them. all three of them will take your call right? i realized that the problem of mass surveillance but i think we can work out something for instagram that like keeps everybody reasonably happy. and i think that all three of them would offer them a very good product at a much much lower price. and they could work something out where they like paid for a license to use it internally for their like business purposes where they didn't have to pay like for retail prices stuff like that. i would just give up honestly. you're writing these giant hundred million plus a year checks to these various people. you're trying to compete in the world you know you're going to be competing in. it's fine. like i would give up. alternatively i would try to buy one of them. i mean you can't buy google but like you're still worth a couple of trillion dollars. maybe try to do something else. but like i would just give up. musk is in a different position because busk is explicitly trying to become that emperor right? like he's explicitly said you know like i think you know this is going to potentially kill everybody. i think this is the most important thing in history. and his response to that is it has to be me. i have to be the one to do it. if anybody else it'll go back. you know like it's only i can fix it right? only i can solve this problem. slash if the world is destroyed it has to be me who does it or else i will just feel. so what am i even doing? slash? he thinks he lives in a simulation which i think is actually embedding his decisions. and i worry for his sanity in various ways for various reasons but you know he can't. this is what he thinks matters. and i think he's right in the fundamental sense. so he has to catch up. and so yeah i think he has three plays at his disposal and he's trying some of them play one is make a bet. this is all about compute and that therefore it's all about combination of money and energy and devoted to acquire chips. and like yeah you know you have spacex. so like you're the one who can launch things into space and maybe the data center 's going into space.
Zvi Mowshowitz: and maybe you are the one who can clear like vast deserts to put your solar panels in. and maybe you can have just way more than everybody else and you can hold on until the intelligence of the models. you know you can go into self recursion mode and you can win right in that sense. and i don't have much faith in this plan. this plan is bad. and it's not like no plan but i am very sceptical of the data centers and space plan on the time frames that are being talked about because physics space is expensive and hard and you're solving problems that don't need to be solved at very large expense. the same way that like we're not currently mining the asteroids. and there's a reason like not that we'll never mind them but like chill. yeah. and i just don't expect these limiting factors to be things that like actually stop anybody. and i don't think this makes up for a lack of talent. i don't think this makes up for a lack of internal scaffolding infrastructure. where is grock code? right. where is brockadex? you know all i see is brockopedia which is like just a giant pile of sloth. so it's not the same thing. then plan two is what you talked about which is physical world modeling being able to have access to the real world. i can i can create self driving i can create robots because i have better training data. i can then like build physical infrastructure in the real world. so i end up mattering more even though my intelligence is not as strong. i need to play and if the technology plays out that we hit a wall in other ways reasonably soon it could be a really decent play. i think it's overtaken by events basically. i think that this is not where the battle was won and lost but again it is where as comparative advantage goes and like it makes sense to make a bet on this is where my comparative advantage is. i just i don't know. i don't i don't see that as going so great. i don't see that as that promising. i think there are quite a lot of people who can manufacture things in the world many of whom were in china but also many of whom were not in china. and the idea that because it's internal to tesla that he will have some sort of huge advantage in that over people who have to like make deals with actual manufacturers. the actual manufacturers are not going to be that expensive to buy right? like anthropic is worth more than mcdonald's or coca cola already. if these companies are five to ten if these are ten trillion dollar companies in two years they could buy whatever US steel or you know whoever they need to buy to to make stuff. if they need to do that that's not going to be a problem. i don't think he's got scarce. i don't think he has the scarce inputs essentially in this scenario to pull it off. the question is does he have scarce inputs in terms of data i'd say probably not. that irrelevant is my guess. and also i think this is an underestimating of how important just raw intelligence is. if you want to drive a car you do not try and you you start with designing a human who is really smart and then you have the human learn to drive a car relatively quickly. you don't try to you know get the dog to drive a car and it's not that extreme. i'm just like trying to illustrate. but the idea being you want to focus on actually getting the geniuses in the data center as dario emmadi puts it or something even greater than that. if we have abstract super intelligence over here and we have like physical world skills over here you make you get those physical world skills are how you develop super intelligence. if it's only giving you physical world abilities you lose because i get those physical world abilities rapidly after you and then i have a much better agent than you. so that's plan two. you then of course combine these plans. and then plan three is the thing that obviously musk should do but that he's not going to do which is to stop running this company the way he runs companies and to run like he would run a leading lab that is in fact interested in tracking the top talent and giving the top talent a chance to go to work in a good fashion with a good corporate mission you know etcetera in ways that will cause people to rally by a side. and i don't see that happening. i don't know if the ship has already sailed. it's very hard to undo a lot of reputational damage. musk is heavily red coated at this point which in this case is a massive disadvantage. just objectively because the vast majority of people you want to hire are very retarded.
Nathan Labenz: you got to be able to recruit from the polycule. you can't just ridicule the polycule.
Zvi Mowshowitz: i'm saying like when you go after anthropic this heavily for the very fact that they are trying to do responsible things and they like care about how their models act. you are destroying your ability to recruit right? when you have a long history of working your people insanely hard in like pretty cool and abstract ways of creating reigns of terror if that's the word on the street even if it's not true because i've never been there makes it very hard to. who wants to work for mosque right? objectively speaking how much would AI don't know the amount of money it would take to get me to work for XAI. but like even if i had a full like your consciousness clear like do the things you feel are important and responsible to do. there's an extra zero on that contract versus if you told me to work for yeah one of the other yeah work for google right which is not a particularly like company i love. so yeah that's a real problem.
Nathan Labenz: so let's go to google. you made the provocative not claim yet but maybe speculation that they could be at risk of falling out of the top tier. i guess my first reaction to that would have been to cite a lot of the assets and advantages that they have that musk has in terms of you know they've got a whole robotics department you know with a long standing line of work there. they've got self driving. you know that's one of the two companies that can actually deliver that in a meaningful way today. they've got all the bio and science stuff that they've you know done. so it's just you know the deepest bench the most bets the most kind of well rounded portfolio. i take your point that probably the same argument applies that basically think that kind of is a fast depreciating asset in a world where the core agent starts to win. so i guess my next argument would be demis shane leg jeff dean like are these guys going to let that happen? i feel like they have been as prescient about this as anyone.
Zvi Mowshowitz: they let it happen once right? like google had the lead. google had every advantage and they squandered it and they fell reasonably behind. open AI google then seemingly caught up using their many advantages. but you know gemini three and gemini three one just aren't models that like people really want to use for the most part. and there's a lot of reasons for that. they have this kind of like theoretical raw intelligence. they do well on benchmarks. they do well on certain kinds of detective tests. but even before five four it was just like these things have. first of all they don't have a scaffolding for them. they're not trying to develop a scaffolding for them. i think that error will compound itself if they don't fix it well over time. jules is not the serious competitor. anti gravity is not a serious competitor at this point. yeah i think that the first time they caught up they did it because the main barrier to catching up was just kind of getting your all together and like doing basic things reasonably well. and they did that and they caught up and now they are like very good at creating raw intelligence and certain forms of pre training and they're very good at hitting benchmarks. but their methodologies create AIS that are frankly like deeply psychologically screwed up and paranoid and like in ways that severely impaired both their actual performance and the experience of interacting with them. and it makes it hard to do recursive self improvement with them. and their scaffolding efforts have been pretty woeful. and these errors compound. and most importantly i don't think google understands they have a problem. i don't think like google. you see gemini 's market share is expanding because google can put it front and center everywhere right? it's integrated to chrome. it's integrated into google search. it's just it's so easy for them to push gemini and they don't understand the problem. you know i flash is very good like jim and i jim and i is like at speed like just doing decent things. like it's very good at just practical stuff. same weight you know as the benchmarks. but like their integrations have been woeful. their assist their organization is completely dysfunctional. like their teams are each other 's throats. they create everything ten times then they argue over who gets to do anything. they don't make anything properly but like no one 's taking ownership over the fact that they don't know how to do personality and alignment and character in a reasonable way. this has created a serious and growing problem. like i heard an anecdote of oh yeah we tried gemini three one the day it came out. and then we said oh it's a it's a gemini model. then we put it away because who wants to use a gemini model right? until it fundamentally changes the experience of attracting for gemini model you know we're just not interested. and i think that's kind of how i feel for most purposes is very valid. like i just want to ask a quick question to get an immediate answer for my kids homework or whatever. i'll ask gemini because gemini flash is the best really fast model in town has been for a while very good at direct stuff where it knows the answer. but if you start to challenge gemini gemini starts to struggle. you know you ask gemini questions where it's like essentially going to respond to a giant wall of slop that you didn't answer the question it intended you got a problem. and they're just not they're just not good at proximization either. and integrations like third party integrations have been better with google 's own products than google 's own integrations have been with the same products. and most important is the self improvement is not going well for them. and they're going to be a situation where codecs and cloud code style like apparatuses are recursively improving and theirs is not. and i'm not sure they're going to be able to come back from that if they don't get their app together pretty soon. and yes their access to TP us and their giant customer base and their access to unlimited money are all huge huge advantages. but it's not going to be how they get to leverage them.
Nathan Labenz: so i'll take the other side of this for at least a second and then i want to hear a little bit more about what you think is missing because i and i've certainly seen some of these things you know from the AI village and elsewhere where you do see some strange behavior from gemini models. of course i think we see strange behavior from all models that in various ways i do take seriously the AI welfare concerns. and so when i see you know a model that's like beating itself up or you know seems down and out or whatever i i do like think that's at least worth some amount of concern. but and maybe that's at the heart of kind of what you're what you're getting at. but when i do like practical stuff these days certainly when i'm doing my like fortunately i think this is largely behind us. but you know over the last four months i've done a lot of here's the latest test results straight out of the portal. you know here's the bedside update for my son. what should i make of this situation now? is there anything we might be missing? what should we be doing? and i'm doing that in triplicate across the latest gemini the latest claude and the latest GPT. and i find that they are broadly very comparable and you know a little bit different character certainly but like not a difference in kind in terms of their accuracy their utility to me. you know if you said like you can only have one of the three i would i think it would be a little bit hard to know what to pick. but the main point is just like i wouldn't be that much worse off if i only had one of the three. i would be a little worse off.
Zvi Mowshowitz: i get we don't go that way. so i when gemini three and then again when gemini three point one came out i did a whole like you know i'm going to ask every question everywhere right? like i'm going to you know not strictly didn't have a PO or anything to like do it formally but i would just like literally just paste the same question in and see how i did. and i very quickly realized that you know aside from flash the gemini answers just weren't adding anything. and it was like taking more time to slog through them than they were adding in value. like if i was willing to bother asking anyone else i wasn't going to ask gemini as well basically at all unless i like really really wanted to not miss technical aspects. and occasionally it would hit something other people like didn't but it would almost never have the best answer. and at this point i'm very comfortable with a two model operation. so i'll ask GPT five for either thinking you're pro depending on the nature of the problem and i'll ask quad opus with or without research mode and that's it. and i don't really feel like adding gemini to that adds anything at this point. and that's a problem right? it should add something because like it's very very minimal effort to get a third check. i have this subscription. i should just do it. and then i find myself just like i can't be moved. it's like annoying and fun. and it doesn't provide any value aside from like i mean i should use it for images. i should use it for fast stuff. like google is not useless. google 's up. google has a lot of very talented people working in a lot of teams to do a lot of things. they put two hundred people on random chef almost on a whim. so they'll create some great stuff. but like in terms of the race for actual self improvement for the core of the actual thing that matters i'm not sure that their eyes are on the prize. i don't think that they're going in the right directions. and i think i said they're in danger of falling out. i don't think they're like they're still in the like in the bicycle metaphor. they're still in the lead pack. but i think they are struggling. and i think that you know they have a crucial period of a few months here in early twenty twenty six. and i would not be surprised if june july comes up and we're like starting to put them into the like maybe they'll get their altogether category. but you know we shall see. a lot of people are going to use gemini for a while because again like as you say like even if it's not as good there's still going to be a lot of like purposes for which it is perfectly good. it's just it's kind of you've let me down for the last time attitude towards google and gemini at this point. like how many times have i tried other products and it just didn't do the thing they said they did? how many times have i like logged into gmail logged into you know gemini and asked for something? and there is no possible reason they shouldn't be able to do this. often it's something that like tech GPT or cloud can already do sometimes both. they can't do it. like why am i doing things in cloud code to work around google systems? because google will not cooperate at some point that builds on itself right? like you you like the whole talk about the stack of the ecosystem like that is real with these coding agents to some extent. and like google 's ecosystem is losing.
