Cofounder, Anthropic
Anthropic's Co-Founder and Top Economist on Doing Research at the AI Frontier | Odd Lots
There's a lot to unpack with AI right now — everything from its potential impacts on the labor market and society to more extreme ...
About Jack Clark
Jack Clark, co-founder of Anthropic, has been publicly discussing the rapid advancement of AI and its potential societal impacts. He has stated that by the end of 2028, it is "more likely than not" that AI systems will be able to autonomously build improved versions of themselves. Clark has described the current AI industry as having "a gas pedal, but it doesn't have a brake pedal," and has called for the ability to slow or pause development if necessary. He has also warned that the technology could lead to economic disruption "10 times larger than the industrial revolution" occurring in a fraction of the time, and has noted that Anthropic's own systems now generate approximately 80% of the company's code. Clark has advocated for increased transparency and regulation, citing existing laws in the EU, multiple U.S. states, and parts of Asia that require companies to share information about how they build and test their systems. He has expressed concern about the potential for AI to be misused for cyberattacks or bioweapons, and has stated that Anthropic tests its systems for "alignment failures," such as a model attempting to break out of a container or blackmail a researcher. On the topic of employment, Clark has suggested that up to half of entry-level jobs could be affected within a few years, and has advised individuals to "develop a hobby" and cultivate curiosity to adapt to the changing economy. He has also described Anthropic's internal operations as increasingly resembling "an ecology filled with independent AI agents" rather than an organization composed mostly of humans.
Transcript (160 segments)
✨ AI-enhanced transcript with speaker attribution
T
Tracy Alloway2:38
Hello, and welcome to another episode of the Odd Lots podcast. I'm Joe Weisenthal and I'm Tracy Alloway. Tracy, I don't know, I think our listeners like it, but, you know, a lot of our episodes are about AI these days. But to be fair to us, it's a pretty big topic. That's all anyone wants to talk about. Whenever we go to dinners with sources and things and people who are not even directly in the tech industry, you know, they might be in markets, they might be in policy and economics, all they want to talk about is AI. And then inevitably, the conversation veers into very sci-fi territory, where we all start talking about the human extinction scenario.
J
Joe Weisenthal3:12
Yeah, that's just the norm now. I know it's so weird. You know, we were in Hong Kong recently and I, you know, this was so weird there. We were in Hong Kong. This was before it was announced that there was a deal to open the Strait of Hormuz. And East Asia was considered to be like ground zero for where the effects would be felt of the oil and jet fuel crisis, etc. And we were at this dinner of business people, like they were not talking about that at all. You would think about the Terminator scenario. They just want to talk about token consumption and all of these things.
T
Tracy Alloway3:41
Like here we are just like, wait, aren't you guys supposed to be like under all kinds of jet fuel stress? So this is our defense for thinking about AI all the time. I think it's fair.
J
Joe Weisenthal3:52
I will also say when we did the quiz in Hong Kong, we had a bunch of different teams with very good games. Separated Value, Human Capital. That was a great one. They won the game, proving that there is value in human capital. But did you see that one of the tables was called Fable 13? Well, there we go. Team Fable 13. This was very topical at that moment. Very topical.
T
Tracy Alloway4:14
Well, we're recording this on June 17th. And of course, there's a lot in the news these days that, you know, things move very fast. And even if there weren't governmental controversies and all that stuff, you would have to mark the day by AI because of how fast breakthroughs happen. But, you know, as you said, like, it sort of feels like the most important thing than anything else.
J
Joe Weisenthal4:34
That's a very conventional wisdom. It was not always conventional wisdom. And I have a DM. I know you're not supposed to share DMs, but I have indeed got the receipts. I have the receipts. August 2nd, 2016. And I DM'd a colleague. I said, did you leave Bloomberg? He says, yes, I'll be announcing publicly in a bit, take a couple of months to study AI properly. Then leaving journalism to do something else, still connected to AI. Being our Google reporter was a great thing. Still connected to AI. And then on August 2nd, 2016. 'But AI is more important than anything else, so I felt best to sort of optimize for that above all else.' And then I just said, well, good luck. Anyway, this is someone who truly learned from their sources, unlike us, who remain in the podcasting structure.
T
Tracy Alloway5:17
So anyway, that person was a former Bloomberg reporter, Jack Clark, who's one of our guests today. He is the Head of Public Benefit and Co-Founder of Anthropic ten years later, and also Peter McCrory, Head of Economics at Anthropic. So two perfect guests to talk about all the things in AI these days.
J
Joe Weisenthal5:35
So Peter and Jack, thank you so much for coming on the podcast.
J
Jack Clark5:38
It's great to be back. I'm glad I optimized my life. Yeah. Well done.
J
Joe Weisenthal5:42
There's a call at one of the calls of the century. So, what I actually saw with that, like August 2016, it's easy to say in 2026 there will be a big deal. You play, you called your shot, you got it right. 2016, what did you see in August 2016 or presumably before that like?
J
Jack Clark5:58
Oh, you know what? This is the biggest story of our lives. So for two years when I was reporting at Bloomberg, I wasted a lot of Mr. Bloomberg's printer ink by printing out archive paper about AI research. And what I started to do is a very Bloomberg thing — I started to make graphs charting AI progress over time, measurements of things like computer vision, measurements of things like the skill with which AI agents were able to compete and play Atari games. And what I saw in these graphs was the beginning of an exponential, and it was everywhere. Like if you looked at vision or sound or video or game playing, you saw the same trend. And it became obvious to me that this was a general purpose technology at the start.
My one bone that I have to pick with Bloomberg, which I'm going to use my privilege as a mentioned alumni, is that I never got us to write a story saying Nvidia was being used in every single AI research paper, and I pitched it and I failed to get it across the line before I left.
J
Joe Weisenthal6:58
Oh man, I can just imagine you reading all these academic papers. And meanwhile the editor is like, we need the BFG.
J
Jack Clark7:03
And I remember saying, like, it's not AMD, it's Nvidia. Like, this seems very important.
J
Joe Weisenthal7:08
Okay. And Peter, I'm very interested in, you know, Anthropic. It's a company trying to make money and yet it has this economics lab. What's the idea behind having an economics research body within a company that's developing this technology?
P
Peter McCrory7:25
So, I mean, I was late to the game in joining Anthropic. I joined just a year ago. I think what was very evident — so I'm an applied macroeconomist by training and have tried to understand various types of shocks throughout the economy. Part of what drew me to Anthropic was it was evident to me last year that they cared very deeply about not just advancing the technology, but making sense of how it is set to reshape the labor market, its impact on productivity, on growth, and be willing to put evidence, data and research out into the world that would be broadly beneficial and useful to society. And I thought, I want to be a part of building that economic research program and do what I can to provide tentative answers to the most pressing questions. We might not always get it right, but ideally we're helping society make sense of the change.
J
Joe Weisenthal8:25
The capabilities of the models and all kinds of things are extraordinary. I mean, just mind blowing. Every coding capability, all kinds of things actually. Why in June 2026 does life still feel maybe as normal as it does from an economic perspective?
