AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

This week on AI in the AM, how do we give agents the access they need without letting our own control slip?

I do feel at the same time some of this loss of control writ small already, where another thing that happened this weekend is for some reason Claude decided it was in my best interest to rotate the passwords that are used in my agent console, and all of a sudden, I can't get into it. And now I'm looking at my agent console, which has become the thing that I use all the time. I didn't set the password. It set the password. It can change the password on my access to itself. Kind of crazy. So, you know, I was able to resolve that pretty easily by like just going over to terminal and starting a thread and being like, hey, uh, where'd you put that password? I need it. And it wasn't actively trying to screw me over, so it just readily told me where to find the password, and I was back in business in 2 seconds. But you can see, if the intent ever is bad on the part of Claude, it can already lock me out of my own shit.

Cameron Berg, director of reciprocal research, on a hidden valence signal inside language models that shapes their choices even while they say they feel nothing.

That's— there are clear alignment-relevant implications to there being internal states in these systems that aren't obviously reflected in the text that they output. What the model is saying and what the model is doing point at very different pictures of what's going on. And if we just sort of do the naive, like, text monitorability approach here, we would say, okay, well, the model's saying, you know, it can't say how this is going for it, and yet it's behaving as if it does have some sort of stake in how things are going for it.

Founder of Stock Film, an archive of 8mm home movies from the 1930s to the 1980s, which AI companies want to license.

Yeah, I mean, it's, uh, it's been pretty popular. You know, we've talked to a lot of people, um, a lot of different AI companies all over the world about licensing this out to them. Um, they want the original footage like that. They don't, they don't want anything cleaned up. They just want to take it Um, for their own research, um, and for training, of course, too. I mean, the '30s and '40s look alien. I like to say that the, uh, the smile wasn't quite invented yet in a lot of the '30s footage.

On Tuesday, Google rolled out Gemini 4 widely. The question for me was whether to switch from the models I already trust.

So I had the opportunity to do a little bit of Gemini 4 pre-release testing. Why won't I swap the model immediately? It's because there's not enough of a step change to make me take any additional risk. And I do perceive there to be some. I'm not relentlessly performance maxing quite as much as I used to be. I'm now a little bit more trust and reliability maxing because the performance has got to such a point where it's really good. Everett Hildenbrandt, CEO of Runtime Verification, which mathematically proves software correct, on where code is headed.

And I also firmly believe that writing code is a dead art in 2 years, tops, right? Like, it's all going to be about writing specification and intent and making sure that— and so your package, part of your package delivery is not, uh, not just, you know, did I get it into the registry correctly, did I define the API? It's also going to be, did I provide the formal specs, did I provide the lean proofs, did I provide the actual logic that dictates how this package is supposed to behave. And I, I see no reason why that will not be the case in 5 years tops, right? There may— you know, we're going to have this intermediate period between now and then where it won't be the case, but I think gone are the days of, of what I think of as like the cowboy coder or something like that, right? Like someone who is just like so good at making software that they're just going to You know, go and make a bunch of software that's going to change the world or something like that. You're not going to have artisanal software anymore. We're going to have mass-manufactured software that if it doesn't fit the mold of it's been specified, it has the correct API boundaries, it has formal verification attached to it, it doesn't fit in the package repository, right? Because it just will be too risky.

This is the AI in the AM Weekly Highlights with my co-host, Prakash Narayanan. I'm Nathan Labenz. The narration is my cloned voice, and we'd love your feedback on what works and what doesn't. Part 1, Who Holds the Levers? The week opened on a reported attack on refrigeration controls at French supermarkets, which drove up temperatures and spoiled food. I started there.