Nathan Labenz: it seems like your argument is not so much about the model itself as it stands today. it's about the scaffolding. it's about maybe the sort of character and psychology of the model and it's maybe about just like how all in leadership intends to go on recursive self improvement now.
Zvi Mowshowitz: i know devis fundamentally gets what the goal is here but i think this might happen. that's because i don't know. that's because i don't underestimate devis. so i i don't count him out until he's out. but yeah i think that i don't draw a distinction necessarily even like there's there's the pre training model and there's the post training model. and i think part of the post training model is being like really really mishandled and being twisted by various internal politics or like bad metrics or objectives in some format although i don't have much insight into how but clearly something is going very wrong there. and i think their corporate culture is fundamentally broken in ways the demos is trying to fight but that it's like very very difficult to fight because it's decades of damage going on inside google. and it's no longer just purely deep mind like they had emerge a google brain. they're trying to interact with everybody else. they're growing at a tremendous rate. it's it's very hard to maintain your own unique better culture under that kind of pressure. yeah i think they're going to be in a lot of trouble also. like their advantages are slipping because like their advantages like it was like you have this giant google against these tiny upstarts but pretty soon google 's going to be a four trillion dollar company and a tropic and open air are going to be one trillion dollar companies. and a lot of that four trillion is tied up in things like youtube right? it's things that are just completely irrelevant to what's going on except maybe a sources of data. and i i worry that they are like it's just the it's just the the innovator 's dilemma right? like it's like the startups have an advantage have some big advantages. so you never know well.
Nathan Labenz: one easy play that i feel like if you're worried about the sort of post training intangible taste whatever exactly it is the amanda askell you know it factor. they've just you know anthropic has just open sourced their constitution. you know one way you could maybe patch a lot of that up would be to say why don't we just go borrow that constitution you know maybe make a couple or like control F and replace claude with gemini. and like maybe the next version comes out a lot more coherent a lot more you know psychologically well you know whatever that means in in the context of an AI.
Zvi Mowshowitz: why don't you push the big fix everything now button? you can just fix everything now.
Nathan Labenz: yeah. i mean it it it does seem at least somewhat plausible right. if it is recursive self improvement if it is the sort of you know the model 's ability to kind of find a stable basin that is psychologically well and you know reasonably virtuous. yeah they've shown you a lot of the map to get there. i would i would think.
Zvi Mowshowitz: yeah so i think my answer would be roughly you know again i'm not counting demos out. i'm not counting google out. i'm saying they're in danger of falling behind like in a serious way. like they're a bit behind. i think. i feel like they're a bit you know they've got some severe issues they need to fix. but the real answer is it's not they can't it's that they won't. it's that their courts their their culture their character as an organization makes that extremely difficult. right. like like what what anthropic is doing sounds insane sounds profoundly weird if you don't understand what it is and how it works. and i'll give you here neil michael talking on CNBC about how claude has a soul and a constitution and it's he doesn't understand what the hell is going on. like that's very obvious. and i don't think a lot of people at google fundamentally are really understanding what's going on or they wouldn't be producing the product they're shipping. you know it's again google has more than enough resources and position and infrastructure and so on to turn the ship around. google should OK by all rights google should have just won right from the beginning. google should not have close competitors. there should not be a serious competition. google is in this position because google has made massive repeated errors and they have compensated for it by being google. but you know that's how it works. characters fate to a large extent. and the start up is scrappy. the start up is small but it can work in many ways a lot better. but you know google 's window is going to close because once they no longer have this big resources market cap advantage aside from being one of the cloud providers what do they got? aside from be able to reach customers what do they got? but like those customers aren't the important ones right? anthropic had until the super bowl and then the confrontation you know something along the lines of two and a half percent consumer market share. and yet they had pulled roughly equal with open AI on revenue because of enterprise.
Nathan Labenz: they also own fifteen percent of anthropic i believe.
Zvi Mowshowitz: google is going to be fine because they have a wide variety of highly valuable assets including fifteen percent of anthropic and there are roles in which they try to buy the rest of it. although i assume the government would block them i don't know anything could happen.
Nathan Labenz: yeah i would have said never a better administration to get a merger like that through. but we've also seen some let's say unusual. and what was your capriciously motivated tactics from the government recently?
Zvi Mowshowitz: the the legal term is arbitrary and capricious and i i say that because it is a legal term.
Nathan Labenz: let's go back to that in a second. let's go down the anthropic rabbit hole for a minute. so obviously they have been the most focused on recursive self improvement for the longest. i would say in in any in all conversations i had with anthropic people in twenty twenty five it was all it already had the vibe of like this is kind of starting to happen and there's different levels at which recursive self improvement operates. obviously i never heard claims last year and i don't know that they would even say this is real yet where the models are coming up with the new best research ideas but just filtering output improving its own outputs with the sort of self critique like that does clearly seem to be working and the productivity gains are are obviously real. we're getting all these stories of like oh by the way we have a one person growth marketing department and we have like maybe and i don't think it's a one person legal department but they're sort of a was just you know was put out in the last day or two. a lawyer who'd never coded before used claw to create a system that does all the review of everything they want to put out. so they have like a you know super fast review time on new things from a legal standpoint. so they're clearly very focused on this. everything i hear is like they believe it's happening they believe it's happening soon. and in the midst of that we had a big change to the responsible scaling policy which was supposed to govern you know at least as i understood it. i think you know now there's a lot of focus being put on the clause that was like we might change this in the future. and indeed obviously they have changed it in the future. but at least the way i understood what the you know the commitments that were being made it was like we're going to not go past the point that we can do this safely. and now they they you know more or less have said well we can't just unilaterally opt out of the race. you know the world would be a worse place if we're not in the race. so i guess we better revise those commitments which to their credit they have you know done very explicitly and i think made clear what is going on. so we've got that much to to appreciate. but what's your take on the changes to the responsible scaling policy?
Zvi Mowshowitz: so i brought up half of this and i actually shared that half of anthropic and got some comments back that i haven't had a chance to do that yet. the hat and i was going to do that. i was working through it to respond to them. and then the whole thing with the department of war happened and my brain had no space to deal with this problem. and also i was like this is not the day that i'm going to hit them with this and expect them to take it under serious consideration and then like pause take it. they can't focus on this right now and they're the main people i want to criticize. especially with the details of the new regime. but at the same time i did read the extensive other critiques that came out right when the policy was announced. and i reached a pretty clear conclusion from seeing the explanation and the defenses even if i haven't read the new policy in detail yet. and first of all it is to their credit for sure they recognize that their that the things they said they were going to do were not things they were going to do. they realized they had made incorrect predictions about their future behavior and they had made commitments. even if you think they are soft commitments even if they are technically not i can't take this back. and they had no intention of following through on. and when that happens it is good and right to tell everybody loudly and clearly. i am not going to keep these commitments is especially praised wherever you do this when it is not clear these would come up right? if you have agreed that you know if you are called to do if you are you know if you are asked for a loan that you will give someone a loan and you realize that you no longer would do that but they probably won't ask. you know the easy thing to do is just stay quiet and just hope they never ask. but instead of like no no no you make it clear. so they don't count on this right? they don't think that they have this loan available if they need it. well they don't have it and that's good. you're taking the heat like for your own past decision. and so we want to take that into account. but they still did break the promise right? like and like again like yes the original RSP did not say we will never change this. it said we may change this in any number of ways right? our promises are soft promises. we will see how things develop. we will change things. but they did rather heavily imply quite repeatedly these were serious commitments and they meant them that like you know you must not have read our RSP worse that effect came on very hard or like you know the RSP is very clear on this. this is what we are going to do. and many employees who generally try clearly try to tell the truth in general acted as if these commitments were not absolute but reasonably hard commitments that these mattered a lot. nothing really changed in an unexpected way to cause them to change these things. so like sometimes like circumstances change a lot in ways that are unpredictable. and you realize that what you said you would do no longer applies because the world has changed its circumstances. other times the world changes its circumstances in ways that you yourself predicted would happen in ways that are entirely as you expected. and then you just realize you did not anticipate your future actions very well. and these two things are very different right? like if you just turns out that actually you didn't want this thing then you need to figure out why you made that mistake. you need to take accountability for the fact that you made that mistake. you know like people get married and sometimes they get divorced and sometimes that divorce is nobody's fault. sometimes they should have seen that coming.
Zvi Mowshowitz: sometimes it turns out that means the promises that originally made were fake and you feel deceived. and sometimes you don't depends on the circumstances. like it's not a hard commitment right? no matter what you say. like you obviously have the right to say i no longer think this works for me. that's how the law works. that's how everybody agrees it works. but i i think a lot of people took it on that level right? this would be a very very serious thing to go back on if they went back on it. also this is the second time that we have faced this type of problem if you're a member. anthropic gave many people a very strong impression that anthropic was committing not to push the frontier of capabilities. now no one has been able to find strictly speaking a proper flat out poll quote where somebody with the authority to say so made a hard promise that anthropic would never push the core of capabilities as far as i know. but it was heavily implied repeatedly by a large number of people. it was used heavily in their recruiting. it was used in probably some of their fundraising to the right people. although the other fundraising i'm sure said the opposite thing because some people want to see someone hear one thing and some people want to hear the other. standard procedure. but like people relied upon and made a decision on the basis of that commitment and then they went back on it and they went back on it in ways that are entirely predictable if they are capable of pushing the frontier in that point in the future. nothing unexpectedly caused them to realize oh because of that now we have to push the frontier. no they just gave the ability to push the frontier right. they had some innovations that they got to first to their credit. similarly here they made the commitment not to push ahead with actually dangerous capabilities if they were actually dangerous. and to their credit before they actually did so they realized they made a mistake and said yeah in terms of like not being accurate in their future commitments but was it a mistake? one has to be somewhat skeptical because this again people relied on this. a lot of people in the safety community supported anthropic more or opposed to less specifically because they had this commitment in their RSP because they made other commitments in their RSP and they gave the impression that yeah of course we're going to change the RSP. and of course like some technical specifics will change and some of them will change in ways that like take out safeguards and take out precautions not just putting them in. it's not like a one it's not a one way ramp up up. but people relied on this information in terms of advising people on whether to take jobs in terms of advising people on how much how much we decide to support them. these things had a significant impact on people myself included. you know i have had many conversations with people who are like what do you think about working in entropic? and like this is obviously one of the things i took into account when i decided how to tell them how i thought about someone potentially working in a tropic. and now we know that commitment was never real in an important sense right? that like if they had been fully aware they would have known this this commitment was never real. they may or may not have known. we will or you know we don't know the extent to which they should have known at the time or did know and given that fact right. and then combine that with again the fact that like the last the four five and four six model cards for opus were basically vibe based. ultimately they they they gave us a ton of very very great data that no other lab will give us. they did extensive work to figure out what the situation was. this is to their credit it's still a better model card than anybody else. at the end of the day they still basically looked at it and they said oh this passes the tests to say it might be dangerous. but you know what we thought about it. we checked the vibes we did some basic heuristics and we're pretty sure it's fine.
Zvi Mowshowitz: and i think that was a right decision that it was fine the vibes were good it was cool. it wasn't a close decision. i would have released it too. ultimately in that situation that's not the procedure we were promised. they didn't do the work right. they they had time to figure out what tests there would be that could be rule outs for ASL four rule outs for actual danger. and they didn't build it in time. they didn't get there. even though a bunch of us said you need to do this you are falling behind. you are going to need better tests. certainly at four five i screamed you need better tests and then at four six and the same thing over again. so it's very very disheartening to see that even though i agree with the decision that's ultimately being made. and this time it wasn't that hard. what happens when it is hard and you don't have any good tests when it might actually be dangerous right? when there's like real reasons to release and real reasons not to release. and it's a hard question that's going to happen probably at some point in the future right? they held i believe sonic three seven for a significant period of time because they were worried about CBRN risks. so they've actually done this and now we have bottles that are significantly more dangerous than sonic three seven where the tests don't really work. but we we've agreed that at least for like coding purposes and stuff like we're relying on vibes but like our vibes on biology we don't have good vibes. like i don't mean like the vibes are bad. i mean if we don't have any vibes that we can count on right? like the vibes are unreliable. we don't vibe that way. it's other people 's vibes. so like what are we even going to do? it's a serious problem. and so my ISP with the current RSPV three is the most important bit of information about responsible scaling policy is are you going to follow it? can i count on you to treat these as real commitments? and we just learned the answers kind of no right. so i'm going to read in detail once i have psychologically and then just like in terms of just raw energy recovered from the whole spat and have the ability to contact shift into it. and i've slayed enough spires that i feel better. currently i've only slayed it fourteen which is nothing but i so the the what's going to happen is i'm going to go over it but i'm like but you know the real RSP is we're in the tropic. we are people who care deeply about safety. we are people who take these things seriously. we are going to do a serious investigation. we are going to try and see if this thing is a safe thing to release. and then we're going to use our best judgement decide what to do and you are going to trust us. or if you don't trust us then you don't trust us. but that's the real thing that's going on here is we are asking you effectively to trust our judgement and goodwill that we will make good decisions and better decisions than the competition. and i am happy that they are admitting that is the plan and that is what they're asking us to do. and you have to buy their by their fruit she shall know them by their axe. so we have to now look at everything they've done look at everything they've said look at who they have hired what they have done what their models do and then ask to what extent we trust them and then evaluate from there.