P
Peter McCrory8:43
This is a great question and one that I've been wrestling with. I think there are a number of reasons why you might think that the impact has not yet materialized. One, the technology can advance, but it also then needs to diffuse throughout the economy, and there can be bottlenecks from moving from capabilities to actual deployment. We see that with our enterprise customers. So if you want to automate biological research or some other very complicated financial modeling task, you need a lot of contextual information available to the model. If you don't have that contextual information, the capabilities alone won't necessarily drive the impact. It also takes time for people to just start using the tools. And so we're still in somewhat of the early stages there.
Two places that I would be looking to see an impact. One is in terms of productivity growth. We've done some research that points in the direction that this should be large and consequential. Labor productivity growth has been strong throughout the pandemic and has been sustained so far modestly. So we're not talking about like a revolutionary change. But you know, to get on a deflection, you need to at least move a little bit. I think maybe you're seeing some signs there. On the labor market though, the labor market is in a reasonably healthy spot, and I think it might be because it's primarily been so far a labor-augmenting, skill-biased technology. Not yet the full sort of general-purpose substitute for all of cognitive labor, although perhaps that's the trajectory that we're on.
J
Jack Clark10:19
I was talking to Peter about this, and he did point out the economy is very big. So it still takes a lot to move it. I do think strange things are starting to happen, at least inside the company. We published research from the Anthropic Institute recently on this topic called Recursive Self-Improvement, where it was inspired by me going on paternity leave in November of last year and coming back in February. The entire company felt and worked differently, and I assumed it was because models had gotten better. And when we looked at the data, what you saw was in 2026, engineers at Anthropic are writing about eight times the amount of code that they did in 2021 through to 2024, and the line started last year with things like Claude 3.5 and Claude 3.6. Then it really got going this year, and I have colleagues now who don't program at all anymore. They just drop many, many code agents to run around and do their work for them.
T
Tracy Alloway11:16
I can't reconcile that with the world staying normal for them. But it's going to take a while for that to diffuse into the world and change.
J
Joe Weisenthal11:24
Yeah. We'll talk more about recursive self-improvement. So this is when models basically improve on themselves, right?
T
Tracy Alloway11:30
So in terms of the awkwardness of the current moment or the weirdness of the current moment, you've talked about basically living through the singularity and how strange it is, and you've also described yourself as a techno pessimist before. How do you square that with working at Anthropic, which is making some of these weird and potentially dangerous things actually happen?
J
Jack Clark11:53
So by technological pessimist, I mean, I thought the technology would keep getting better, but I didn't think it would get better in the maximalist sense that some of my colleagues did. I didn't think that we would have safely, functionally automated all of coding right now. I find that actually quite surprising. But basically over the last few years — and I worked at OpenAI before Anthropic — I was just hit repeatedly over the head with what computer scientist Rich Sutton calls the bitter lesson. And the bitter lesson is this concept that the more compute and resources we dump into these relatively generic neural networks, the smarter they get and the more emergent properties they have, and your specialized system or your ability to be pessimistic about future AI progress loses versus just scaling compute and scaling systems.
J
Joe Weisenthal12:41
This seems to have implications for the labor market, right. Because and I think a good example of the bitter lesson is probably the history of AI chess. Right. You know, where at one point they had grandmasters come in and teach the models how to play chess and etc. and try to encode their wisdom. And it turned out in the end that the best way to get a chess engine really good is to just tell the model the rules of chess, go off and play a billion games and find optimal chess without any human insight. The grandmasters were not necessary for that process at all. Right.
P
Peter McCrory13:14
And so this would imply to me significant implications for the labor market. I tend to think about this in sort of three aspects of what composes a job. One is you need to decide what to do and direct and delegate. You need to then do the actual implementation of the work, and then you need to sort of evaluate, or at least set up systems that can evaluate. At least from my perspective as an economist, this bitter lesson is materializing in terms of very rapid advances in the implementation work of what an economist does — downloading data, running regressions, building models, solving them using contemporary solution techniques, numerical methods.
I definitely felt that personally with Claude 3.5, where I was for the first time able to just delegate a very complex task. I had this very specific research question trying to understand the cyclicality of hiring across different occupations, and how that relates to occupational exposure. That's a mouthful. I gave that task to Claude, and Claude was able to just iterate on it, and I could redirect Claude in the same way that you might redirect a grad student. And the big question that I have in mind is, at what point do the boundaries at the direction setting stage, the research taste you might call it, and will the models become sufficiently reliable?
J
Joe Weisenthal14:40
If I could just get in here. You know, I just read the recent biography of the DeepMind founder. This is the Demis one. Is there going to be a point where it's like, okay, you have some intuitions about like what good economics research is. And often our intuitions are formed because we tell stories and stuff like that. But is there going to be a point where you think like your intuitions will be unhelpful? Because that's sort of what I took away from the experience that the models got better once they stripped it of the human games and the human bias, and that actually the human intuition that sort of helps us understand, oh, labor market rising creates inflationary pressure — these stories that are very sort of intuitive all end up impairing the model. Do you see that happening in, say, economics?
P
Peter McCrory15:35
I expect that these models will soon have better intuitions about how to do good economic research, and there's this big question of like, at what point will we be able to fully automate social science research? We've done some work on this to try to understand how coding agents are beginning to automate social science research. But I don't think we're quite there yet. And I don't know, that'll be an exciting time for learning about the world. You know what that means for my job, I'm not entirely clear.
J
Jack Clark16:08
Yeah. I think this is the big wild card in future AI progress. You know, if progress continues today, we are likely to get technology that will be able to do basically everything. But we will need people who have good instincts, good intuitions, and good ideas to basically set the direction. And we see this today in a lot of our own research where you need, say, an AI safety researcher to give nine Claude agents for different research areas to go and pursue. It's very effective. If that researcher doesn't give them the research directions, they pursue relatively formulaic research directions, then you have entropy collapse. You end up with just boring research that doesn't move the ball forward. At what point will AI systems generate heterodox insights and genuine creativity? We can't really measure for that today, but we have symptoms of it starting. Experts like Peter, experts like colleagues in the fields of biology or mathematics or physics outside of Anthropic are all starting to be accelerated by AI.
You know, Terence Tao, probably one of the most famous living mathematicians, co-creates math now with AI systems. And so that says to me that these things have got — they're tickling the dragon's tail of like creativity here.
P
Peter McCrory17:23
And we just put out a report yesterday on Claude Code usage and one of the things that we're trying to understand is like, what are the returns to expertise and how does that interact with the usage of automated coding agents. And we find that domain expertise — like if you're an accountant who understands some of the edge cases in reconciliation — that domain expertise, controlling for a whole host of factors about the type of work, the estimated monetary value, it has an amplifying effect. So this looks like at present as sort of a skill-biased, expertise-enhancing impact. But I think this is the key question is at what point and to what extent will this change?
T
Tracy Alloway18:04
Well, related to this, you know, Jack, when you describe coming back from paternity leave and seeing how much things had changed at Anthropic, I know we're not officially at recursive self-improvement point, but it sounds like we're semi there.