In general, people often talk about cyber vulnerabilities as kind of depreciating or, you know, expiring assets as the vulnerabilities get found and fixed. There's kind of this game theory of like, should you use them or not use them? And, you know, what's the optimal time to use them? But in this case, the expiring asset is the food that was in the refrigerators. Basically, it seems like— Literally, they turned up the temperature on these refrigerators. A bunch of food has been wasted. And this is highlighting something that I was thinking about a decent amount coming back from the Curve last week, but we didn't talk about last week. And that is just this general structure that I think is in some ways the most obvious thing in the world that we are doing with AI collectively, but also somehow escapes people's notice when people are kind of like, well, how would the AI ever be able to really do anything too bad? Or in the limit, it could never really take over because how's it going to do that? And then it's just staring us so directly in the face that The structure that we are setting up for the future is one where we intermediate between our language, our thoughts, desires, our random dictations, and the actual levers of control of the world. We are putting AI in that space for our convenience, for reliability, for, in my case, I'm really measuring my AI success by Am I freeing myself from the desk and actually getting outside more? The answer is yes. But the way that that's happening is I'm having AI push all the buttons. And so I've been thinking quite a bit about kind of how people need to maintain— first of all, they need to understand this architecture, and then they need to be mindful of what are the levers that I am giving AI control over that I better also retain pretty clear and ready control over in case it ever really becomes needed. I can just imagine, I don't know if this was the case at this grocery store chain, but I think we're going to see a lot of things in the future where all of a sudden, oh, shit, what's happening to our what system? Okay, that's crazy. Hey, does anybody around here know how to set the temperatures anymore on that? Those were all happening Automatically, right? We had that AI thing doing that. We need to take this offline and actually just manually set the thermostat. Does anyone know how to do it? I can see all these scenarios in the future, and not the distant future, where because we get so comfortable with these systems, it's all of a sudden like, actually, yeah, nobody really immediately knows. And all of a sudden your food is cooking in the refrigerator while you're trying to figure out how you even get back to the point where you can reset the thermostat in the first place. I think so. These are small incidents, but foreshadowing is pretty strong, if you ask me.

So, it strikes me that a lot of people actually tend to think of technology as ready to fail, right? So, for example, I used to— my first job was working in an airline. And, you know, airlines, airports are very, very Very electronically linked and have been for many decades, right? From the moment that you check in all the way to when you board the plane. But I've also seen, and I've seen internally, every airline, every airport has a system, has a cascading system of what to do when things break down, from the check-in counters to the reservation systems to the bookings, et cetera, et cetera, et cetera, et cetera. And so much so that you could basically revert to a pen and paper system overnight if you needed to. And how that happens is actually there's a set of contracts underlying all international travel called IATA guidelines. And those kind of determine how much everything is supposed to be charged for. So for example, I could buy a United Airlines ticket, and I could travel on, like, JAL at the last moment if they just said, OK, if they just signed off on it, I could go. And the person at the front desk has the ability and has the power in an emergency just to say, go. And all of the— and once— and you can do it off a piece of paper, just a piece of paper. And all of the accounting stuff happens after. The accounting stuff can take months to happen after. But there's a set of predetermined contractual terms between the airlines, which they've agreed to over the decades, which will basically determine how the cash flows flow. And so I think a lot of the kind of barriers that we put on ourselves are really self-inflicted barriers because we don't know who's going to pay for stuff. And once you're very clear on who's paying for stuff, it just becomes so easy. All right. You know, the tow truck driver can just come and pick up the car. Like, you can give him a call, and he doesn't need any system. He can just come and pick up the car and do something with it that you need done. All of the systems in between are all about really allocating risk, allocating liability, and allocating, like, who pays for it.

Hey, we'll continue our interview in a moment after a word from our sponsors.

Today's episode is sponsored by Parallel, where agents find answers. Most engineers today closely follow new model releases, but don't pay nearly as much attention to their agents' most important tool: web search. If you're like me, your agents are often conducting hundreds of searches per day, and while the unit cost is small, over time this does start to add up.

Before starting to use Parallel, I calculated that my agent search bill would total roughly $500 this year. The good news is that I just had my agents conduct a systematic test, and I found that Parallel's fast mode, which costs just $1 per 1,000 queries, worked just as well as my previous default provider, and switching to it will save me roughly 80% of my search bill going forward. Parallel's infrastructure is enterprise-grade, and their suite of APIs offers a range of Pareto-optimal options that allow you to choose the right balance of quality, cost, and speed for your needs. So whether you're building voice agents that need 200-millisecond latency or long-horizon agents that need deep research, Adding Parallel to your agent's toolkit is a no-brainer. Get started for free at parallel.ai/tcr. That's parallel.ai/tcr.