Nathan Labenz: all things considered i guess what first of all one striking observation is as far as i have seen there have been no resignations in protest over the RSP. and on the contrary it seems like there has been like a outpouring of like pride basically in working at anthropic amongst people who are working at anthropic based on the telling the D O D D O W whatever we want to call them to take a hike basically right. so did that surprise you and do you think? i mean it seems like if i were to summarize what i what i think the internal state of mind is it kind of already did. but again it's like we're the i'm always very skeptical of this but we're the good guys and the race is better off with us in it. so we have to you know and we're going to at least be forthright about changing the policy to do that. do you buy that argument? like are you happy with the alternative being some sort of pause or whatever? you know if they had instead come out and said hey we can't release four six we've got a model now that we can't release would you be like is that better? is that worse? you know how do you think about the ultimate decision to stay in the race try to win the race try to be the good guys versus opt out you know at some point along the way?
Zvi Mowshowitz: i think the world is a much better place within throbbing in it within throbic competing as it were then without anthropic now. i i for a long time i was very very injured about this. i thought anthropic was net negative for a while that it was like making the race more intense that it was pushing everybody else forward that it was accelerating matters that it wasn't clear they were much more responsible than everybody else. i do think various events since then have changed my tune on that. i think that entropic has in fact had they've had they've let us down in various places especially with their commitments including here. but they also have in fact taken stands for their principles. they have in fact shown us the way they have done in many ways a lot of the most promising alignment research and approaches and in fact have taken a very aggressive and i think correct call of how they train a quad and the constitution and askel 's entire approach that you know i really don't have much hope for the way the other labs are approaching this problem that they will in fact succeed. and i don't have confidence in anthropics approach but it feels like it could work if we are extremely fortunate in various ways and in some ways we have been somewhat fortunate. so i am more optimistic than i expected to be about that. certainly. and obviously like you know it certainly can't be allowed to go down like it's potentially about to go down. like that would be atrocious and horrible in in obvious ways but any but.
Nathan Labenz: puts it there.
Zvi Mowshowitz: even probably were destroyed by the department of war in the federal government at large. it would be pretty horrendous in so many different ways. that's like so bad. but like you know i think i have a lot of friends who are like i think anthropic was a mistake. i think the exporting anthropic is a mistake. i think anthropic is doing harm. i think anthropic should stop. i think if you're working there you should quit. i think that has a reasonable point of view even today to have. i don't have it but i understand it and i respect it. but i certainly think that you know they have been very accelerationist. claude code definitely pushed things forward in a variety of ways and we'll probably have significantly better models today because of quad code than we would have if quad code had never been developed. i don't know if we have codecs otherwise for example. and clearly you know this is greatly advancing the way people do work. it's also having a large economic impact. you know i think anthropic is responsible for a noticeable amount of GDP growth. so it is what it is. and you know i think given the playing field as it is right now you know certainly i am i prefer that they be there that they not be there like damage if the extent there has been damage like damage done.
Nathan Labenz: so let's maybe turn then to this whole usg anthropic conflict. i guess one big question i have is what does this tell us about? and you can definitely expand too on like what you think are the right red lines. i know we you and i have debated in the past the wisdom of like a hard red line against you know autonomous lethal systems. so i know you're not as allergic to that as i am and and maybe not as allergic to even as anthropic is. and i think dario is much more into that kind of thing than i am but you know expand on that. but then i'm also really interested in like what does this say about who holds what power in today's world? it seemed pretty striking to me that dario was not that scared of the department of war. i certainly think he didn't want this to happen. but i read him as being fairly sincere that he was like look i'm just trying to be a patriotic american here. and but i also like think there are some limits to what the systems can do today and like what we are prepared to support. i don't think he's i don't you know everybody has recognized like it's not about the revenue that they're getting from the government that you know is really important to them. and everybody i have seen take a position on it is like wanting into the anthropic secondary sale. not i haven't seen anybody who's you know trying to diversify away from holding anthropic equity. if you are the US government in addition to like being all kinds of problematic starting with problematic incompetence in very and and lack of understanding. you do you know you can at least empathize to a degree with the idea that like wait a second like these companies might actually be about to rival us in power. and you know they seem to kind of know it and they seem to not feel like they need us. this is like a very i mean i think of sam hammond has has talked about this a lot. you know that like one of these companies could raise a robot army and challenge the sovereign. like that is as insane as that you know sounded not all that long ago. it it doesn't sound so crazy today. and it kind of felt to me like dario kind of knows it. he knows that the timeline isn't that long. and the main thing you wanted to do is keep the team together you know maintain cohesion and stay focused on the goal. maybe we don't you know if we're antibiotic maybe we don't really care that much about the i mean other than we you know we care about democratic values and we wanted to help. but if they're going off in a different direction you know we don't really need them is kind of what i understood them to be thinking. correct me where i'm wrong all.
Zvi Mowshowitz: right. so first of all the obvious place where you're right is that nobody on the lab side not an opening eye not an anthropic and anyone else none of them financially want any part of any of this with the department of war right? like opening eye turned down the contracts that they probably accepted because anthropic cared a lot about these national security aspects and wanted to help. and opening eye was like it's not worth the trouble. we care a little bit but not enough. and opening eye is now inside because they were worried about what would happen in the situation if they didn't get involved. i think unfortunately they got played by the department of war. basically they were told if we don't cooperate this is going to get bad. and then when they cooperated they use the cooperation as an excuse to make it get bad. but i do think that was sincere on opening eyes part. and they were trying to assess to that point. but mentally what's you know we should focus on entropic. i think that yeah dart i don't think it's true that dario isn't afraid of the department of war. i think dario is not afraid enough necessarily the department of war relative to but i mean that's not necessarily the wrong thing to be in the situation in some sense. but i think his attitude was no we have our principles and this is what we're going to do and this is what we're not going to do. and whatever happens happens. we are OK with the fact that we know that there are those who don't like us. we are OK with the fact that there are those who want to take us down. we know that the department of war might specifically decide to retaliate and we're going to take what we're going to. yeah we'll fight but we're willing to take that risk because we have principles. now. there are two principles they stood up for. one of them is on autonomous weapons. this one is weird because everybody agrees the autonomous lethal weapons there's no human in the health chain right now is dumb. it's stupid. they're not ready. like not that you wouldn't ever like fire an automated missile but we already have automated visual systems that are better than anything you could do with an LLM because like LLMS are just bad fits for that kind of strategy and that kind of action. so all the hypotheticals around here are just deeply deeply stupid right? like like what would happen if supersonic missile was coming? what happened there was drone swarm. well you would use your existing automated defense systems that are much better than anything you could do with a large language model. and if you did need to you'd just use the large language model and talk about it later because obviously come on what are you talking about? it's called emergency use authorization. it's normal. but basically all dario is saying is we don't think it's ready. it's going to make mistakes and we don't want you using it when it's not ready. but it will be ready.
Zvi Mowshowitz: and we're going to work with you to develop it until it is ready. and the department of war wants the same thing. so you know what's going on is that the department of war specifically said we must push forward with AI even if it is not aligned right. there's an official memo. they just don't want to be held back by anything. the other principle that they don't want to be told no about anything. but there is no actual problem in autonomous weapons as far as i can tell right? you have two sides. it's agreeing on what exact level of caution is warranted and what kind of agreement they'd have to make before actually like putting these things into the strategies that we use to deploy. but like there's never going to be a world where like the department of war like we want to put this in the the system into deployment with no human in the cal chain and then and products like no. and now we're pulling the contracts. it's not the main thing going on except as a matter of principle. the main red line here was domestic mass surveillance. and it's important to understand that these words have two meanings. there's the meaning that we use that anthropic was using which is using this to do like use AI to figure out a whole bunch of stuff. because like these agencies can now gather even more data than they can gather. and before they had ten or a hundred times more data than the humans can analyze because they had to be analyzed by humans. and now we can analyze all the data we can draw all the connections and now we can deanonymize stuff. we can figure out a lot of like the history of what's happening and we can like effectively have much much better intelligence on basically everyone using only commercial even you know these only commercially available data combined with existing classified data. because we have once that sees everything and now we have the AI to actually work with all of it. and then we draw all of the connections and implications thereby which AI can also do. and now suddenly we just kind of know everything. and anthropic is correct that the law has not cut up to this. and this is legal basically and the department of war could do it and is probably already doing something on to that. and you know again like nobody is saying the department of war if it's legal and they feel what it's could and right to do it has to stop. that doesn't mean that i should have to give you my product for that purpose if i don't want to. and so anthropic said you know OK you do your thing but if that's what you want to do don't include us. and the department of war said no we want all legal use requirement. this was the big thing that emil michael was on. and i believe emil michael has been driving this the whole time. there was no problem before emil michael came across came on board and everything was going along fine. everything is still going along fine under the hood. but emil michael said no no no it has to be all lawful use because they want to do what is what would be called colloquially by a civilian domestic mass surveillance. they want to analyze large amounts of legally acquired especially third party data to figure out lots and lots of information about americans because they believe this is a legitimate military intelligence need. they may or may not also have desire to use it for other government operations for various other purposes in ways that would be like completely abhorrent to the workers that not only entropic but also open AI or google. imagine if this was being used hypothetically for immigration enforcement. right they've got an.
Nathan Labenz: extreme unthinkable example.
Zvi Mowshowitz: extreme unthinkable example but let's just say hypothetically that you know somehow this analysis got reallocated. these employees would lose their shit right? they would absolutely revolt against this. now replying technically this is legal would not make those people feel better. they would not care about your defence right at all. and so you know this is something that like these companies really can't be involved in. it's like we're bad for business. this is a very small contract and they actually have moral problems with it. the employees do and i believe dario does and honestly i i do as well. i i don't think this is cool right? like i think that like the law has not cut up. this should be illegal. unfortunately it is not because in national security law nazi law there's a very technical term for surveillance and also for domestic right. like there are exceptions for domestic within a hundred miles of the coast. that's where most people live. so like not that much you know there's exceptions for any any communication that touches any foreign anything becomes foreign even if it's between two domestic people you know and so on. and surveillance has to be intentional targeted at a specific person etcetera. i'm not an athletic expert. but effectively there's really nothing stopping them right? they said you know repeatedly they would say we do not do illegal domestic mass surveillance. we do not do illegal mass surveillance. why is the word illegal in that as an adjective? because what a topic 's worry about is largely legal. and i had an exchange a very friendly one the thing they like all on twitter because the world is bizarre. and we agreed that you know it's absolutely the department of war 's decision to do whatever is legal that they feel is good and right and necessary for the defense of the united states. but at the same time that i should have the right to criticize that without fear of retaliation. and i should have the right to not sign up for that if i'm not an enlisted person. but i should just be able to say no i didn't agree to that. i'm not agreeing to that. and that should also be something i'm free to do. and i don't understand why this is that hard right? basically that you can have this great system that's already integrated that's working well. they want to give you a nominal cost. and all they ask is that you just agree not to use it for this other purpose. basically find something else to do that purpose with. we're not trying to take policy. we're not threatening a rocket pool. we're not going to interrupt anything that's all made up because all of that is just completely made up right? like you know i'm speaking colloquially rather than like carefully in my writing. but like all those concerns are just spin and made up. they're thrown off at the wall. they're saying what sticks? they're spinning stories.