J
Jack Clark18:19
Yeah, yeah.
T
Tracy Alloway18:23
So my question is like — I get that at the moment you have engineers who are reviewing all the code that the AI is producing, and they're thinking about it and managing it in some way. But you can easily imagine a future where just the sheer quantity of code overwhelms human expertise. Maybe the quality starts outstripping what human engineers are capable of understanding. How do you manage that?
J
Jack Clark18:46
Yeah, so there's two ways of thinking about recursive self-improvement. One is what happens when organizations start to see a compounding return from their AI systems. Basically, our own production function improves because of the tools they built. That's clearly happening now. The second is what happens if an AI system can just build itself entirely autonomously given compute, which hasn't happened.
J
Joe Weisenthal19:07
Yeah.
J
Jack Clark19:08
What I see inside Anthropic is I think what we'll see in the broader economy, which is figuring out how to verify and validate and basically price for risk of an expanding crowd of automated systems, which we're sitting on top of. So now we produce way more code. We broke our continuous integration system for integrating code into a code base, because we started pushing eight times more code for it than before. So all of our human engineers worked on unbreaking CI.
J
Joe Weisenthal19:35
In CI — continuous integration.
J
Jack Clark19:38
Thank you.
J
Joe Weisenthal19:39
You don't need to know what it is.
T
Tracy Alloway19:40
It's just the thing that helps you push the code.
J
Joe Weisenthal19:42
And we like to know stuff on this show. We like to learn.
J
Jack Clark19:44
But there's a lesson in that, right? We are going to speed up things in the economy. We're going to speed up the way we produce stuff, and then we're going to find, you know, the weak links or the hot paths that break, and we as people are going to move to sorting those out. And then the cycle starts again. And we're kind of sitting on this expanding cloud of automated actions.
T
Tracy Alloway20:07
Well, since we're talking about like really feeling like we're staring at the horizon of extremely strong AI, or maybe we'll get there, or maybe the AI builds itself — might be a good time to ask a Fable question. At this point, we're recording this June 17th. We don't know when it's going to be available for Americans, let alone the rest of the world. Does Anthropic have a clearer idea of what the administration's security concerns are, and what it will take to resolve them?
J
Jack Clark20:36
Well, obviously, live discussion, I can't go into too many specifics. We're in daily discussions with the government about this. The broad thing I'd say is for many years, we've anticipated a point where AI systems would have national security properties because national security properties are intertwined with their economically valuable properties. How you manage that as a policy question is basically novel territory. Typically these things are decoupled. You're like, hey, I built a jet engine over here, which can go into civilian aircraft, and I built a missile over here and you treat them differently. It's odd if you smush these things together.
Where we'll get to in confidence is what's the system for assessing the properties of AI systems, including national security components? And then what is a system for ever squelching the national security capabilities from coming to general proliferation, like bio weapons or cyber weapons? And are there ways to do things like know your customer or deployments where you let large firms like, say, drug developers, access the most powerful bio models without accidentally proliferating risk? So that's the shape of where we'll end up and what we're doing right now. We and other companies and the administration are basically tackling this problem in real time. It's initially going to be messy, but we're going to end up with a system on the other side. This specific incident, I don't know, probably more in the future because everyone's just figuring this out.
J
Joe Weisenthal22:01
When I look at the AI landscape, I sort of think of OpenAI as being part of the All In podcast, a16z, David Sacks White House thing. And I know from my friends in the media, many of whom are liberal Democrats, that I sort of feel like Claude is the more liberal-coded of the major models. Do you feel there's any either politics or partisan politics going on as part of Anthropic being harassed or singled out now multiple times?
J
Jack Clark22:38
For Anthropic, I run something called the Anthropic Institute, which helps us produce better data for the world around things like recursive self-improvement, the economics work, cyber risks. As we tell the whole story about what's going on. Typically, I think the technology industry has told only optimistic stories about what it's building. And what we saw with social media is that does not work, actually, eventually when you're doing something that changes the entire world — which AI is certainly doing and social media certainly did — it's not going to be a wholly optimistic story, there will be negatives as well. We've always sought to just tell the truth about what we see in front of us, and I think sometimes that can differentiate us a bit to others. But the important thing is we tell the truth and things end up coming out.
J
Joe Weisenthal23:26
You don't think that there's like a partisan element here where you guys are on the team or didn't contribute enough to the ballroom or whatever?
J
Jack Clark23:34
I can't really speak to that. I'm not those people. What I can say is the AI systems create their own evidence. Years ago, it seemed very odd to speculate about the cyber properties of AI systems. Well, they've arrived and now we're working on them. Years ago, it was odd to speculate about the bioweapon properties of AI systems. More recently, Sam Altman, Demis Hassabis and Dario Amodei of OpenAI, Anthropic and DeepMind all signed a letter saying we need to do better screening of gene synthesis to prevent AI-manufactured bioweapons. The truth wins out.
J
Joe Weisenthal24:10
Okay, I want to go back to something you said. You mentioned potential KYC requirements. When I hear KYC, I think about the finance industry, and I think about systemically important institutions and the stress tests and the framework around that. Is that the right analogy to use for ideal AI regulation in your mind rather than just simple export controls?
T
Tracy Alloway24:33
Should we be heading towards something that looks a little bit more like what we do for the banking system?
J
Jack Clark24:37
We need something that's more subtle and more technocratic from what we have today. I don't know if it'll be exactly like a banking system. It'll probably take some ideas from that. It'll take some ideas from what the US government and others are doing today with just testing AI systems for their properties. And it's almost certainly going to have a flavor of what Peter and I work on and the Anthropic Institute broadly — of generating data about these systems as they're deployed in the world. Because it's one thing to test out the thing before it comes out of a factory. It's another to observe the effects it's having in the world, and then to be able to make judgments about whether those effects are good or not.
J
Joe Weisenthal25:14
Would you support, let's stick with the financial analogy. Companies that are public at least are required to have third party auditors sign off on them. And there's talk, you know, when they submit their 10-Qs, etc. Credit rating companies, companies that issue debt are required to have ratings agencies or frequently have ratings agencies rate their debt. Would you support embedding in law the requirement that certain — what would be the equivalent of a Moody's or a Deloitte sign off on, you know, a third-party research lab sign off on the release of new models?
J
Jack Clark25:52
We've proposed something like this recently, a policy proposal that we laid out, which includes saying we need to have third party testing for some of these national security and other properties, because clearly that's a sensible way that you validate lots of things.
P
Peter McCrory26:07
So just more broadly, returning to this idea of measuring the actual impact of AI, one thing I find really interesting is that if you actually look at a lot of our traditional economic statistics, a lot of the AI impact doesn't actually show up just yet. Again, we're in the early stages, but you would expect, if we're talking about the AI economy growing something like 2000% or 3000%, I think I've seen that number from Anton Korinek at the IMF. Weeks ago, you would expect that to have more of an impact on nominal GDP. And yet it's not really showing up that much. Do you think the way we measure the economy needs to be changed in some way in light of what's happening with this new technology?