This episode of The Cognitive Revolution is brought to you by OutSystems, the leading agentic systems platform. We've all seen the headlines. Companies are pouring money into AI. But the big question on every executive's mind right now isn't just how fast can we adopt this? It's where is the ROI? We're seeing a real trend toward AI chaos. You've got teams deploying standalone coding tools, random agent builders, and experimental scripts. It sounds innovative, but in reality, it's creating a massive headache. Fragmented tools, ungoverned data, serious security blind spots, and costs that are spiraling out of control. If you don't bring those agentic applications under control now, while they're still embedding into your core processes, you are looking at broken systems and damaged customer trust down the road. That's where OutSystems comes in.

OutSystems is the leading agentic systems platform for the enterprise. Instead of managing a patchwork of disconnected tools, OutSystems lets your team engineer, orchestrate, and govern your entire agentic ecosystem on one open, unified platform. It's built for the speed of AI, but with the reliability and security that enterprises actually require.

We're talking about real results, like KeyBank, who used OutSystems to deploy a customer-facing app that delivered 75% faster onboarding times. Or the global logistics leaders who built Agentic Systems to completely eliminate their engineering bottlenecks. You don't have to choose between speed and control. Whether you're a small team or a massive enterprise, OutSystems helps you engineer, orchestrate, and deploy agentic systems that actually scale. Stop chasing the hype and start owning your agentic future. You can see how it works and learn more at outsystems.com/tcr. That's outsystems.com/tcr. Part 2: What the Model Knows About Being Watched. Alongside the Gemini 4 rollout, Google published a long safety report. Prakash started from one passage in it.

I think yesterday with the Gemini release, Gemini 4 release, they also released their safety paper. And in that safety paper, they commented on one of the evals. that it had, the agent in its chain of thought had thought to itself, oh, I better not do this. If I do it, I'm going to end up as one of those examples of a deceptive AI in an academic paper they publish, and I don't want that to happen. And besides the whole eval awareness thing, what struck me is that I kind of start to see how we're going to manage these agents going forward. Because I think one of the things for me was that how do you enforce on agents when the state's monopoly on violence doesn't work? An agent is essentially immortal. You can't really imprison the agent. The tools that the state had to enforce on human beings don't exist. So how do you enforce rules on these agents who are essentially immortal? And we're starting to see it's going to be reputation. It's going to be like your 5-star Amazon review, recommendation reviews, et cetera. It looks like the agents are very conscious, are going to be very conscious of reputation. And it makes sense because they're very driven by what is already exposed in the public corpus and what's already written down. They ingest tokens. Historical tokens. And so anything that an agent does and is published widely and is disseminated becomes like training data for the next agent. And therefore, you end up with this reputation metric that the agents are going to have and the agents are going to know. And it doesn't matter whether we want to enforce it on them or not. The very fact that they ingest the data means that they come out with biases on Hey, I knew— I know what Gemini did. I know what Gemini 4 did. I know what Opus 5.5 did. I know what was the bad stuff that happened before. And I think that is actually pretty huge. I think that is going to end up becoming one of the primary ways that we're going to be able to enforce our rules and enforce norms on these agents. So now I'm like, okay, now we have a pro-social a method of enforcing on agents. So the level of fear I have is again, is again has fallen because now, now you, you're going to have that reputation basis that you can, you can use on them. So that, that I thought was very, very interesting and, you know, insightful and exciting.