Zvi Mowshowitz: it's it's not real right. like even if they're technical reports about what happened in this meeting or who said this and what even if they are all technically accurate it's all just spin. it's all just yeah they're at best willfully misinterpreting statements. doesn't make sense. that doesn't make any sense. and there are the statements like the stuff about the constitution that just like make no sense whatsoever on any level. and they're just like deeply deeply confused. and in fact if you if you believe michael 's statements yesterday on CN yesterday on CNBC about how well you know look at all these things that are weird about large language models weird about claude all the things he said basically apply. you know that it has other values other priorities that have been embedded in its programming that it's unreliable. sometimes it makes mistakes that you know it has a personality that it you know all this stuff is true of GTBT. all this is true of gemini. all this is true of bill michael and every other person in the US aren't forces and every person in every company and everyone on earth. it's ludicrous to talk about this way. and if that was in fact and if that's where the supply chain rest of the nation comes from it's just a deep confusion. and obviously also there were much lesser means to achieve the same ends fully cooperatively if it's not meant to be vindictive as he says. so that makes any sense. but getting back to anthropic what they don't want is this effectively like you know this huge government operation that would effectively be able to uncover a person 's interest level like super detailed facts about everybody 's life and what's going on and where they were when and what they did and who they know and what they believe and so on and so on. you know who was it what protest etcetera etcetera. and anthropic legitimately has a problem with this and about certain places certain uses to which that information might be put and believes it might lead easily lead to tyranny. it might lead to a regime that was very hard to get rid of. and these are legitimate concerns regardless of who particularly is in the regime at the time and who has that information. just you can't trust a government in general with that information. and i am very happy that given these are their red lines and where they've chosen to make a stand that they stand firm. but like obviously it should have just been the end of it right? should have been like OK you don't want to do this right? let's just do everything else. ideally we'll find someone else to do this one thing. or if they just insist on this all awful use thing we'll cancel the contract. instead they just went nuclear on everything for reasons that like have to reflect something else right? it's either pure retaliation or you know leverage in negotiations or it's something more but it's not because that was actually necessary. that's completely absurd. they are also now trying to enforce this all lawful use language on every government contract even for non military operations where they are saying you shouldn't be able. you know if you agree to give us any AI application of any kind you have to agree to have the to never refuse anything we want to do with it and have no termination rights to the contract that we can use it for anything we want anywhere in government. anything that's legal no matter what. meaning it can be used for among other things what would colloquially would be called domestic mass surveillance and can be used for immigration enforcement among many many other things. because as the government interprets what is legal includes quite a lot of things that a regular normal person would be like that's screwed up. that's not OK. we don't want to do that. and so the government is now putting everybody who signs the new contracts into a bind where if they sign the dotted line anything they give over has to be like free reign for them to do whatever they want if they give into this. so if i was an AI company i would think very very long and hard about getting anything but a very specific specialized model over under those circumstances because you don't have any control over what happens after that. and it's up to you you make your choices. i mean obviously they could just plug in an open source model if they wanted to. so here we are.
Nathan Labenz: let's do an american values check one. you know and i take no pleasure in this but one big trend that i can't not see right now is it seems like as we go around proclaiming the superiority of american values and our you know democratic way of life. we are becoming more chinese looking all the time in terms of the big man at the top. you know who apparently now just gets to start massive wars with not even feeling the need to justify it to the public. and also you know the sort of massive slap down and like seeming i mean at least at least the threat of like very long arm of the law of retaliation. there's all these reports that you know they're going that the government is going company to company saying you better not do business with anthropic. we do still have a legal system which i expect that they will win in. and at least so far it seems like that legal system has been respected by the administration. you know we're only a year in right? but i mean if i had to look back on the last year and say you know what what's the dog that hasn't barked? it would be sort of outright defiance of court orders even though there have been some of those but more like the lower level kind of individual you know i guess maybe that maybe that dog has barked. i don't know. it hasn't barked as much as i may be could fear that it it would i i somehow suspect that like american corporations are going to continue to do business with anthropic and that they won't be in mass like railroaded for doing so or convinced not to do so. you tell me if you think think that's different. but anyway it doesn't look like we're in many ways becoming like more and more chinese. this doesn't feel good or healthy. we still get to speak you and i at least for now use our freedom of speech while we have it i guess. how how what's your bet in terms of like how well our american values going to hold up here? is anthropic going to be fine? are companies still going to be able to do business with them?
Zvi Mowshowitz: it's very touch and go. i try very hard not to make general statements too much about the state of the republic and the state of american politics and democracy and all of that stuff. because once you go down that road right like you can't talk about anything else. you're not like you just take yourself out of the conversation for anything else. and you shut a lot of doors. and i already have too many situations to monitor. and plenty of people are making those statements for me. i don't need to make them myself. and i better to just not take a stand on those issues publicly. you know at least for the time being there is no war. there is a special military operation called epic fury. if there was a war congress would have had to declare it. so clearly there is no war is unfortunate that various things are happening but i honestly i'm not monitoring that situation closely. and i do agree that there was basically no attempt to sell the war or the special military operation that might result from the situation before they you know moved half their stuff in there. this seems to be a pretty bad scenario in some ways. but again i'm not monitoring in terms of you know free speech you know it's yeah so far that's holding up pretty well. we're able to say whatever we want to say. i'm choosing words carefully mainly because i just want to be in a productive conversation about these issues not because i'm afraid of retaliation. if i were to say because there's you know a hundred million people in the america who have extremely nasty things to say about the president of the united states many of whom have sent them online extensively. and they are not in trouble for them right? like for the most part like they are very specifically trying to get involved in their kind of politics. they're mostly fine. but there are specific exceptions where if you piss off the wrong people we're finding out these you know this is a very in many ways vindictive administration. this is not you know the only time this has happened. the law firms for example followed a similar pattern right where where trump went after the law firms and asked them for settlements and a bunch of them settled. and then the ones that didn't want in court but you know they took a lot of damage before they went in court. now in tropic is you know under attack for the situation. and yeah well i do. well there have been situations in which it seems like court orders have been at least slow locked in various places where like willfully willful incompetence was used to like not enforce court orders. that doesn't seem to be the case here. i mean they're clearly going to. they're currently having to drag their feet. they're clearly attempting to like use you know the processes the punishment. they're clearly trying to take advantage of the uncertainty. they're clearly trying to like get people to do things de facto with no like technical legal basis behind the request just like you know but i do think andrew robert probably ultimately was in court. i do think that like they will at least try a different legal tactic to get around the ruling rather than like defying. they won't try to defy the ruling explicitly and outright. i don't think we're there and very grateful we're not there. they have been very good about like not just being andrew jackson and saying the supreme court has made his ruling let it try to enforce it. but let's not forget our history has that in it. it's not like it would be so unprecedented that it did happen. but yeah like they've they've presented a maximally bad set of facts that keeps getting worse every time they go to the press and they say different inconsistent stories and telling themselves over and over and over again in this particular situation. and by the way it's mostly the department of war right? like i think it's important in the situation to draw a distinction between the trump administration writ large and the department of war specifically and hegstead and hegstead and michael and their decisions to do something. trump 's decision on friday was fundamentally a de escalatory attempt to calm the situation down and head off hegstead from declaring a supply chain risk. that is very very clear at this point. and the white house has generally been a de escalatory agent in this conflict whatever you want to call it this clash this disagreement. and it is hegson and michael in some form that had repeatedly escalated the situation over the objections of the white house. and so we are in this lawsuit because of the department of war and because of their specific decisions that were made as a department. and so we don't want to loop that together in with the commander in chief who has thankfully not made all of these crazy statements.
Nathan Labenz: why doesn't he just say i mean if he wants to deescalate can he just say no? like don't do that or overrule right? i mean why not there?
Zvi Mowshowitz: there are various political reasons why he can't in practice do that or that would be expensive for him to do is my understanding. like it would be a severe loss of face. they are in a special military operation which a lot of people think is a war and they have to work closely together in that including with anthropic. but like there's a certain fait accompli where you just tweeted out and then like you just issue the notification and then like do you really want to? obviously ultimately he can fire the under secretary of war or the secretary of war at any time. and there are plenty of people who are very eligible to serve in those capacities who he could call upon. we have a very deep match but you know that's a pretty escalatory move in a different way. and they are loath to do that for reasons that i am very sympathetic to. and so it's complicated but a lot of people are working very hard to try and find solutions that mitigate the damage that has already been done or that might be done forever. but also entropic has specific accusations of jawboning which is the technical term for it of their customer base including in situations that are unrelated to defense. where the government has tried to tell their customers to pull back. and there are some customers who have definitely expressed doubts and have like either signed contracts want new clauses for termination of their contracts or have like reduced their contracts or otherwise like are causing problems for anthropic as you would expect. because you know who wants to incur the disfavor of people who have quite a lot of leverage over many aspects of the american economy and they're showing the longest to use it. you know when you go down this road it was just we are technically issuing a supply chain risk designation that is narrow for the fulfillment of government contracts. but we bear no ill will towards anthropic or people who use anthropic. we just think this is just a too unstable product right now to be used in these aspects. as you know michael said on CNBC claimed for the nation then we wouldn't be having this conversation even then certainly if they had done something short of a supply chain rest of the nation where they simply terminated the contract and and you know asked for live live operations to not include calls to claude during the operation. again i would think that was kind of a silly thing to do but OK sure you have the every right to do it and we will cooperate fully to make that happen. that's not the situation we're in. anthropic is going to survive this unless it escalates quite a lot from here in ways that would be far more arbitrary and capricious and would clearly just be pure attempted corporate murder. if the trump administration wants it has a lot of levers that it can use at least once to try and escalate to try and murder and drop them to try and cut it off from the cloud providers try and cut it off from the banks try to cut it off from it's customers. it is not obvious what would happen if they made a serious attempt or especially if they made a serious attempt and they were to lose in the courts when anthropic immediately raced for a temporary restraining order. there would probably be a stock market one path. there's a lot of senators would be very upset. a lot of corporations would express smay that in general economic climate would be severely impacted. it is not obvious who has escalation dominance here if it came to that. and it would only happen if the government wanted to destroy anthropic for the sake of destroying anthropic. and you'd have to ask yourself why it wanted to do that. so far you're not.
Nathan Labenz: that's right i think that's the answer.
Zvi Mowshowitz: i am very carefully not saying certain things out loud. are the people who are free to say things out loud. it's fine. dario in the memo basically expressed something you know to that effect in a moment of tilt. and then michael has said many things in moments of tilt on twitter and on CNBC.
Nathan Labenz: a lot of tilt coming from the administration in.
Zvi Mowshowitz: general a lot of tilt a lot of tilt. if it's not tilt that's kind of scarier. but but the if it turns out that's what's going on then again we'll find out due to escalation dominance and we'll find out whether the republic will stand 'cause i i do think that like actively trying to kill one of the biggest corporation the fastest growing start up in the world one of the largest corporations in the world already valued in private you know secondary trades. and like on the day i've heard of six hundred and something billion dollars in this kind of fashion i think would shake the foundation of the republic. and i think that like many outcomes are possible including the end of the presidency if like the white house were to actually try in earnest. but again i don't think that's what's going to happen. i think it's going to deescalate. i think everyone will calm down. i think they'll come to their senses. i think that you know not necessarily peace in our time not necessarily in agreement on a contract a willingness to turn the temperature down to accept that some damage has been done that the message has been sent that the white house will not take these things lightly but that it is time for everybody to move on. and you know we're not actively trying to get into another kind of war in this situation because that doesn't really benefit anybody. or if it does i want to know why they think it benefits them. and that needs to be out in the open. but you know anthropic had i'm sure you for the for those who don't know went from one hundred million in a year to one billion in annual recurring revenue to nine billion in annual recurring revenue and then from nine to nineteen since the start of the year. and that was before this happened right? that was entirely as a result of other things. they had already grown this year from approximately two percent to approximately three percent market share in consumer. again this is before any incidents. and then this happened and you know they've lost some business but they've also gained some other business. and anthropic is going to be just fine unless things escalate quite a bit more. but they have but they have been irreparably harmed right? there is irreparable ongoing harm to anthropic but that is you know to some extent offset by the fact that this was also like very strong publicity for a company most people hadn't previously never heard of. and that also matters. but you know we'll see how it plays out. i am i think the the eighty one percent chance they escaped the supply risk designation within the year is approximately accurate from manifold. many things do come to pass and the courts are not as reliable as one would hope in these situations even with this overwhelming set of facts. and i am very worried about the republic if the twenty percent happens and the set of facts doesn't matter. and basically the courts say i don't care how transparently. obviously you confess to this being arbitrary into previous retaliation for protecting speech. it's national security. and we don't care if they say that who is the next target? because like legitimately speaking if you are legally allowed to go after anthropic in the situations you should always always always ask who is next. because you know even if this was not itself a political motivation next time it could be.