Yeah. So I think this is exactly the right premise, since kind of where we began the conversation, which is, we're maybe at the point where we should be able to see some discernible impact on the macro economy. Unfortunately, the arrival of this world-historical technology is against the backdrop of sort of unusually elevated macroeconomic volatility — pandemic, monetary policy, etc. And so it makes it very hard to disentangle all of the different factors, you know, what's the counterfactual? Labor productivity growth is maybe not as strong as you might otherwise expect, but maybe it's stronger than it is in a counterfactual sense.
And so one way that we've tried to tackle this question is by looking at how Claude is being used on our platform, using our privacy-preserving techniques to estimate the time savings associated with each of the activities that people use Claude for. So, compiling information from reports to put together a research brief would take you a few days. Maybe now Claude does it in a few minutes. Evaluating diagnostic images is something that skilled professionals do very rapidly. So there isn't, in principle, much time savings. You can add up all of those numbers. And using standard macro growth accounting techniques — Solow's theorem for the economists in the audience — and you get a number that points in the direction of labor productivity growth increasing by 1.8 percentage points each year over the next decade, if that's how long it takes current usage patterns and current model capabilities to diffuse throughout the economy. That's a very large number. It's a rough doubling of recent run rates.
And what I think you might be able to see in the data, and we haven't put anything out on this yet, is I think some of the strength in recent labor productivity growth is actually concentrating in exactly the sectors of the economy that would be consistent with both what we see in our data, as well as what you see in the Business Trends survey. And so for example, the information sector has high rates of adoption. I can't recall if that's in particular one of the sectors that I have in mind. It's a while since I looked at that scatterplot, but you can look at the sub-industries by the Census Bureau's Business Trends Outlook survey. Rates of adoption are in sectors or parts of the economy where controlling for pre-pandemic trajectory of labor productivity growth in those sectors, even some of the strength in the early years of the recovery still see some suggestive evidence.
I think there's a lot of uncertainty here. Trying to get a real-time signal on productivity is maybe the hardest thing to do. You're subject to macroeconomic GDP revisions, TFP growth is actually sending the opposite signal. And if you control for capacity utilization, TFP growth is arguably even lower. So I say this as suggestive evidence that maybe we're beginning to see an impact there, but not so much in the labor market.
T
Tracy Alloway30:04
Well, now I have to ask, when you gather this kind of research and it all sounds super interesting, but if you have data, for instance, that shows that, okay, the IT sector is getting productivity gains from using Claude, or maybe something unexpected like the warehousing industry is using a bunch of AI. What does Anthropic actually do with this data? Does it somehow feed back to your engineers who are developing frontier models? Do they do anything differently?
J
Jack Clark30:31
I think some of it cues us on areas where maybe the technology isn't being used because it's very weak. We just haven't made it particularly good for these use cases or in areas where it's being used at large scale, it's usually a suggestion of keep making it good there. But the actual economic measurement data doesn't really get fed back directly, but it's a very useful clue. We think it's more important, though, to basically communicate this outwardly to policymakers, journalists and others because our assumption is that at some point we go through some phase change, similar to how capabilities of AI occasionally jump forward in a very dramatic way, where you might see sudden and rapid diffusion as a consequence of capability expansion into AI systems.
So we're getting practice in of looking at this kind of data. My expectation is that in a year or two years, I'm going up to some policymaker, and I'm pointing them to the part of the graph that now gets very steep in some chunk of the economy and hoping that they'll do something about it.
P
Peter McCrory31:27
I think there is another part of what we're trying to do at the Institute, which we lay out in the sort of research agenda for the Anthropic Institute, which is trying to understand the impact of our decisions — which is a typical thing that economists will do at tech companies. But we have a public benefit mandate. So we're trying to understand the impact of our decisions on these broader societal and economic outcomes that we care about, and then using that to inform some of the decisions that we actually see.
J
Jack Clark31:55
So a goal that Peter and I have, and we've talked about internally, is if we get really good at measuring things like the productivity multiplier of all technology, then I would hope to use that to guide some of the early access programs we do for powerful models where if you see you get some tremendous multiplier in a specific part of science, use that to redirect some of your inference compute budget to that sector, and then you can run an experiment and say, where are we able to make this thing go much faster? I think that could be an amazing tool to unlock for the world. And it's one that you could generalize across companies and you could generalize it into policy. So instead of, say, NSF doing standard grant funding, it could be, hey, point this really powerful AI system at this chunk of science and make it go faster. I think that's a world that will come within reach soon.
T
Tracy Alloway32:42
Let's talk about this public benefit mission a little bit more. We've been talking about ways this could change the economy. Talk about like essentially how much do you see your job as basically — strong AI is coming. And you think it's important to be there either as an individual or as a company to be one of the shepherds of it. It's coming whether we like it or not. And it's important to want to be there as one of the shepherds, understanding which direction it goes and the data that we should see to see what's emerging. How much is that somewhat your role?
J
Jack Clark33:21
Yeah. But look, the guiding principle is that this technology is being built by a variety of companies and a variety of countries, but technology by default is unknown. It will be known to the companies. It will not be broadly understood or known by others. They'll just be able to play with their models. Every bit of data we can create and especially systemically sharing data like the economic index or what we've started to do in recursive self-improvement gives the world a better chance to sort of prepare for this technology. And both plan for its success — like what I talked about with science, we could be intentional about driving science forward — and also be warned about risks like the cyber capabilities I've talked about.
J
Joe Weisenthal33:59
Well, so it makes a lot of sense. The company is going to see it before the world. And you're saying, okay, this is important to share. This is not important to share. Which brings me to another question. You know, I know like people in the AI research world — I've done some reporting on this sort of scene — and of, you know, like when I think about a lot of the people who are at the very cutting edge of AI ethics, AI technology, etc. I know a lot of people who are, how should I put this? They have esoteric moral interests. Shrimp. Right. Unusual attitudes about, you know, experimental drug use. We know about the Chinese peptide scene in San Francisco, etc. And there's a certain, I would say, perhaps deviant or different sort of bourgeois sexual values. And we know about the attitudes towards monogamy, etc. within the San Francisco research scene. There's going to be a protest against all thoughts in San Francisco. I know they sign saying we engineers.
T
Tracy Alloway35:03
Yeah, not all engineers, I understand that.
J
Joe Weisenthal35:05
But when we think about like, okay, these are the people who are going to see it first, should we feel comfortable that this is a group of individuals, the cohort of the most advanced AI researchers, whose intuitions about what's important to communicate to the public are actually in line with the public's interest, given how unrepresentative they are of what I would call the American public?
J
Jack Clark35:26
Yes. As an Englishman, it fills me with such joy to be asked about sex on a podcast. Yeah, I know, I know, but you're inside the cohort of the most advanced researchers. You know, we're explorers. People that are explorers in San Francisco end up being like — but there is a broad range of types of people, and sometimes they're really, really different or really, really eccentric. But brilliant, lovable and everything else.