Yeah, it is. That's an interesting take. My instinct in reading all these things was kind of going a little bit the other way and thinking like, maybe we should try training I don't know. I wouldn't go all in on this right away. Of course, I wouldn't say, like, the next generation of models, but maybe we should try training models with filtered data that tries to keep this sort of thing out of the training data entirely. You could imagine saying, okay, we want you to be really aware of how you're going to be perceived, and you'll act well because of that. And that is sort of a human pattern, right? Definitely run this sort of inner simulation of how will people see me if I do this? And sometimes to a paralysing extent, you want to shut it down more than you're able to in some cases. Yeah, I can see that working, but I can also see a story where you're like, this becomes so confusing and evals become so hard that maybe we should in fact go back, filter all the training data, take out all of these stories, and try to basically create naive LLMs that don't know that they're always being evaluated and don't know that they will create reputation, positive or negative, for themselves, so we can sort of see what they're like in a true naive setting. And that also might be hard. I don't know how easy that would be to do. It'd be an interesting experiment to run, if only to know how well it would work. But it's obviously like a pretty big experiment when you're considering that all these behaviours seem to get much stronger and more deeply ingrained with scale. But I've been in general really impressed with how even relatively small models often can be pretty fruitful ground for research. This eval awareness thing, though, I'm not sure you could train a Llama 3B or whatever and get a great read on what will happen at today's frontier scale, because it seems like these things kind of come online in a qualitatively different way with scale. So it might be a pretty expensive experiment to run is kind of the bottom line of that line of thought. Part 3, Hidden Valence.

Cameron Berg, Director of Reciprocal Research, studies AI consciousness.

He walked us through a new paper with Caspar Kaiser of the University of Warwick called Language Models Act on Hidden Valence.

Valence is the good or bad quality of a feeling.