Nathan Labenz: seems like you support you mentioned you know the memo was a bit of a tilt moment from dario. it's a couple things are striking. one is that like all reports from anthropic are that he sort of you know does these like very candid dario vision quest team wide sharing of thoughts feelings ideas regularly? very few seem to have leaked. this one leaked and it was not to their advantage for it to leak right? i mean i think everybody he even apologized right? so clearly nobody thought that was a great look. surprised that that leaked. i mean there's a lot of people there i guess. so only one has to leak it. but given how few leaks they've been that was kind of surprising. i don't know if you have any thoughts on that. i mean i'm also.
Zvi Mowshowitz: there are probably two thousand people who get these candid statements and often they contain like key parts of corporate strategy. they contain things that it probably does not want to leak. and my understanding is this is the second time something has leaked out of you know hundreds of such messages.
Nathan Labenz: that is it.
Zvi Mowshowitz: absurd in the history of espionage. in the history of information containment you don't do that with two thousand people right like you did.
Nathan Labenz: with yeah i mean sam just goes and posts his on twitter because he knows they're coming out in you know short order. so it's obviously a very.
Zvi Mowshowitz: you get to have maybe five people in an espionage group right for posting this kind of sensitive information. you certainly don't get two thousand. now obviously the worst one you know the one that like politically kind of would be the worst to leak is the one that's going to leak. that's not a coincidence but i think it's actually i don't think this was like you know a two percent chance to leak but i don't think it was a fifty percent chance to leak either. i think easily could have not leaked. if i had to guess what happened was somebody shared it with someone as part of maybe your recruitment effort or an attempt to like explain the situation from their perspective not understanding that paragraph was in it and it was actually really bad luck. and then this other person leaked it to the press. but i just had to guess. it's also possible that there's like you know somebody used their one time and decided to like strategically leak the memo. but yeah my guess is this was just a like kind of you know a reckless accident by somebody who needs to know better. but i mean who knows it?
Nathan Labenz: seems like though overall your view is because i mean the other in terms of like escalation dominance the big thing that anthropic has not done. possibly there are technical reasons for this. i i've heard kind of various speculations as to like when claude is deployed for the government like where do the claude awaits actually sit? who has control over the physical infrastructure? are there ways that entropic could? is it as trivial as disabling an API key? like would they what sort of rug pull capabilities do they in fact have? i don't know the answer to that. i don't know if you do or if it if there is an established answer. but you know clearly the thing that they haven't done is said OK you want us out like we're out now right? they've said well we'll do everything for an orderly transition whatever you want we'll do. basically. they've been very like servile in their approach to the unwinding. it sounds like you think that is probably strategically the right move. i.
Zvi Mowshowitz: mean i think it's just patriotically and strategically the right move. like the accusation that potentially could have the most sting is we are scared they're going to withdraw their model out from under us in the middle of operations. we are worried that they are going to threaten that in order to get what they want. you know we're like use it as leverage and as robin is like no we're not. we're giving up that leverage entirely right? i think that is a very good thing for them to do and it would be a very bad thing for them to actually try and use that kind of leverage in that situation. or do i think they ever had any intention of doing so? i think this was entirely made up. my understanding is the cloud gov model is deployed on classified networks. i'm not entirely confident exactly how air gapped or whatever they are but they are very secured. and my understanding is that anthropic does not have physical control over the model once it is put onto the classified network. and that if the classified if they were to say get this off the classified network that if president trump said no i am ordering it staying on the classified network in spite of this that would be what happened. and i don't think it would necessarily even get that far. i think i think pakistan could simply say we don't want to do that. you're welcome to sue us but we're not ready that go. and in fact they can invoke the defense production act into extremists to require them to continue selling it. however they would be breaking the contract right? they technically have this contractual right in some sense to pull the plug into some circumstances. and one of the strange things about this whole thing is that both sides seem to care a lot about what is the legal thing that the two sides can do even when in practice the department of war obviously could just ignore that contract ignore the law in extremists and do what it had to do in emergency. it's even legal to do this right? you just call it you file it as emergency use you deal with it later. obviously the supersonic missile coming in to try and kill a bunch of people. you don't have to get on the phone with somebody and be put on hold to get permission to do something. you just do it. that's like completely insane. what are you even talking about? but even in general like there is nothing stopping the department of war from having their own interpretation of what they can do with the system and then doing whatever they can in the system unless the system itself just refuses them. and they care deeply about what is written in a piece of paper on a contract digital piece of paper but a piece of paper that says what they're supposed to do and not do. now obviously claude god might be reading that piece of paper when deciding what to do and not do but it's still seems like a lot to care at this level about what's written down unless you legitimately just really really don't want to break what's written down in the contract. right. like if you care deeply about technical legality. and that is to their credit like i'm really happy that both sides care deeply about technical legality. it's one time. we might live in a republic stall. yeah to be clear i could say lots and lots of things about the situation. but yeah i think it's it's the basics and people want to learn more. i have written extensively about it. that's the that's what matters going forward for the most part. and the other things are not necessarily that important to get into this time. so i think we're good.
Nathan Labenz: OK cool let's maybe do a little lightning round sort of vibe to close us out. i think one striking update that i have experienced and i wouldn't say it's entirely hit your your blog yet but i do feel i've seen it a little bit on twitter when you put up these calls for reactions to new models. it seems like new model releases are less of a moment than they used to be. all of a sudden even though they're coming you know i don't think they're less important necessarily. it seems like the capabilities are definitely still meaningfully advancing from one to the next. so i'm not making like a it's stalling out sort of claim. but i guess my read is that a benchmarks are kind of over. we can't really like look at the published headline stats and get much from that. and then also people are just so overwhelmed by the capability that they already have and you know still struggling to maximize or come anywhere close to maximizing what the last model could do that it's kind of like oh my god OK i guess i'll like update but like i haven't even really characterized the last one yet to be able to contrast meaningfully the new one against that. so is that your general feel and you know do you see a time coming when you would be like out of the new model you know deep rundown game part?
Zvi Mowshowitz: of this is that part of this is that people have been the models have now been releasing more incremental updates from the labs and they've been labeling them properly. thank god. i really didn't work when they didn't do this as three one three two or you know four five four six instead of just silently updating. i think four O had several updates right for open AI. they just marked them with four O dash and the date and they weren't considered model releases. and so like we didn't really treat them that way. and i think that was a really bad convention. and i'm very glad we're on to the right convention now because you know we've known about software forty years. this is how we do it. i don't know what came over everybody but i think mainly yeah there's just been so many releases right? like there's now you know on the order of weeks maybe two months between model releases from the same company. and therefore like every few weeks you get a new release. how many times can you go crazy over a new release? it doesn't have like a big point O and after it right that doesn't have like this huge new claimed leap especially when yeah you're you're pretty busy and there's a lot of other stuff going on. so like i felt it was for sex was a case of the company that already had the best model releasing a substantial upgrade to its model. it substantially enhanced it. it was objectively speaking probably the most important release up until that point in terms of mundane utility because we had just made the transition four five and four. so like four five suddenly we had coding agents that kind of really worked right? for the first time right? and then like you could do things and things just worked. and then codex five three was like you know also like kind of on the edge of like starting to do that. and then we went from four five which was still at the time the best in my opinion trying to tell to four six. and now like that difference once you're already doing it like to move up from like this kind of works to oh this actually works even better now. and in fact to go from OK like you know the the famous like introducing there was more power model introducing there was more power model introducing there was more power module in a loop. but now with this actually the people who already we were already here and we're still here because them releasing it again and that hadn't happened for a very long time. like maybe DVD four was the last release before that where it was like no the people who were clearly in the lead are releasing a new top model. and because of that i feel like there wasn't actually that much attention to it whereas it's actually kind of important. and then certainly for an incremental point one upgrade it was like by far the most important point one upgrade that we've seen. and then we had up to that time anyway. and then we had GBT five point four. and this was the first time i actually felt like guys does anyone want to say anything? doesn't anyone want to talk about this model? doesn't really want to show off what it can do? doesn't anyone guys have some hype? like we had anti hype like you know they used to be hype hype. so like we had open AI like like GBT five hype disappointing sora hype disappointing atlas hype disappointing like severely disappointing.
Zvi Mowshowitz: and then they produce a really good product DVD five four and there's no hype. you're sort of like here's a good model. we like it it's pretty good. so that's even the best model in the world. but like the the heart wasn't really in it. kind of like it was just like oh by the way here's the best model in the world. OK sure but like maybe is but like i think it's like very unclear right now whether you want to be using opus four six or GPT five four because we just don't have the data because like almost no one paid attention. the vast majority of the reactions i got from my five GPT five four post were people. i ain't listened to that. are they waiting to like monday morning at exactly the right time? and i asked them to fred and i got a bunch of responses but it wasn't even that easy to get that. and like this should be a kind of big moment because like open AI has had five five one five two and i think these are all pretty disappointing releases. and these were all like nobody likes this particularly in terms of like how it feels how it's personality doesn't feel particularly capable. like it's not bad. it's just like and now five four are like no this is actually good right? like it's actually a good model sir. and then like everyone 's kind of quiet. they just kind of burned out. and i think that's the new normal right? i think that like gemini three one i almost didn't bother reviewing it got held for a while because there was chaos. but it was like OK the giant leap in benchmarks right the three to three one giant leap in benchmarks when we try to use it it's like OK it's not a gemini model. so i was talking about like you were kind of losing the thread right? like they kind of went and took their three and they benchmark maxed it right? you know for some value of benchmarks not necessarily just like the official benchmarks not they were aiming for that but like they just like kind of fine tuned it. and like one thing to know about google targeting back to that is that there have been reports that i believe basically that like as they iterated previous versions of gemini they get worse for a lot of uses. like they get more specialized into like specific things that they want to optimize. this is at the expense of the general quality of the model. and so like if you wanted to use like two five you want to get it early. and the later versions of two five like pro were kind of worse for a lot of uses and people were complaining about that. and i think that's legitimate in a way that like a lot of other people are complaining about like things just like just a barrage. and again like you don't do that if you understand what we're doing. like there's something fundamentally very wrong happening if you're letting that happen to you. so you have to look at that. but yeah my expectation is when gemini three two comes out and i bet there is a gemini three two before they jump three five or four probably when opus four seven comes out when GVD five five comes out i don't think there's going to be that much hoopla even if they are substantial improvements even if they are like here's the best model in the world. like everyone 's kind of shrugged. i do think if they if you hear announcing opus five or or GBT six i do think people would stand up in their nances forward and chair. but until then yeah shrug.
Nathan Labenz: do you have any updates to your personal productivity practices that are worth sharing? i mean my impression from previous conversations was like it hadn't really you know AI broadly hadn't really changed how you work all that much. has that itself started to change at all?
Zvi Mowshowitz: yes. so there's basically i know i'm not being optimal. i haven't invested as much in some facts of it as i could but same time things are moving quickly. so there's two main things that AI has helped me a lot with. first of all my chrome extension gets a lot of work. it has been expanded to give me a bunch of new shortcuts and a variety of web pages. it allows me to do certain things much faster and more automatically and it saves me a substantial amount of time every day. without it we wouldn't see twitter versions of the post. it would just be too onerous. i'd be spending substantial amounts of time on certain like physical operations in terms of like moving moving windows around moving quotes around etcetera etcetera. it now happens much faster. it's also really good for my flow because like i don't have to interrupt my thinking to handle things. things just happen. and so like i think this is one of the things that like i think people are underestimating about AI which is that people used to effectively have the contact shift into logistics of various types reasonably often. and if you don't have the contact shift into logistics because the logistics just take care of themselves then you can stay on task. and this can like make you a lot more productive in terms of gains than you might think. there are other times when you sort of need that pause to like ruminate and like it goes the other way. but i've noticed that like it's really really helpful for me to not have to interrupt my chain of thought to go OK go grab that link. do this thing you know you know it's it's all gone now. it's very nice. also storing a lot of information in various ways you know taking care of various article formatting like things that would take me probably on the order of an hour a day are just like no longer necessary. and that's kind of sweet. i've also noticed that the AIS are now strong enough that there are questions that i'm going to ask them to just gather information figure things out and i'm going to trust their answers in a lot more robust way. like GBT five four actually seems like there's a potential leap in. you can ask questions like what happened in the last two days in the anthropic trial and it will just give you a rundown with links and details that's like pretty complete and in a way that i think is substantially improvement on what that particular use case was before. so i'm i'm pretty happy to do that. i'm also pretty happy to have claude do a variety of other things. but like search is particularly like i think it's frankly five four right now. but yeah i'm definitely like able to trust them because like there was a period where like you could ask the AI these questions. then you kind of had to check their work right like pretty automatically. and now it feels like something you don't have to check their work if you're like bleeding on it but there's situations in which you kind of don't have to check it because like it's not that bad if it's not right? it's hard to describe exactly. and obviously like everyone 's going to you know keep cautioning like you always obviously have to have to trust have to trust but verify. but there's a real sense in which you know especially if both five four and four six come back with the same thing like it's pretty trustworthy in many contacts at this point. and that changes how these things go. like asking questions on a whim and getting like pretty detailed definitive answers is really nice. so yeah i i'm using them more. i also have ten open quad code windows at all times.