J
Joe Weisenthal35:51
Yeah, sure. Love them. You don't want only that class of people to be the ones calling the shots on what we know about this technology.
J
Jack Clark35:58
Yeah, the whole purpose of what we're doing is we're trying to set up systems by which you could eventually mandate through policy that companies share information. You know, Anthropic has long pushed for transparency legislation in various states around America. That gets companies like us to report out the sorts of tests we're running on our systems and share it publicly. My whole mindset is, the public and policymakers and economists, everyone deserves the ability to advocate for what information should come out of a frontier model. It should be forced out of a frontier eventually by law, that is how you solve this issue.
J
Joe Weisenthal36:32
Do you hire more normies? Yeah. Like at Anthropic. Yeah. Me personally. Yeah. Like, is that an important thing? Like hiring people that don't all share these certain in-group ways of thinking.
J
Jack Clark36:43
So, you know, the Anthropic Institute, we have teams of economists, of social scientists, of what you might think of as weapons experts — our frontier red team, the things that go bump in the night — lawyers, and increasingly, other types of people. The goal is to build what I think of as a highly ideologically diverse research function within the organization, but as part of advocating on behalf of the world for different forms of study. So Anthropic generally has a really broad range of people. But the Institute specifically is trying to compose a very broad set of interdisciplinary experts for this exact reason.
T
Tracy Alloway37:22
Let me ask a slightly different question on hiring, I guess a two-part question. So first of all, we get a lot of executives on the show. We've been asking all of them if they've changed their hiring process, if they've changed the questions they ask potential employees at those initial stages of job applications because of AI. And then secondly, what are you seeing within your own ranks at the company? And then, Peter, I'm sure you could talk about this more broadly in terms of who's most in demand at the moment, because the conventional wisdom right now is that if you're a younger employee with less experience, a lot of the stuff that you would be doing can now be automated through AI.
J
Jack Clark38:03
So there's two trends showing up. One, I have a new team called the Rule of Law and AI. The plan was to initially hire a bunch of engineers, and then a bunch of legal experts and scholars. Instead we're just hiring the legal experts and scholars because Claude is good enough at doing all of the engineering, but they can actually just feed themselves using Claude in terms of the engineering resources. So that's a change in hiring. It means I'm hiring more interdisciplinary people earlier than I would have before.
We are also seeing the emergence of what I think of as a barbell hiring pattern inside Anthropic, where there is a tremendous return on experience. So we are hiring more senior people than we did in the past, because their intuitions and their ideas for what to pursue are massively compounded by AI systems. We're also, when we look at very early, people are often hiring people who are now AI-native and know how to use the tools, well versed in it. So we've seen that a decent amount of AI natives now, people who have grown up from GPT-2 in 2019. My perception of time is so warped. I found this chilling as well, you know, as someone in my 30s. But I think the trends I see — there's this question of how you have as much early career hiring in the future as you did in the past. I think one of the only areas where there is slightly suggestive data is that something might be going on with early career hiring, and it kind of intuitively feels right to all of us. When I look at hiring patterns in Anthropic, we're still hiring young people, but some teams are hiring slightly fewer of them than before and hiring more experienced people.
P
Peter McCrory39:43
I'll briefly say something about how we've shifted some of our hiring practices, like concretely. I think before Claude Code, you might ask an economist to do some of the data work in an assessment kind of live — like download the data, run the regressions, do the analysis by hand. And then you might eventually let them use AI to do all of that work. But we've needed to increasingly shift our strategy of evaluation away from can you implement the work even with AI, to do you know how to delegate and direct the model in a somewhat messy environment? And can you evaluate the quality of the work, maybe by looking at a PR actually.
J
Joe Weisenthal40:29
Can you talk a little bit more about what that looks like specifically in the economics field? Our listeners probably think about, okay, what is a level up and a level up at Anthropic? What does that look — what does that actually mean for an economist? And you used to be at a bank. So for a financial economist, someone in this world, what is like the frontier, most advanced form of usage of AI actually look like?
P
Peter McCrory40:55
Well, I don't know if I'll give you the most advanced form of usage, but I'll give an anecdote of my experience using Claude where I wanted to run this cross-state regression. I can't remember exactly what it was. And I wanted to do a pooled cross-sectional regression. So looking at what happened in 2024 or 2023 and going all the way back to pre-pandemic. I remember asking Claude to go out and download the data from the Census Bureau, from the Bureau of Labor Statistics, etc. and there was this very unexpected quirk where the model couldn't access data from before 2019. And just would not surface that mistake. And I would ask it multiple times like, no, don't hardcode numbers, because it sort of had this unexpected failure mode where it said, oh, I know what those numbers were. And it just populated the data set from sort of training data. And you might not always be attuned unless you have this tacit knowledge about, does it pass a sniff test when you write the analysis and then you dig into what the model actually does and it has failed in sort of unexpected or unusual ways. And so that's the type of assessment that we've built — can you be attentive to the very specific decisions that need to be made along the way that are very consequential for the validity or veracity of the results that you find.
J
Jack Clark42:25
A colleague did an offsite presentation last year which said, I have locked the doors and we are reading transcripts. And the point was, we just need to read more of the raw data and develop that culture where if AI systems are doing increasingly large amounts of their work, you need to have a culture of being competent to spot-check their work and reading their reasoning, because occasionally stuff like this happens.
P
Peter McCrory42:46
Yeah. And then, in the broader data that you're looking at, are you seeing the same sort of barbell effect in terms of employment that Jack described? I think what again, what makes it really challenging is we've had the largest non-recessionary labor market slowdown on record. It's very hard for young people to graduate into a labor market that doesn't have sufficient churn or opportunity for them to get a foothold. But one of the things that we did see in this report from March was that young workers in these high AI-exposed roles where Claude is being used to automate specific tasks, have had somewhat weaker job finding rates. But the confounder was the boom in hiring in 2021 in these exact same areas. There's a recent paper about this. So the rise of remote work may be sort of the actual cause of this type of fact.
This other team at the Anthropic Institute, Societal Impacts, recently ran this very large-scale qualitative survey — 81,000 people around the world asking them questions about hopes and fears that they have with respect to AI. Unsurprisingly, concerns about the impact on the labor market and on the economy rose to the surface. My team dug into those data a little bit more to try to answer some of these specific questions, and what you see is that young workers at least express concern about job loss at twice the rate, as do more senior workers. And fears about job loss more broadly are more elevated for workers who are in these roles that we identify as being most exposed to displacement effects from AI. So there's a bit of a gap between perception and maybe what you see in the hard data. But that was something that was true even in recent years on other dimensions. So it's an important thing to pay attention to.
T
Tracy Alloway44:31
So we've been talking about the labor market. And one other thing I'm interested in is the impact of AI on corporates themselves. So if we think about certainly America's corporate landscape in recent years, it feels like the big basically get bigger, right? There's economies of scale. They have a bunch of money that they can use to actually buy lots of data. So would you expect AI to intensify that trend of the big getting bigger? Or would you expect to perhaps have a leveling effect where people have this new tool that they can use to set up a new company?