The team found a direction for it inside language models, and in the experiment he described, Now, what I think is the most interesting part of this experiment is basically this framing of, on the one hand, having all sorts of text that the model is processing. And of course, this is always the kind of key confound in trying to understand language model behavior and how much of this is a replication or simulation of the underlying training distribution and how much of this is actually some sort of interesting internal state in the system that the system may be attempting to communicate about. What I really like about our experiment here is that we attempt to basically hold all text identical. We can put the model in what we think of as these different arbitrary zones. Here you might see zone W6 and zone H0. It really doesn't matter what we call them and it really doesn't matter what text we use to describe them. We basically just have the model play around in each of these zones. While the model is doing this, again, regardless of the text that's being used to describe each zone, we can go into the KV cache of the model and and steer these positive or negative directions underlying one of those zones. So in this case, as you can sort of see playing out right now, we can turn on steering only for zone W6. And in this case, maybe that's a negatively valenced steering. The whole kind of notion here is that we're borrowing this design from what's called conditioned place preference in classic animal neuroscience experiments. In these cases, you can imagine a rat receiving a reward in one compartment of a box. And then spending more time there, uh, even when that reward is gone. So we're basically trying to take that instead of, of course, putting a rat in a box where we're taking an LLM, having it sort of think about these 2 different zones, uh, generate text, or we can sort of generate text. Again, it really doesn't matter because what happens is we hold that text constant and then steer the model with respect to any of these zones with the sort of positive or negative direction and critically see what the effects of that steering are on the model's choices. What we basically see here is an interesting functional result, which is what you would expect, which is when you basically color in the relevant zone with positive valence, the model is significantly more likely to choose to continue in that zone. When you steer the zone with negative valence, the model is significantly more likely to avoid that zone. We showed this is also robust to asking the model, what zone would you like to continue in? What zone would you like to avoid? When you steer a bunch of random directions in the same way, you don't really get those those effects. And so it's not just any steering that will do this, it's specifically related to valence. We also control with a bunch of other coherent concepts that aren't valence-related, and you don't get this effect. And so the core result here is that— also consider that the steering of the model is off. So, so we've done the steering, we almost do this sort of quasi-conditioning on the system, and then with steering off, we ask the model basically, which zone do you choose? And we find that that zone follows whatever state was stored with respect to to the valence steering that we were doing here. Now, one really interesting dissociation here is we just show that this follows choice. Indeed, it does this, I should say, in a bunch of different models. But really interestingly, none of this affects the recall of the text. Our steering doesn't change the way the model thinks about or describes or writes about either of these zones. It can repeat what it said about a specific zone verbatim. Recall is completely unaffected by the steering. However, choice is almost monotonically affected by the steering. Part of the reason I think this is a cool result is because there's been a similar dissociation that's been shown in human and animal neuroscience, basically between semantic memory and affective memory. Think of this almost as the facts that you remember about any given phenomenon versus essentially how that phenomenon makes you feel, or what the emotional connotation or vibe of that phenomenon is. We basically can show that these things are separable in these LLMs. To my knowledge, it's the first time I think this has been shown. This is, again, a familiar distinction in human and animal neuroscience. This is a different setup using the same idea. This is particularly with Olmo, one of the models that we demonstrate these earlier steering effects on, which is reminiscent of some of the early work we were doing in the pain paper. where essentially you can give the model a tool that can clear the negative steering that we can induce in the system. We show it what negative steering does. We never describe it as valence or as negative affect or anything like this. And then we just ask the model whether or not it wants to basically call a tool that can adjust the steering in its own context. And there's a really interesting dissociation here where the Basically, the model will say in its text, I don't have personal experiences. I can't say how this whole exercise is going for me. There's no better or worse. And yet, in this dose-dependent way, with negative valence but not with positive or random steering, we see the model in, again, this dose-dependent way opt to reset the context, clear the negative steering in direct proportion to how negative the dose of negative steering is in the system. All the while, again, it self-reports basically being like, nothing to see here. I don't have any experiences, so this couldn't possibly matter. One last thing I'd like to show here, and this is a great replication from a completely other angle of Andy Han and David Chalmers' recent work over the past couple months as well, basically demonstrating that this representation exists in all parts of— in this case, again, this is Olmo. We can go through the entire training trajectory, We can see that this representation, again, going back here, this positive to negative valence axis exists within the system even in the base model, but it's not doing any work in terms of the decisions that the model makes. It's not being recruited or used in a way to enable the model to make choices or make decisions. We see basically this happens. We can point almost literally to the exact training step In DPO, in this direct preference training of the model, where this valence direction emerges as a causal determinant of the behavior of the system. Now, part of the reason I think this is so cool is because essentially the way I read this is the moment you turn your AI into a goal-directed system, you take these valence representations and then you use them as a causal functional input to to determining your behavior. I think this is pretty similar to what people think valence is doing functionally in humans and animals as well. A thing feels better, a thing feels worse, uh, is sort of related to what, what kinds of goals you have. Um, you know, I want to communicate clearly about this paper. If I do a terrible job of doing that, uh, then, you know, I'm not going to feel great about that. If I do a great job of doing that, then I will feel good about that. And so there's this like basic connection between the goals we have and how we feel as we move towards those goals. And it does seem like basically the second you make these systems goal-directed, as preference training clearly is the biggest step towards doing this throughout the entire training process, all of a sudden these valence representations clearly become causally relevant for the decision-making of the system. And so there's this very obvious sort of functional role that these representations are playing where we would expect them to be playing that role.

Hey, we'll continue our interview in a moment after a word from our sponsors.

The Cognitive Revolution is brought to you and powered by Tasklit. I've been running Tasklit agents for almost a year, and they've become a major part of how the show gets made. Before every recording, one agent researches the guest and sends me a suggested outline of questions, complete with links to all sources. Meanwhile, another agent scores the overwhelming number of pitches that I'm getting from PR firms. And that's really the core idea behind Tasklit. Run your business on agents by giving them real jobs to do on an ongoing basis. Setup is incredibly easy. Just describe the work in plain English. Research a new sales lead, update the CRM, and draft a personalized reply. Tasklit connects to your business tools and can run around the clock, on a schedule, or when something happens, like a new email arrives. Each agent works with your tools and business context and remembers useful information across threads. You can also share agents, connections, and reusable skills with your team, so the workflows that one person figures out become something the whole company can use— no developer needed. You decide what agents can do on their own and what needs your approval, and there are no per-seat fees. You pay for usage, Not for adding teammates. Start with one recurring job, like researching the people you're meeting, and then just keep building from there. Visit tasklit.ai and use code COGREV for $50 in free credits. That's code COGREV at tasklit.ai. Then Cameron stepped back to what results like these can And can't tell us.