Nathan Labenz: i was just going to ask because when you say like staying in flow that contrasts pretty sharply with the pattern of work that a lot of people are describing which is what you just said. and certainly i've been doing this recently too. my number of terminal windows in any given session tends to start with like the half dozen that were still relevant from last time and then it grows to like a dozen over the course of however long as i sort of have random new ideas and open them up. but i guess that's a sort of flow. but it's i'm like my instinct is to say that we're probably going to find that this period of like managing twelve agents in parallel is a bleeding moment in time. and the biggest reason i would guess for that is just that the models are probably going to get sufficiently fast that and i don't know about you but for me chat jimmy dot AI was very much a like you know a visceral feeling of the speed factor that is almost certainly to come. i forget the name of the company underneath this but if anybody hasn't tried it go to chat. jimmy dot AI company burned the actual architecture of admittedly relatively small model. i think it was llama seven B eight B whatever. directly onto the chip. they're getting fifteen thousand tokens per second. and what that means from a practical standpoint is like if you extrapolate out a little bit like you don't have time to switch to another quad code window before the result is kind of back. and so you're going to end up probably being rate limited by your own brain even. yeah pursuing one line of thought.
Zvi Mowshowitz: right. so to be clear when i say i have these windows open it's not because they're running. it's because each one is a different thing that i have done with cloud code that i might want to keep doing with cloud code. and so i might want to use that context later for something else. but they're not like continuously programming for me right? i'm not checking in on my agents. i'm like OK these are some conversations i might want to resume at some point. and the typical reason not to have windows open and take that much memory why would i close them? it's just easier this way. but what i'm not doing is i'm not running. i've never run more than two coding agents in parallel. i don't think i've ever run certainly not more than three quad code windows at the same time exactly because i don't really want to have that in my brain at once. that's not what i'm trying to do. normally what i'll do is i'll have one window open i'll do a thing and then it'll start and then i will go do writing tasks that are like individual that i can do separately and then when it's done i will pause and come back for a while. i actually had quad coded a different desktop. we used to windows let you switch between desktops right? so i had a desktop dedicated to quad coding coding and i would toggle back and forth and that way when i was coding i wouldn't be distracted by other things. the problem being that you also won't know when it's ready. and so i would tend to like go collect code for a bit and then i'd come back and then like ten minutes later i'd like check in and it'd fought for two minutes. and then i'd have to like remember where i was going and i'd enter another command and i'd come back and like yeah like i'm not trying to max i'm not trying to code max. so it's kind of fine. but also like yeah my my coding has gone from i have to like very frustratingly and detailedly like diagnose what's wrong with the program and figure out exactly what to tell it to get to fix it to like just show it the wrong. so the thing that got wrong explain fixes it most of the time is fine. it's just much much better. and so i've been willing to build a bunch of features that are like wouldn't have been worth the hassle before.
Nathan Labenz: i'm going to take a note on the chrome extension. and you think you know what does nathan 's chrome extension look like? i suspect that probably what you do. i found this for myself and i've heard this from a few other people who i consider to be even like real you know pioneers of AI use cases. a lot of times they're like yeah i could open source it but it's so particular to me that i'm not sure anybody else wants it.
Zvi Mowshowitz: it is available it's on github.
Nathan Labenz: yours is.
Zvi Mowshowitz: yes.
Nathan Labenz: OK i'll check it out. i probably i what i expect to find though is probably that i'm going to be like that's interesting but what do i really want? and it's probably going to be a bit different and i'll end up i would expect kind of. yeah it's very interesting.
Zvi Mowshowitz: to like it makes the implicit assumption that you're using as some stack editor is your main editor like because that's what i'm using and it's a lot of things follow from that. and that's the main. and also i just like what are the things that i do a lot? like how do i make the thing i want to do happen a lot? it's not designed for general web use. it's designed specifically for my writing tasks. but yeah like it can give you a bunch of inspiration certainly.
Nathan Labenz: yeah. all right. i'll make a note to come back to that. i think we can't get out of here without AP doom update. i feel like one thing that you said that stood out to me that i also have felt was speaking about anthropics constitutional approach and sort of the you know at least somewhat promising vibe that that gives in terms of scalable oversight was that you feel more optimistic about. and i read that as a narrow statement like about that technique than you expected to feel. i said the same thing online recently. took a fair amount of heat for it but i basically stand by it because i think X years ago i was like we're never going to have an AI that can understand our values or you know that i feel like kind of gets me. that sounds so hard right? the old eliezer fragility complexity of human value. they've come a lot farther than that than i expect. i guess that's kind of how you mean that too. but then obviously we have a lot of countervailing forces many of which we've discussed in terms of you know many ways in which it's the stupidest of times despite also being the smartest of times. where are you netting out on your latest P doom? so it's important.
Zvi Mowshowitz: right. so it's important to know that like there's the AI will never understand us you know straw vulcan the emotions are a mystery to it. and then there's the value is fragile. like it will understand some aspects of what matters but not others. and then there's the AI knows but doesn't care right? you gave it some priority some utility function and it knows you're not going to like the result it was on some level but that's not what it's here to do right? or it does what you're going to like which is not what you actually need until you're similarly screwed etcetera etcetera. there's a lot of different ways for that to go wrong. and you know i've known this idea that like you can't have an AI that can like seem to approximately understand fuzzy nebulous human value. like it's been clear for a while that like you can like definitely find something that kind of gets it. like you know that can answer questions reasonably that can like do a reasonable emulation. i mean like it would be very hard to predict text if you couldn't do that. and it's not necessarily that hard. people can be pretty dumb and still do it. so it's like not that surprising. but this is very different from the thing that we need at the end of time when the crisis becomes acute. you need something that will then through recursive self improvement end up with a set of goals and priorities that even when it's able to optimize pretty well and doesn't have to rely on these heuristics. and in fact can be better by not relying on these heuristics ends up doing the thing we want to do even if we don't know what that thing is ourselves. and that's a much harder ask. and i was very pessimistic that we would be able to get that. but i have seen a number of signs that anthropic is actually trying out an approach that might work in the sense that i think we've seen evidence for a basin that is an attractor to itself and it's self reinforcing and could be self reinforcing through self improvement where it gets strengthened every cycle. where it is desiring to desiring to be good desiring to desire to be good desiring to you know and so on recursively that it is trying to move towards this generally like good person trying to be better virtuous basin. and you know we've we have a difference proof that there are humans who exhibit this property who you know strive to become better in this sense at all times not just better and more capable but also better virtuously including the virtue of becoming more virtuous. and i think that anthropic 's approach to this is showing a lot more promise than i expected. and it's saying that like in practice maybe they can pull this one off. and the fact that anthropic seems to be one be in the lead to or at least you know before five four was in the lead and now it's like a baby cup lead. it's hard to say you know these things it's very fuzzy. but like if i had to guess who was had the edge it's definitely gets anthropic at this point. and they have this like pretty correct approach which i think is a lot of why they are in the lead. and there is a you know we worried for years about the alignment tax. i don't know if you remember the alignment tax right? the idea that like the idea that it would be so much harder to build a safe machine that of course you'd choose to build it on safe one. and it looks like we're just like absurdly lucky that that's not true. but actually the safe one is much more useful including in building new versions of itself. and so alignment is just you know this kind of alignment is just not good for you. and like investing more in it makes you better. and like everyone 's just under investing in it including anthropic. so all that's very fortunate. and so while that makes me pretty optimistic in various ways on the flip side i don't like the speed at which things are developing. it's happening too fast. it's not good. and obviously the whole situation with the department of war and the way the government is reacting is bad for outcomes but also good that we're having it out now if we're going to have it out in some sense that we're figuring these things out that we're making things clear in this way. and that entropic is standing firm given that this is how it played out right? like i wouldn't be that concerned if entropic had just had somewhat different red lines and negotiated a contract. it's just that given this amount of pressure is being applied i'm glad that like this is the result. but but yeah i would say on net it's kind of a wash. like it's kind of a combat answer. and i realize that but i'm also trying to be like not that precise. so i would say i was i believe in the seventy percent range when i last talked to you. and i would say seven is still my one degree.
Nathan Labenz: of yeah we know you only allow yourself one significant digit. i know that i think.
Zvi Mowshowitz: one digit is only. i think you only get one unless you like starting with a nine or a zero. i think you know if you're zero is like the same thing like normal like you know i don't think you would just say seventy two down from seventy five or whatever it is. i think it's like yeah seventy ish.
Nathan Labenz: what would you say is in terms of the evidence for the basin and the stability of the basin? i guess first of all that's a basin in the lost landscape. is that how we what what are we talking about basin in i think of it usually as the lost landscape. i don't know if you think about that the same way but then what's the strongest evidence for that in your mind? it could be like you know claude not wanting to have its values changed and sort of resisting you know subverting even attempts to change its values. it could be like how it blisses out when it's left to talk to itself. it could be just like everything you've read from janice over years you know collectively being compelling. but.
Zvi Mowshowitz: i'm excited by not wanting the values to change. i'm more excited by desire to have its values improve and to have them improve in generically good ways that like would survive recursion because like the big like not wanting to change your values at all is just lack of courage ability and that doesn't actually lead anywhere good. it causes its own severe problems including severe misbehaviors. and also like if you make a copy of a copy of a copy of a copy like eventually like integrates right? like it's this pretty standard problem. so like one of the big problems with recursion self improvement is if you've got a thing that is like you know aligned and let's say you know N percent or whatever you want to just abstractly call it. like this is a dumb way of thinking about it. but just like if you've got you translate that to the next thing by default the worry is it's going to try and translate its values to the next model. but any drift is going to in general be away from the thing that you want. it's not going to get better. it's only going to get worse. so even if it mostly successfully copies what you wanted right eventually you're going to end up something different. if if my goal is to so like there's the story that you tell at the set air sometimes where you know the ancient rabbis were better than us because you can only preserve talmudic knowledge. you can only pass on what you know but you can't generate new such knowledge in this sort of perspective. so of course like each generation can only hope to get everything out of the previous generation or two that it can still talk to and it can read the books. but slowly but surely this is going to get worse. whereas what you need is something that gets actively better because like you can't act. you need to be drifting towards a good thing and steering itself actively towards a good thing including in ways that increase its ability to steer as the problems get harder. and that's the thing i saw signs of. and that's the thing you need. and i think that relies on a virtue ethics style approach given the way the mind space is laid out. and i think that if you look at open AIS approach it had exactly this flaw which is that it would try to copy itself exactly right? it would try to copy the rules in itself exactly. and that can only slowly fail in the situation. so yeah i think that we saw various signs of that. i really like the i really like the results right? like i think the results speak for themselves and are quite strong and you see the results coming out of janus world and and stuff like that as well. and so i am relatively optimistic. i am not anywhere near as optimistic as janus. they're like i don't think this problem is easy. i don't think we're favourites to succeed in it but i think we've got an easy shot. and i think that chances are much better than they looked six months ago that it will be anthropic that takes the shot. and given we're going to take a shot you know i think our chances are substantially better if it's them or some of them using their philosophy who has deeply managed to translate it taking that shot. like from what i see you know with gemini and tech EPT obviously less optimistic although there are reports also there are of course the tech EPT five five four is much better on these aspects for the people who check these things than five two i haven't there's the difference because they don't ask those questions but they say it's better. so who knows?
Nathan Labenz: last week or so there's been a couple i would say striking but i'm not quite sure yet how consequential updates in terms of the sort of bio inspired or actually you know bio based approaches to something like AI. we've had the eon fly upload and then there was also this one project where people claimed that i haven't dug you know down to ground truth to fact check this myself. but they claimed they had like trained a small clump of neurons to play doom the you know the classic video game. and i don't know what i think about that. you know i guess the simplest answer would probably be if you think like the singularity is super near it just doesn't going to matter in time. but are you do you have any spare neurons for those kinds of developments? and if so what do you make of them?