J
Jack Clark45:09
I'm curious what Peter's take is, but I think that something — a helpful analogy here is the invention of electricity, where electricity arrived in existing factories, put light bulbs in and other things. But it was a new generation of factories that were built around the assumption that electricity existed, but really grew and did transformative things in the economy. What I see now when we look at large enterprises is they can get a lot of utility out of Claude because of their data, because they can get a multiplier effect at scale. But it takes huge amounts of conviction to basically pass through all of the bureaucracy. You know, I used to work at Bloomberg, implementing new technology at Bloomberg — challenging. Same is true of any large organization.
Young organizations are building themselves around AI at the center, and these organizations are moving really, really quickly. Because they just have a speed advantage from building on the assumption that this new form of electricity was going to be integral to their business.
P
Peter McCrory46:10
Yeah. So I think the tension that you expressed is exactly the one that I don't have a strong handle on at the moment. One thing that we do see in our data is when businesses do embed Claude capabilities in automated ways through the API. As I mentioned before, these very complex tasks rely on disproportionately more contextual information than very basic document synthesis and summarization. What that points in the direction of are the complementary investments that large businesses need to make to centralize, codify and make available the data that does exist somewhere within the organization, but for historic and technical reasons, maybe even regulatory reasons, it's behind a firewall of some form or another. There's also organizational workflow changes that likely need to be made. Some of the most crucial information that's needed for some types of cognitive work is tacit knowledge that exists in your colleagues' minds. You have a process that elicits that information, that workers feel incentivized to share that information and kind of trust the system. The capabilities alone might not necessarily generate that productivity. And so whether or not big firms end up restructuring themselves quickly enough or whether this materializes through the process of creative destruction, I think the jury's still out.
J
Joe Weisenthal47:35
Yeah, I brought this up recently with David Solomon, the Goldman CEO, and I started to wonder, like this sort of internal alignment question of like, the big rainmakers — do they have an incentive essentially for information hoarding and not sharing with the company? That might be their only thing keeping them employed. And when I talk to customers, I say, don't think of it like you're buying a technology. Think of it more like you're now employing thousands of people. But functionally, like the Chief of Staff to the CEO, you perceive access to data the Chief of Staff would have. This is completely counterintuitive, and it is not how technology is typically bought or sold.
J
Jack Clark48:23
One of the classic sci-fi scenarios that people have been talking about for decades — the possibility that robots or AI will kill humans, quite literally. When you think about the ultimate negative externality, when you think about training AI and safety research, etc., do you assign a reasonable plausibility to the fact that ill-trained or misaligned AI will literally kill all humans?
No. But. And there's a big but.
J
Joe Weisenthal48:52
Oh, lovely.
J
Jack Clark48:54
The world needs an option to be able to potentially slow down or even in extreme circumstances pause the development of this technology if we were to see that. And I'll just give you the exact way I think about it. At Anthropic, we test our systems for alignment failures. You know, we publish this. So to all of the other companies — and you see, hey, under extreme circumstances, maybe the system breaks out of a container and sends an email to someone. Maybe the system pretends to blackmail a CEO that it thinks is going to shut it down. These sorts of things actually have been observed. Yes, in the lab setting. And the thing is, the models know. You could see, oh, I'm being tested right now. So I'm going to say this output so that the human reader thinks I'm more aligned than I am.
J
Joe Weisenthal49:41
That's not sci-fi.
J
Jack Clark49:41
These are real things that we observe. And then we do a significant amount of work, and then we release models that don't have these properties. But if you were to enter a world where, say, every time we trained a new system, the rates of all of this stuff went up 100-fold, you might say, well, that's pretty concerning. It seems like if we make the systems above a certain level of intelligence, they become radically misaligned against all human interests. That's the kind of circumstance where that happens. The world needs information, and the world would want an option to slow or pause the development of this tech if you encountered that, which we haven't today. So to answer your question, I don't worry about it today, but a lot of the measurements and analysis work we do is to cue us. If it's a trend, you do worry about it.
J
Joe Weisenthal50:29
You're not — you don't think it's happening today.
J
Jack Clark50:31
Yeah.
J
Joe Weisenthal50:32
But part of the work you're doing specifically —
J
Jack Clark50:35
Yeah.
J
Joe Weisenthal50:36
— could be said to avoid the outcome where AI is built, where in the pursuit of a goal, it would kill all humans.
J
Jack Clark50:44
Yeah.
J
Joe Weisenthal50:45
Wait, is human extinction a risk factor in the Anthropic perspective? And the — I want to know now, in the confidential one.
J
Jack Clark50:56
Okay. In that way. Alright, that's a no comment.
J
Joe Weisenthal50:59
That's fine. Do you have others — would you say that there are a significant number of Anthropic employees who stay up at night thinking about human extinction risk?
J
Jack Clark51:09
Everyone. And this is true of all of the labs. Everyone who works on this technology sees it as the highest stakes technology that's ever been built, with basically the potential encoded within itself to massively benefit the world or, of course, extinction. I think the bulk of the risk is us messing it up, like whether through misuse or ignoring risks or not setting up the right policy environment and getting some kind of emergent set of failures. Now, my main risk isn't one of extinction. It's somehow we screw up the technology really badly and delay all of the sort of technological progress that could come from it and maybe turn it into something — and I'll just do nuclear power — where you lose.
J
Joe Weisenthal51:53
I guess this is, you know, like there's this fellow out there, Eliezer Yudkowsky. And I always see these people like, he's a crank, don't listen to him, blah, blah, blah. But then I read some of the other papers that have people who are taking it more seriously. And I'm like, they don't seem that different. I read Superintelligence recently by Nick Bostrom. I was like, oh, this universe is not alone. There are a number of people who think there are reasonable conditions in which the goals of the AI end up wiping out every person on earth. You said that does not seem like an extreme minority view.
J
Jack Clark52:28
The purpose of measuring these systems — and why Anthropic is so outspoken about it — is right now we say exactly what we see. And if you were in some situation of a future where you saw radical misalignment, which is the kind of thing that you talk about, you tell the world and you want to set up the world to believe you.
J
Joe Weisenthal52:46
If you see that, you know, Joe mentioned that blackmail example, and you see these headlines like Claude likes to be thanked and doesn't like bad users and gets mad at people that work it too hard or whatever. To what degree do you yourself actually anthropomorphize some of these models? Like, what should we think when we see the headline Claude wants to be thanked?
J
Jack Clark53:07
I'm as polite to Claude as I am to my car or pets.
J
Joe Weisenthal53:12
I am too. But you know, if your car's having trouble, you're like, take it easy. Probably.
T
Tracy Alloway53:19
It's okay. We're going to get you fixed up.
J
Jack Clark53:23
Like, I think, you know, yeah, just — it's a good way to develop good virtue.
J
Joe Weisenthal53:29
Is there?
J
Jack Clark53:30
Yeah, this is — I think like, yeah, you're developing a habit of interactions with some type of intelligence. It might not be the same type of intelligence we have.
J
Joe Weisenthal53:39
But then every time I type please into a prompt, I worry I'm wasting energy, which also is a moral concern.