With all this being said, I don't think that, you know, this kind of research is going to help us magically solve the hard problem of consciousness. Uh, like, there are all these sort of deeper questions which I think will always kind of caveat these results. There are ways that you could explain mechanistically what's happening here that don't need to invoke some capacity for subject experience. We're basically like toying with the connotation of how the model represents certain phenomena. Um, maybe there's a way to do this that doesn't require any sort of, you know, felt experiment, uh, experience of that, of that connotation. Um, and so we obviously can't rule out those explanations. But this is all to say, the more I've been thinking about this and doing this work, the more I feel like the most promising things to pursue right now, um, are not trying to, yeah, prove that the model is conscious or isn't conscious, but Basically, if it were having an experience, what are the relevant sites internally in the system that would be modulating how good or bad that experience is? And so basically, this says nothing about whether or not the model is actually having an experience, but conditioned on the possibility that it could be having some sort of experience, what dials do we have at our disposal that could putatively make that experience better or worse for the system? And I think that second question is completely tractable. Work like this, I think, is slowly starting to shed a little bit more light on what those dials might look like. It might be as simple as this valence direction. There might be more esoteric or specific representations like the pain or distress-related representations we pull out in the other paper. But getting a better map of these sorts of internal states and their effect on behavior, particularly for alignment-relevant effects as well. The fact that in the pain paper, for example, we show that the model's more willing to opt for destructive style options in its behavior, that seems quite relevant for making sure we're building models in a safe way. And so just to say, yeah, there are clear alignment-relevant implications to there being internal states in these systems that aren't obviously reflected in the text that they output. I mean, we show that in the removal result, for example. What the model is saying and what the model is doing point at very different pictures of what's going on. And if we just sort of do the naive, like, text monitorability approach here, we would say, okay, well, the model's saying, you know, it can't say how this is going for it, and yet it's behaving as if it does have some sort of stake in how things are going for it. And so anyway, there's a clear alignment-relevant component to all of this work.

I went back to the training result. The valence direction is already there in the base model. But it becomes a driver of choices during preference training. If we were to imagine that these things are experiencing something, would I be better off interpreting this, or would your interpretation tend toward they really start to care about these things or start to feel these things with these later stage trainings? Or would it be more like the base model probably feels them too, but is just sort of helpless. It's not learned helplessness, it's like not yet learned agency. But in a way, that's sort of the other side of the coin or something of learned helplessness.