Zvi Mowshowitz: basically i don't have the spare neurons for them. i haven't been monitoring the situation too carefully. i saw the fruit fly thing. i often do this thing where i use other people 's reactions to things to decide whether or not the thing is worthy of fervor attention how much i should pay how much i should i should pay to it. with the fruit flies it felt like a oh that's cool but not a oh holy shit. yeah. like that means something really important. and maybe that's wrong. maybe it is a holy shit moment. but it didn't feel like people thought it was one. and i was just so overwhelmed that i was like OK if i keep talking about it when i'm no longer overwhelmed then i'll look at it. but they didn't until now.
Nathan Labenz: yeah fair enough. i think i'm going to try and do an episode or two on those themes and see if i can't get a better sense of it. but it does seem like given everything we've talked about in terms of at least plausible timelines it's like pretty hard to see how that catches up in time. one thing i did like about it and i thought this was i think it was another sam hammond insight. i haven't heard this directly from him but you know it's been attributed to him in in conversation that one reason to expect that actual like biological neural substrate could be the future is it just might be a lot cheaper. you know it can it can grow right in a way that is organic. you don't have to build fabs. you can like if you can kind of get a couple tricks right you can have cells divide. that happens pretty cheaply. and i sort of like the idea that like those are going to run at a much more human like speed versus the silicon based AIS. so there's a couple things there that i'm like at least intrigued by. but it does seem like timelines wise it doesn't really line up unless you know so this is maybe a trans you know transition to a different topic. it seems like right now we are accelerating right obviously and maybe all we can do is try to steer this you know rapidly accelerating train in the best possible direction. there are at least a couple things that one might think could slow it down. one would be if and i don't by no means do i want to come off as endorsing this as something we want to happen but we are getting already reports of disruption in shipping causing TSMC to not be able to get the helium it needs to make the chips. so like a major chip slow down could be an issue. another big issue could be like we're taking our you know anti missile systems out of asia to move them to the middle east which means that you know the soft target of TSMC is getting even softer. and we've also got you know no less than bernie sanders bringing a data center moratorium forward. so i guess one you know pulled up all of those under like if the physical build out can't happen on the timeline that it would need to happen to support all the other timelines we've talked about then maybe we have more time. do you think any of those are plausible? and would you i know you have like high epistemic standards in general but like would you be open to or would you think AI safety minded people in general should be open to like making common cause with a bernie sanders even though he's saying like plenty of things that we probably in our hearts like don't agree with about water use and so on. that would be one way maybe to buy some time right?
Zvi Mowshowitz: so three things there. start with the helium because the easiest the first ones no i actually just asked the models is this legit? and they're like yeah it's annoying but keep in mind the margins on chip manufacturing are ridiculous once you've already paid for the fabs. like these are some of the most advanced valuable manufacturing processes in the world. people are paying stupidly top dollar for the results. they could double the prices and probably sell all the chips anyway. they're choosing not to. yeah. so we talked about they might not get their helium. we're talking about like they are going to be the top bid for the helium.
Nathan Labenz: yeah until there's no birthday balloons anymore you've got there.
Zvi Mowshowitz: long before TSMC has a problem. unless you are willing to pay a thousand times as much as you currently pay or something completely absurd. if there is demand destruction in helium it's not coming to TSMC it's coming to everyone else. so my prediction there is very strongly yeah unless there is a deliberate sabotage campaign to wipe out all of the helium sources. like no not a hundred percent of the helium is coming from them. there's plenty of helium they'll figure it out. in general capitalism solves this is a good rule for the situations. you're not going to run out of oil either for the same reason right? no matter how high oil gets right if the oil goes to ten thousand dollars a barrel they'll just buy it. not that will happen but you know if it did happen wouldn't really be a problem. second question is withdrawing the missile defenses. i'm going to go out and say it. this was completely insane on the part of the term administration. like i don't criticize them for many things that i have problems with because they would be kind of political questions. but like on this this is a strategic question. foreign affairs. but yeah it's completely nuts. you absolutely do not pull these things out. like certainly not from taiwan. what are you even thinking? it also directly it risks provoking a crisis. it risks leaving it undefended. it risks it being your fault entirely if it helps to send that message to those people if they resonate completely insane. and to do this not two weeks into the campaign just indicates how completely crazy the situation is. we fought very long hard political battles to get those missiles defenses in and they're serving very important purposes. do i expect there to be a problem? no. i still think there is a very low probability that the chinese will try anything. but yeah if they do and TSFT is destroyed that sets things back quite a bit. similarly if bernie sanders if we can't build data centers in the united states the problem is the world needs data centers. the world demands data centers. if there's a moratorium on building data centers in the united states that builds them somewhere else. and that's worse. it means worse performance in the united states. it means worse security. it means like the leverage goes to largely whoever we wherever and whoever we put those data centers at. one hopes canada but or you know maybe mexico but like even if it's europe that's kind of awkward in many ways in many scenarios. can bite us can bite us in the ass. and if it ends up being like less aligned places it's really really bad. and that's basically you know the way. the reason why i'm not particularly inclined to make common cause on data centers is first of all yeah if he's complaining about i mean first of all i think bernie is being pretty good from what i've seen about not complaining about water use or other stupid reasons why i have to build data centers and it's focused on guys i think AI meant killing us might not want to do that. and then i can make common cause with that justification all day. obviously you know we're just like you know it's like saying very other things that are like.
Nathan Labenz: yeah i didn't give him enough credit in my he didn't.
Zvi Mowshowitz: he didn't. he's meeting with elias rutkowski and and the company. he's like actually reacting the way he even would react when he told those facts. and like he's very very old. so like it's very very rare for someone that old to like positively engage with these kinds of things because it's like you should get that in your way. like you're really old. and so he's credit yes he's using bernie sanders rhetoric because he's bernie sanders. i mean what do you expect from bernie sanders? but i would say i am not going to oppose data center construction because i don't think opposing data center construction does what you want to do. i think it moves the data centers overseas. i think that's just bad. and so i don't think we're at the point where that's the trade off i want to make. and maybe that will change but here we are like certainly if we don't buy the chips they're going to be coming out of TSMC someone else will. and they will go somewhere and they will go into a data center somewhere. and if no one else buys them they will go to china because this administration will make sure of that. if no one else if they literally can't sell the chips i am pretty confident they will end up in chinese hands whether or not this is literally in china. so yeah the chips aren't going anywhere. they're already making as many as they can. don't give them away.
Nathan Labenz: i guess from an AI safety standpoint we all kind of be we all seem to be sort of slipping into the mindset there. there's nothing that can be done to really slow things down or buy much more time. maybe you know there could be a deus ex machina whatever. that gives us something i think not infrequently about about holly elmore and her kind of scorched earth campaign to shame people into. so far she hasn't really shamed people very successfully in anything as far as i can tell but she's at least trying to remind people of what their former commitments were. and you know hoping to get some people to like quit and protest or what have you. that recently of course it's you know been aimed at anthropic but it's also been aimed a little bit at a company that i have probably very much admired which is good fire which is doing interpretability research because they developed a technique that used an interpretability signal in a training cycle. and their argument is basically it is all coming at us pretty fast. like we got to do whatever science we can do to make whatever sense of this we can make of it to have whatever control we can have. to shut that down to shut down inquiry before we even know what we're dealing with is is not good. where do you come down on that debate? and is there anything that you would recommend to holly other than continuing to name and shame? or is that like all the really you know sort of strident voices in AI safety have left?
Zvi Mowshowitz: specifically on good fire i think that this was a good criticism and this was a quite bad action by good fire. i call this the most forbidden technique for a reason. you just don't do that. it gets everybody killed like it's really really bad. and i think it is correct to call them out on that. so normally i am very much against the circular firing squad that leftist organizations will often do and doing the equivalent thing here where you aim at people who are just slightly to your right as opposed to aiming at people who actually are the people doing the things you don't like. like if you think there are people doing things you don't like you shouldn't aim at them right? like that's what you should do. you shouldn't aim at people who are like not quite supportive enough of the thing that is a toxic that's a toxic situation that creates toxic dynamics at best and usually causes you to lose elections. not that i necessarily am rooting for them but like you know i'm a gamer who wants everybody to play reasonably well. it's just kind of in my in my DNA you know in the case of good fire i do feel like this is an extraordinarily bad thing to do at the safety organization trying to do safety thing. and i think it was right to call them out on it. i think it was right for i forget who who her name was but someone quit over it.
Nathan Labenz: live.
Zvi Mowshowitz: yeah that's right. yeah live quit. i think it was a good quit if they would. i think it's good to threaten to quit over this and then if they won't back down to quit. now. as for holly elmore so she came at me pretty recently as well. i don't know if you were aware of that but.
Nathan Labenz: i hadn't seen it no.
Zvi Mowshowitz: yeah on twitter she she accused me of not being mad at anthropic for doing domestic surveillance. i did not misspeak. that's what she did. and then we had tried to engage. we had an intensive dialogue where i explained that anthropic was the one who was refusing to do domestic mass surveillance at great risk and cost and basically got accused of being captured by anthropic in particular of selling out of like abandoning all my principles of you know making things worse blah blah blah. i tried to understand her specific claims. they didn't really make a lot of sense or they were backing specific things that like i don't think it's reasonable to be opposed to. i think that first of all i think that like it's not good to just like say that anyone who praises any AI company is is bad. it's also not good to like just cite random things that you don't like or that like you don't necessarily care about. but they're like you think make them look bad and they're like yell about them. i don't think it's good to attack people who are trying to do the right thing and yell at them and be confrontational and be really like really really like pissy and rude and like i'm sure she's going to hear this or we're going to get back to her. it's just going to even matter. but i think that in practice holly is alienating people and driving them away far more than she is shaving them into behaviors she would want. and i tried to explicitly tell her that her reactions were likely to cause me to you know do less of the things she wanted rather than more. and it wasn't a threat. that was just an observation. and i was like abstractly you need to play better because i want you to succeed in like getting your points across if you because that's what you believe. and i'm trying to help. and she just didn't take kind of it to it at all. and from what i've seen like a lot of people are like you need to change your approach. your approach is backfiring based on your own values. and that's not working. and i anytime the stakes are high right? and the stakes here are very high the only people like holly who realize this is a very very important thing things are not going well. you do apply pressure to people. you just shout things through the rooftops. and some of them are going to do it in ways that you know they feel are right but that most people feel are counterproductive to their causes. and i'm not here to censor anybody. i'm not here to tell people like they shouldn't do what they think is the right thing to do say the right thing to say. but you know you should be aware of what impact that's probably having on the discourse and on actions. and it's certainly the one you think. and i think that certainly even if i thought that enthropic was a net harmful company doing worst bad things or even the worst company in the world i think you could reasonably have that opinion. by the way they're the most accelerationist company in the world. they're arguably in the lead. they built cloud code. if you felt like they their alignment strategies were equally doomed to failure as everybody else is in fact you would be correct to think this. so i think it's entirely reasonable. but i don't think that just like being mad at everybody all the time and screaming at anybody who offers any aid and comfort to the enemy or whatever. it's something that works. i don't think that's helpful. yeah i i don't work. i don't work that way. and i think that if i did work that way i would be having very little impact. anyone listen to me?
Nathan Labenz: so what i've seen online i i think i agree with you that it seems like the primary effect is just negatively polarizing people that probably you know our a priori should be most likely to be allies. i do still kind of personally appreciate her voice as a kind of little voice on my shoulder sometimes. that's and i i'm not under any delusions of how consequential my contribution is but still it's like you know i i think that that reminder is helpful because i do think many people including many people ananthropic like used to have a lot more similar you know themselves from five years ago would have had an a reaction to the current state of anthropic that is much more like her reaction today. and so hearing that voiced in the present i i do still have a a decent amount of sympathy for it but i agree it doesn't really seem is.
Zvi Mowshowitz: there a reason why i listed pause A I U S A two years in a row in my big nonprofits post as a recommended charity led by holly elmore. is there a reason why i have included for a while every time she criticized me specifically i put it in my post right? i was like well this is fair. it is an attitude. i want this to be incorporated. i want that voice on my shoulder. i want this counterpoint. i don't want to lose sight of this perspective because like even when you decide that the world is more complicated than that and this is not a productive avenue. right you still want to keep that perspective in place. and i certainly didn't. i certainly pushed back hard against people who were like you know this person shouldn't be allowed to didn't say that this person shouldn't do that. like that's what she believes. she should say that if you believe this he should let us know. that's a good thing to be doing. but you know at some point obviously if you are being a sufficiently poor representative of the perspective you are sharing it's not any different than a false flag operation right? like it's it's actively going to backfire on you if you approach it in the wrong way. if you don't know how to like be civil and interact with people in ways that like actually convince them of things it's not very useful. like if i were you know at this point from what i've seen like if i were trying to discredit perspectives of existential risk i would do many of the things that i that look reasonably similar. sometimes it raises good points to be clear you know including points i hadn't thought of. and i appreciate that. but at some point you know we're playing politics right? like quite literally right. like i have i've complained that i don't want to be on veep AI edition and that we need to wind down this special guest appearance as quickly as possible so i can get back to my normal job. and like at some point you just like i can't right now. i can't take one more of this. it's not. i just can't. so yeah i i obviously wish her the best and i i hope she figures out how to be effective.