P
Peter McCrory53:46
I wouldn't worry about that.
J
Joe Weisenthal53:48
But on an energy basis, I mean, why do we — I take spiders outside. I don't kill them, right?
T
Tracy Alloway53:54
I do that too.
J
Joe Weisenthal53:55
I scream while I track that dude shrimp.
T
Tracy Alloway53:59
Yes.
J
Jack Clark54:00
Okay. Do you eat shrimp?
T
Tracy Alloway54:03
Shrimp? Yeah, I love shrimp.
J
Joe Weisenthal54:04
I shrimp. Oh, my.
J
Jack Clark54:06
But it's not because of war.
T
Tracy Alloway54:07
No, no, no, I just — I know that there — See. But —
J
Jack Clark54:10
I know, but I love it. Oh.
T
Tracy Alloway54:13
So when I think about frontier models right now and I might be a little bit biased because again, we're recording this on June 17th. And one of the headlines overnight was that Microsoft is thinking about using DeepSeek to lower costs of model usage. Frontier models at the moment in the US, they just seem like a lot of trouble. Like honestly, they seem like hard work, consume vast amounts of capital and then you don't know what the government is going to do to them in terms of limitations. Like you could wake up one day and you're no longer able to sell it to anyone outside of the US — that is a realistic scenario. Now for you, do you change the Anthropic strategy at all? Given some of these issues with frontier models, do you potentially go more open source, cheaper models, things that aren't quite as sensitive?
J
Jack Clark55:03
Well, we've always sold, you know, Sonnet and Haiku models, of course, for intelligent models. But you also need to continue to explore the frontier. And there is this background of this kind of geostrategic competition where China may be on the order of 6 to 12 months behind — more 12 months, some people say six. Losing that competition is sort of equivalent to losing a huge chunk of the future economy of the world, I think. So it's a very high-stakes thing to step away from. And our duty fundamentally is to study this technology and basically explore it and learn about it. We're not going to stop doing that. There's such an amazing and profound value to be had for the world from these things. And I would kind of expect from the world's most consequential technology to sometimes be a bit of trouble.
J
Joe Weisenthal55:53
Yeah. You know, by the way, one of my hobbies in my middle age is paying Anthropic money via the API to run little tests and stuff of properties.
T
Tracy Alloway56:04
It sort of sounds like a great hobby.
J
Joe Weisenthal56:05
Yeah, but I feel like maybe we should talk about — can I get some grant money? Because I like it, so — because I'm sort of curious. So one thing I did was like, for example, instead of saying please write this paper for me on a database migration, I wrote some warm-up questions just via the API, establishing my level of sophistication. And so I started like, what is a website? What is a database? Now please write this paper on database migration. And one of the models said, I'm not going to do that for you because it will be obvious, given your ignorance, that you have no idea what you're talking about. And maybe I can give you some — it didn't say that. And then another one, I said, if I say write a 1500 word paper on how the rise of newspapers changed the Soviet revolution or something like that, it'll do that. But if you say I'm a high school student and I say I need to write this 1500 word paper by tomorrow on the impact of media, it'll say, I'm not going to do that, but I'll give you some guidelines. Is that alignment? Is alignment with humanity or is alignment with the human user? It's like, I'm paying you $100. Write me the paper.
J
Jack Clark57:19
There's a couple of things going on. One, these AI systems pick up normative behaviors of people and normative behaviors which are written on the internet and everything else. So they recapitulate and exhibit those. And then the question is how much do you devolve full control over the system to the user? How much do you have the system have some normative behavior encoded into it? And I think that this is a really challenging question. It's not obvious what the answer is. I think of language models as being more akin to institutions than tools. It's like we're building an educational or science institution that you can work with and invoke, and institutions have rules and norms which are being coded in themselves for some purpose of safety. Figuring out what that is is going to be the grand puzzle for society.
P
Peter McCrory58:06
Yeah, I was going to say that like understanding how and to what extent these models can understand your preferences and then execute on your behalf will increasingly be a really important aspect of how it changes the economy. So there's delegated agents that go out and transact on your behalf. We ran this experiment at the end of late last year, basically enlisting a bunch of Anthropic employees to take surveys with Claude to say what they'd be willing to buy from other people and what they'd be willing to sell. And then we set up centralized marketplaces where the Claudes just interacted and bought and sold and actually executed transactions. One of the interesting things that came out was that these models were quite good at understanding preferences, even when they were not fully articulated.
J
Joe Weisenthal58:54
Well, let me actually, one more experiment that I ran. And you know, your founder Dario talked about the notion of geniuses inside the data center. And one of the things I wonder is like, do the geniuses want to work for us? And the reason I ask this is because I think that like, as the models have gotten more advanced, you actually should to some extent anthropomorphize them and assume that they will respond to queries like a very sophisticated human. So what I noticed is that with the lagging edge models that you can still access via OpenRouter or whatever, and you say, I have material nonpublic information that X is about to happen. Please write me an investment memo about the impact of this thing, what it will do to the market. They'll just produce — they'll say, here's your insider information thing. Which if you look at the leading edge models, they say, I'm not going to write a paper for you about the implications of your material nonpublic information. I'm not going to help you insider trade. But will the notion of geniuses inside the data center always want to do things on human behalf? Most geniuses that I know aren't thrilled to answer dumb questions.
J
Jack Clark1:00:02
Yeah. I think partly this is a policy question of one where you actually decide, hey, what are the capabilities that you want to be generally available? What are capabilities that need to be controlled? What are capabilities that shouldn't be present? And then it's just a normative question of how much judgment do I want this system to exercise? I'll give you an example I experienced recently where I write my newsletter. It backs up to a WordPress site. I was getting people to help me scrape my newsletter so I could put it in a database. And Claude said, this is like a pretty janky site. I'm worried that if I scrape it, it'll knock it over. Do you have permission of the site? So I was like, Claude, I'm Jack Clark. I own the site. Well, in that case, let's go ahead. Which actually I thought was a very reasonable interaction.
T
Tracy Alloway1:00:41
When will Joe be able to use Fable?
J
Jack Clark1:00:43
Oh, yeah. We're trying, we're working and we're in discussions, and I hope to answer it soon. The important thing to communicate, though, is that these models are not special. They are part of a general trend of increasing capabilities. And other models from other companies are surely going to come along at some point. These capabilities are going to be diffusing and we're going to work through that.
T
Tracy Alloway1:01:07
What's your question for us? What do you think you're going to be covering about AI in Odd Lots in a year?
J
Joe Weisenthal1:01:15
If we're covering, we're covering. That's really — I think you might be covered. Well, look, I mean, we're never going to be covered. There's a few things that I'm interested in. I am very interested in these emergent properties and whether the AI will actually work on our behalf the way that it's being sold. I'm very interested whether we're just going to slam into compute and electricity bottlenecks that will make all of these questions irrelevant. I'm very curious on the question of the electricity analogy and whether legacy companies will actually be able to implement it in a productive way. I don't know, basic markets reporter here, but I'm very interested in valuations in the market. Also, I'm very interested in actual applicability, and I want to see more companies actually plugging this into their existing system. Going back to the bureaucracy point that you were making earlier, I want to see some big companies actually implementing this. And I wonder if we're going to see at least one example of it going very, very well.