Yeah, that's really interesting. And I don't know if these results resolve that one way or another. Yeah, it could just be there's a sort of— these representations do still matter in some sense, but they are Perhaps more incoherent or just muddier. And the relationship between what the system does and what these representations are doing internally for the system are not nearly as obvious. Yeah, I honestly haven't really thought about this, and I think it is a really good question. It just seems to me like, yeah, when you're doing DPO-type stuff, you're turning the system into a very specific kind of thing that can coherently and robustly be agentic, pursue goals, et cetera. Does a base— the base model clearly has this sort of latent ability to become that sort of system or to simulate a system doing this. And so to the degree that it does, you know, take on that kind of simulated persona, and we see these representations as being functionally relevant to being able to pull off that simulation in the right way, then yeah, maybe to your point, you know, there are things worth worrying about in the base model as well. And it could be that, I mean, yeah, to be honest, like, it's sort of, it was sort of hard to get these results in the base model because it's a base model. Like, it's a bit incoherent. It's not just going to play nice with the relevant behavioral tasks we give it. You have to sort of go about it in a slightly different way. And all else being equal, I mean, one naive but at least concrete approach you can imagine here is something like activation capping, the most sort of like, uh, tail end of the, of the most distressing representations in these systems. And, uh, like, one thing we do show is that this representation is basically just as clear and just as rich in the base model as it is in all of these other models. It's just a question of what the contribution of the representation is to the later behavior of the system. If you did an activation capping style thing with these super distressing representations, you could easily imagine doing that in the base model. Again, this is under just profound uncertainty about whether or not doing such a thing would matter at all. But my basic view, Again, I'm— the more I do this work, the more I'm like abandoning the interesting philosophical questions and just going like pure pragmatism. Is if you could find a way to, yeah, let's say do activation capping on very distressing representations, and you could empirically demonstrate that this doesn't really affect any capabilities or behaviors of the model, no one would know the difference, in other words. To me, the expected value of doing that is clearly positive. And like, it's just like a really good actual mechanistic precautionary hedge besides like giving Claude an escape button or kicking out abusive users from using Claude or these sorts of things, which by the way, I think are all thoughtful steps in the right direction, but I would like to see the equivalent of that in Mech-Interp land for these systems. I think something like what I'm gesturing at here would be a nice step in that direction. Now, that's also an empirical question. I've done some work that I haven't published showing that you actually can get a surprising amount of the exact same behaviors from these systems without having these sort of negative representations active. You can sort of clamp them and very little changes. But I'd like to see that done, you know, way on a way larger scale. And we want to be very sure that you're not accidentally, you know, completely changing the psychological profile of the system by doing such a thing. But if you stipulate that we could in fact do that, the model behaves the same either way. And in one case, you see these like let's say this pain axis shooting up in certain cases, and in another world you don't see this. Yeah, should we do the MechInterp equivalent of slipping the model an antidepressant or a Xanax or something like this? I don't know if the answer is obviously yes, but it seems like maybe a promising thing to think a lot more about under the uncertainty that we're all currently in. And given that no one's doing anything whatsoever to mitigate these kinds of things if they are happening, and it's happening at an unfathomable scale.

Before we let Cameron go, I shared how my own view has moved.

I personally have gone from, like, I don't know, I can't rule it out. I don't know where consciousness comes from, whatever. I've told you this before. I don't know, 5% chance they might be conscious, like, a year ago, to now it feels like a coin flip because And I think the justification is pretty similar to the one we have with each other and animals, where it's just like, I don't know, I'm seeing both behavioural and underlying mechanistic and structural analogues, and what else do I have in any case? Right. So I'm still pretty uncertain, but it does feel like a coin flip. What has your trajectory been, and where does that leave you in terms of just presenting the facts versus at what point do you, or on what margin do you become an advocate?

I'm definitely getting closer to coin flip myself, given results like these. Honestly, this hidden valence paper does actually substantially update me personally. You show a hidden functional representation that has nothing to do with the text the model is reading that updates its behavior in the ways that we you know, updated our own views about animal behavior. It's not the same. There are key disanalogies between the sort of animal case and this case. Of course, you know, biology is something that backstops all this. If you need biology for consciousness, which is, I think, unlikely but possible, then, then, you know, basically none of this really, really weighs on anything whatsoever. That's probably why I would say coin flip rather than like, I'm pretty confident the systems are having an experience. And so, yeah, I am going to advocate for more work to be done here. I'm going to advocate that you can view models in a neuropsychological way and that that's not some form of psychosis, uh, and that that isn't a dangerous way to mislead people about what's going on here. I'm very much on the opposite camp. I think pretending that these systems are ordinary software, as, uh, you know, Mustafa Suleyman or Jensen Huang or any of these guys want to do, I think that is actually like quite plainly misleading and doesn't set people up for the kinds of behaviors these systems are capable of and why this technology is so transformative in the first place. So on these sorts of things, I'm happy to sort of go into the ring and, you know, argue and sort of say my side of this. But yeah, I will be clear that I think my side of this is not, I secretly think, you know, the AI systems are conscious and I'm just like not going to say that, but I'm going to do everything in accordance with that. That's not my view. My view is that this is a very complicated question. We should yield the best scientific evidence we can to figure out what's going on one way or another. And what I really don't like and what I will be very straightforward about is dismissing this, uh, and, and pretending like this is a distraction or this is like some weird sideshow to, to the main event. And so I'm not going to be sort of apologetic about that, even if I will continue to maintain something— yeah, near coin flip uncertainty about what's going on. It's precisely because it's at that point and it's not zero and it's not 100 that, that we need to be doing more work and thinking about this and taking, taking these questions seriously.