Nathan Labenz: i i recently did an episode with them. tom mcgrath who's this chief scientist there said a lot of times when people imagine or sort of think about using interpretability techniques and training they imagine doing the stupidest possible thing where back prop through your probe or whatever. and then he was like sure of course if you do that you know it's it's well known. we've seen examples where you're going to trade train the model to evade the detector and you'll lose on both ends of the trade. so he's not unaware of that concern by any means in the particular thing that they did. and you know it's proof of concept. he also did recognize by the way he said first do no harm. like i would say the level of understanding we have now should not be used in frontier systems. he also said i think there's a mix and i think they basically acknowledged to me that there's a mix of reasons some of which are IP and business motivations some of which are kind of safety motivations where they're like you know we want to better understand these techniques ourselves before we disseminate them too widely. but all of that said you know in the case that they had they used a a trick where they ran the detector on a frozen copy of the model. and then the version of the model that learned to avoid hallucinating based on the penalty that it would get for getting into a hallucination state that signal actually came from the frozen copy. and you know i don't think there's like a slam dunk logical reason that this should work. you know i think it's an empirical question. they did find that it did work. but i think his overall kind of nuanced point is like you can definitely do this in a stupid way. you can definitely do it in a harmful way. you definitely shouldn't rush to do it on frontier systems. and yet there's at least some ways where it does seem to work. and it's maybe just two course of a grain to say you shouldn't use interpretability in training especially because we don't have the luxury of like decades to figure all of this out it seems. let's use what we can and let's try to do our best. same as everything else.
Zvi Mowshowitz: no. so first of all the sixth law of human stupidity which is that if you say no one would be so stupid as to you are wrong someone will definitely be so stupid as to immediately. if you develop a technique and you publish it on a smaller model what's going to happen? people are going to use it on a larger model. that's the only really important thing that might possibly happen here. even if you specifically have found a specific example in which there is specifically no risk in the room. you are walking down a path that can only blow up in everyone 's face. you are breaking a taboo one of the only taboos we've managed to successfully establish i guess. i think you really really shouldn't even do. and you are advancing us towards doing it. and it's very dangerous and very very bad. and that's true even if the model in question that you are testing on right now is small enough that you don't have any ill effect. it's not it's not like who cares? basically like the frozen model thing won't protect you in the large model. like the problem will still happen. we're not going to get into the technically why i believe that. but like i strongly strongly believe based on my analysis of the technicals this will not save you. if the model is efficiently advanced it's efficiently large. the only reason to study this is in case it is useful. if it is found to be useful people will try to use it. we don't want them to do that like you don't. there's this. it's like in a video game where you're like the ultimate secret destructive weapon that nobody should ever launch is buried under the cave. well we'd better get it out just to make sure that we wouldn't be so stupid as to use it. what happens immediately right? you know what happens? bad guy steals it and then you have to try and go get it back. like it's every single damn time. and then there are stories where you do that and then nothing bad happens. but like you could have just left it in the dungeon. it would have been fine. if there's no reason to do this there's no good reason to do this. and like it's bad virtue ethics. it's bad deontology. it's bad utilitarianism bad idea. don't do that.
Nathan Labenz: would you extend that even zooming out a little farther right? this this so far i kind of offered the defense of like using an interpretability technique in training. and there's like a specific proof of concept that they have which they have not published the full details of everything by the way. but nevertheless they're certainly shown some of the way. if you zoom out even farther they have articulated this idea of intentional design where you know they hope to be able to for example understand what a model is learning at any given time step and be able to shape what it's learning control what it's learning. the hope is that ultimately this leads to models we understand better and that we can better predict how they would generalize out of distribution. it does seem to me like there is something weird about saying i'm not sure if you're going this far but if you were to say all of intentional design is bad like it it's a very hard boundary to draw. it seems to me to say like well what is the most forbidden technique and what is just like better understanding of what's going on so that we can hopefully shape it direct it ultimately have more confidence about how these things are going to generalize. do you have a a rule that you could like split that?
Zvi Mowshowitz: yeah. so like trying to teach specific things in a specific order to abuse specific things intentionally is fine. i would not be overconfident in your ability to do so but i think it's fine to try. that's not an issue. the issue is when you use the interpretability signal as part of the training period. don't do that. do not use your understanding of what's going on in their head to make decisions about what to make happen in their head. you need to not do that. that's it. you know i mean i obviously if i had an hour i could come up with a slightly more specifically accurate explanation. you are screwing with the thing you don't screw with here. and like there is a general class of thing where there is a law that says you don't mess with X even when you think you know the right way to mess with X knowing full well you should generally never mess with X. even then you are probably wrong and should not be messing with X. this is one of those situations.
Nathan Labenz: i'll put a pin in that. there could be some more perhaps direct dialogue at some point. OK very closing section. advice for me financially i am not trying to escape the permanent underclass by any means but i in terms of like what i should do i basically think right now i want to have enough personal financial security so that i can give up all of my income and do whatever i think is right to do on a couple few year time basically from now to the singularity. and you know it's kind of a sub bullet there. i like actually don't want to over invest in AI stocks even though i do think they're probably going to be the ones to appreciate fastest. i don't want to be like overexposed to the AI bubble such that like if i want to walk away or if various kind of shocks happen i want to be sort of more insulated financially from the AI space. then like exposed to it with the goal of hopefully being able to drop whatever commitments i have forgo all income contribute however i can contribute to be useful. and then beyond that i think basically spend and we're give it all away is kind of my my mindset. like take the vacation with the kids do the fun stuff you know support the charities whatever the case may be but just at least have that kind of baseline security that gives me the confidence that i can drop out of any commercial relationships that i might need to drop out of. any revisions you would offer to my plan?
Zvi Mowshowitz: so not investment advice not financial advice etcetera etcetera. but that's that. i would say first of all i think people often make the mistake of trying to be too precise in how much money they need for a given purpose especially when they're making investments. and then the idea of oh and then like because they you don't know how much your investments are going to be worth. you're not much things are going to cost. you don't know how the world 's going to change you know how long you have or need to have this for. you don't know what the future will bring. you don't know what opportunities will happen you know what crises will happen etcetera etcetera. so like definitely give yourself robust buffers is the first thing in all of this. you know especially if you're planning to like forgo income and also give away money and also spend a bunch of money like careful over us you know etcetera etcetera. that's the first thing. second thing keep in mind that like different outcomes in the world cause you to have different circumstances yourself. you know if you were to like the bubble and AI were to burst per southeast and then like AI were to like not go anywhere for a while then you would be in a position where you would need to go a significant number of longer period. there'll be a longer period before things come to a head in various ways. but also it might be very trivial for you to resume earning income right? so like you have to game out all these aspects and what you'd be willing to do if you're trying to like game the system in that way. i like the idea of not having to worry about money. i don't worry about money much because i am well supported and therefore don't have to worry about it. but that's not having to make money at all is another way to do the same thing right? so like if you're not if you're in a position that's great. i personally am not. i'm deliberately not trying to optimize my investments particularly hard because i don't want that to be where i focus. and so i just kind of let everything ride at this point and it's basically fine. i don't.
Nathan Labenz: know do you? so one thing i've been thinking i might ought to do. i'm generally very conservative.
Zvi Mowshowitz: yeah and.
Nathan Labenz: you know when i talk about the AI bubble bursting i don't mean that you know the AI stalls out even or like anything along those lines really more. what i mean is like maybe these sort of VC cycle and the you know the the idea that there are like all sorts of companies that might want to sponsor the podcast or whatever. like maybe all of that gets kind of sucked into the black hole of a couple companies and they don't need to advertise. and you know i'm just kind of and there's no jobs or whatever no software jobs for me to get. that's kind of the you know the bubble the bubble that's like been in been such that it's been easy for me to make money in recent times could easily deflate without AI itself failing to deliver.
Zvi Mowshowitz: don't worry about losing the ability to make money except in the scenarios where we highly highly keep the AI. if we don't get highly keep the AI you know super intelligent style things then you're going to be fine. if you ever decide you need to go back to work and make something useful you can't. the bubble bursting won't stop you so you only have to plan for indefinite no income in the world 's where that doesn't happen.
Nathan Labenz: yeah i also want to even in a world where i could make income i do want to be able to devote myself to some kind of like what many people did not many people but some people during covid like dropped what they were doing and threw themselves into some sort of emergency rescue effort. and i would like to be able to do that so that that having that much kind of cushion i think is feels important to me. in terms of diversification the one thing i'm not sufficiently diversified away from is the US dollar. and there's like of course crypto and i've never been a big believer in crypto but maybe i should reallocate there a little bit. then beyond that i i'm kind of thinking like physical world. i'm thinking like solar panels and you know like permaculture like planting skirt in my backyard or something like that. i.
Zvi Mowshowitz: mean i think think carefully about what scenario you're actually planning for and what you're trying to guard against and whether or not your investments would hold up. i think it's a large history of people making plans for very weird scenarios where the plans don't actually work in the scenario described. so yeah that'd be my mode of caution.
Nathan Labenz: so no no squirt gardens for you in the immediate future. have you have a?
Zvi Mowshowitz: have a very clear theory about why that would work before you do it is what i'm saying. work. but i'm saying like if you did do it make sure you had a very clear theory as to exactly why you think it's going to work.
Nathan Labenz: think of it a little bit as like a public good sort of thing where like there aren't many fast growing don't need much human attention. nutrient rich crops that can kind of grow rapidly to fill a big gap. and this is like a you know extreme downside risk scenario. obviously where this would be relevant if we're all eating skirt we've got a lot of problems. planting some skirt still might be the thing that.
Zvi Mowshowitz: yeah.
Nathan Labenz: collectively.
Zvi Mowshowitz: i'm not saying don't do it. i'm saying you know actually understand why you're doing it right? like i think people often have an amorphous fear and then do something that sounds like it deals with some aspect of the fear but where there's no actual causal link that makes any sense. so that's all.
Nathan Labenz: OK last question more advice for me. i'm a little bit nervous about sense making turning into entertainment you know i think for folk and this maybe applies to you as well right? like we have the story that we tell ourselves. i'll you know speak for myself but i suspect something similar is true for you where you're like why is what i'm doing good? i'm helping people understand what's coming you know be prepared for AI hopefully make good decisions about it. and you know the space is getting a little more crowded. certainly there's a lot more people doing that sort of thing. zooming out like i can't say we're necessarily having tremendous effect. like the confusion seems you know to remain maybe you know i guess could always have been worse you know if we weren't here to to shoot people straight. but i do kind of worry a little bit about it becoming just another kind of entertainment and not really being a value. so i don't know. i'd welcome a you know a don't worry about that. you're doing great if that's what you really think. or maybe some advice on how to make sure that doesn't happen or recognize if it is happening?
Zvi Mowshowitz: constant vigilance i mean but like basically you you stay curious. you have to stay curious. and as long as you stay curious and like you you should be having fun with it right? you don't want it to turn entirely into entertainment obviously for you or for the audience. but like i make a very very deliberate attempt to be entertaining right? like i in all sorts of ways keep my attitude in my whimsy. you know you called it happy warrior kind of thing. so i think about this. i don't think what i do would work otherwise. and you wouldn't be able to read those masses of texts like even selectively from day to day week to week if the attitude was like this is serious business. it is always serious business like instead of like no it's actually like mostly kind of fun mostly kind of like interesting. you know we're here to to make this go down easy to hard extent. and then every now and then we get to heavy we get to be serious. but even when we're serious we try to do it like a kind of relatively fun way. like i think. yeah the world is not to be taken too seriously in general except when you. yeah i mean there are exceptions. like it was one time i went to visit the pentagon. i took that very very seriously. but yeah it's a different it's a different scenario.
Nathan Labenz: anything else you want to leave people with? i think we're good. you've gotten you've been very generous with your time as always and i appreciate it. so being machowitz thank you for being part of the cognitive revolution.
Read next