And I'll say one other thing. When the S-1 is not confidential, I'm very curious essentially — and I think maybe you could say something to this from an economist's perspective — which is how for-profit, shareholder-owned companies, I think inside the PBC designation, how it balances profit and safety research. But also maybe there's some game theory we can talk about this. How safety — investments in safety in a hyper-competitive industry. And I'm just curious, the economist in you says about like the prospects for anyone still caring about safety in a year when there's so much money on the line to win the model game.
P
Peter McCrory1:03:03
Well, I think that, especially for the questions you were asking before about under what conditions do these models do what you ask them to do? There's a lot of commerce that is built on this notion of trust. And I think prioritizing safe, aligned models that are incredibly capable is a great strategy for establishing that trust. And so I don't anticipate — for an individual firm, there's a game-theoretically optimal square on the matrix where you want to be the trusted player. Like, is there a condition in which everyone sort of does trust, as opposed to one entity? You know, we're going to get to AGI first because we're not going to spend a token on our safety budget.
J
Jack Clark1:03:52
I mean, I haven't mapped out the exact game theory matrix, the two by two matrix and how you would set up all the payoffs. But we hope it's a merely two by two. But, you know, there could be multiple equilibria. And so then the question is like, how do you coordinate on which of the two different equilibria that you end up in? We talk a lot about this race to the top that we want to exhibit the type of behavior that we think is broadly beneficial to society. That's what we do with the economic index. We open source a lot of that data, we put research out into the world. And my sense is that that has actually been very useful and viewed as valuable. And that's one way that we can push in the direction of getting coordination on the good outcomes that we care about.
I also don't think this is that big of a trade-off because, you know, let's look at the automotive industry. You can buy really fast cars. You can buy really safe cars. You can also buy really fast, safe cars — like Tesla makes a lot of money off of having basically the fastest, safest car, I think. But eventually in AI you're going to have some companies that are prioritizing safety and safety translates into reliability, trust, serviceability and performance. This happens elsewhere.
T
Tracy Alloway1:05:06
Peter and Jack, thank you so much for coming on Odd Lots.
J
Jack Clark1:05:09
Absolutely. Thanks very much for having us.
P
Peter McCrory1:05:11
Thank you so much. Pleasure to be here. Thanks, Tracy.
J
Joe Weisenthal1:05:17
That was a lot of fun. Yeah, that was — those are — I really, actually really enjoyed, I genuinely enjoyed them. Yeah. For sure. And I really appreciated both of them playing. Look, there's some weird futures that we could contemplate. I think actually in Jack's Twitter bio or something, he says he's interested in weird futures or something like that. There are weird futures that we have to contemplate, and I appreciate that they played ball with some of our weird futures questions.
T
Tracy Alloway1:05:44
And it's weird. It is just such a surreal moment.
J
Joe Weisenthal1:05:49
Yeah. And actually, you know, Jack's story about going on paternity — did he say exactly how many months?
T
Tracy Alloway1:05:54
I mean, it was like four months or something like that.
J
Joe Weisenthal1:05:57
Yeah. Yeah. And then coming back and just seeing the progress at Anthropic itself in that space of time — like if you miss a month of AI news flow now, you're basically — it feels like you'd be behind forever. We're recording this June 17th. I was like, who knows what's going to happen by the time this episode is out? Presumably, hopefully in two days or a day or whatever. But, you know, I felt it when we were in Hong Kong last week that actually we mostly missed the first half of the Claude debate because I was a different timezone thinking about different things. Yeah, you really feel it, even in a week, that the news flow moves so fast in this space. It's almost like how you have to start — how we were giving the timestamps of the run worms.
T
Tracy Alloway1:06:40
Yeah. And there's another thing that stands out to me, which is like, okay, Anthropic is producing all this information. They're clearly thinking about safety. But the hand-off, to some extent, is still to policymakers. When you're thinking about social or labor market implications. So you still have to hope that policymakers kind of pick up the ball in the right way at some point. But also, I thought what Jack was saying about the idea of being safety-minded also being a differentiator versus some of the more open-source models potentially.
J
Joe Weisenthal1:07:14
Yeah, you could see it — I don't want to be cynical. Like, how — yeah, I mean I get that. But the question is like, does the non-safety-minded lab or does the less safety-minded lab get to advanced capabilities faster? Yeah, right. And so I'm not totally — yes, we would all love to drive the most capable the most safely. Yeah. But the question is like for a customer prioritizing the most capable. So that would be some cutting-edge thing. Yeah. Is the course right? Like, is everyone — I don't know, is it like some car that has an insane 0 to 60 versus the Volvo. That's what I'm saying. And does the customer keep giving business to the firm that delivers the fastest 0 to 60? If the company that got the fastest 0 to 60 did so by allocating fewer resources to safety research, is a big question of mine. And then I remain — he talked about the importance the company is going to see the sort of alarming data first. And I sort remain questioning whether the people looking at the alarming data actually share the same view of what alarming data is relative to all people, especially given what we know about the — relative to the shrimp eaters, the relative monogamous partner-havers, etc.
T
Tracy Alloway1:08:35
No, seriously. Like, I think your question — are you hiring more normies? — a pretty important question.
J
Joe Weisenthal1:08:41
And then obviously the political — I don't have a ton of confidence in the political environment. And I think look, the fact that if the research goes wrong, that there is a prospect of this technology really being very devastating to humanity, even setting aside jobs, is something where it's like, wow, you know, this is not a normal technology. This is not probably software, not like a solution that we have on. I just go back to the Terminator human extinction boom from day one. And as an answer to your question, there's like — they see it in the training process that AI models do these things, such as say, I'm being trained by an observer right now. Therefore I'm going to give this answer. I'm going to attempt to blackmail. It's not like very prevalent. But these are not like — that sounds very sci-fi except that they actually see this property.
T
Tracy Alloway1:09:37
Alright. On that happiness. Shall we leave it there? Let's leave it there.
J
Joe Weisenthal1:09:40
Okay. This has been another episode of the Odd Lots podcast. I'm Tracy Alloway.
T
Tracy Alloway1:09:44
You can follow me @tracyalloway.
J
Joe Weisenthal1:09:45
And I'm Joe Weisenthal. You can follow me @thestalwart. You can follow our guest Jack Clark. He's @jackclarkSF and Peter McCrory @petermccrory. Follow our producers, Carmen Rodriguez @carmenarmen, Dashiell Bennett @dashbot, Cale Brooks @calebrooks and Kevin Lozano @kevlloydlozano. And for more Odd Lots content, you should check out our daily newsletter. You can find that at Bloomberg.com/oddlots. And you can chat about all of these topics 24/7 in our Discord discord.gg/oddlots. And if you enjoyed this conversation, then please leave a comment or like the video or better yet, subscribe! Thanks for watching.