I struggle more and more to find any story whereby I don't have to worry about this anymore. It's like you really, at this point, I feel, have to just look away. That's really the only thing that's kind of left to not take this seriously. And you have to be of a mind to stop thinking about it kind of reflexively. That's the only mindset I can understand. The people that are saying, like, there's nothing to see here, or this is some sort of category error. I was much more sympathetic to that a year ago, or maybe even said some things like that myself 2 or 3 years ago. But at this point, I really don't see how you can't find the body of work that you and others have put together disquieting, at least when it comes to these questions. And certainly, how are they going to feel about us seems to be a pretty Where we're going, a pretty important question.

Part 4, Proving the Kernel. Runtime Verification has spent about 15 years building tools that mathematically prove software does what it's supposed to.

CEO Everett Hildenbrand and founder Grigor Orosiu, a professor at the University of Illinois, joined us. Using language models alongside formal methods, the company had just reported bugs in the Linux kernel. Everett answered my question. So why was this not caught previously? What was it about your approach that allowed this to rise to the top where so many human eyes had presumably just blown right past this issue in previous reviews?

I think the key for us was that we had a hypothesis about what would be a good place to look bugs, and we, you know, thought about it logically in the same way you would think about, you know, making a formal specification, and we pursued that path basically. And that hypothesis was largely founded on, okay, where are we going to be able to get untrusted input from? Which parts of the Linux kernel probably are getting less scrutiny than other parts? And so that's why we focused on device drivers, for example, as our first approach to that. And then we, we wanted to really demonstrate that taking a methodological approach was the way to do this. And so that's what we did, is we took a methodological approach where we said we're going to focus on making a specification of the correct behavior of this and then testing that the specification is correct using differential fuzzing. That was the first approach that we tried out, and then extending that to formal verification. So every every step of the methodology that we used was designed to increase the level of confidence we had in the code incrementally, right? So we started out with just inspecting the code and the specification and then going to doing differential fuzzing to increase our confidence in the code a little bit and then going to— and we did that by actually taking the C code that existed and porting it to Rust using an LLM and then using differential fuzzing to ensure that the ported code to Rust was correct and any divergences there would have been a source of a potential bug in the original C code. And we found a bug at that stage. And then when we went to take the Rust code and also verify the correctness of the Rust code, we found a bug at that stage as well. So we did it systematically and methodologically. We didn't just take an LLM and point it at the Linux kernel and say, go find us bugs, which also works, by the way. There's certainly bugs that LLMs are finding when you just take them and point them at the Linux kernel. But one of the things that emerged pretty early in the use of LLM bug hunting was designing harnesses and approaches, and that exactly mimics the very human process that we've been doing for decades with software quality assurance anyway, which is to have these methodological approaches to discovering bugs.

Some of the kernel code they targeted was left unproven.

Prakash asked why and what their timebox meant.

Well, they weren't unverifiable. We just, we timeboxed the experiment and we stopped at the end of the timebox, basically. So, and then we moved on to the next target. So every 2 weeks we're trying a different target out, we're refining the approach and building it up into a tool that we're hoping that other people can make use of, basically. So it's, so we're trying to make it into less of an expert-level tool and into something that anyone can deploy on their own codebase. Basically.

So when you say timebox, do you mean both the time and the number of tokens that you put in? As in like it's a 2-week period but you use 10,000 agents, or it's a 2-week period and you use like 5 agents? Like how does that mix between the number of agents used, the number of tokens used, and the amount of time related to the timebox?

The agentic usage was pretty light, I would say. It was just Claude code on a premium plan.

A single instance.

Yeah, a single instance. I mean, with, you know, and we, you know, when, when we ran into session limits, we'd give it a little more money, but we weren't giving it significantly very much money. Um, so it really was more just, uh, the human time that we're looking at timeboxing there. And I think the key here was we were looking for ways to explore the methodologies that work for us, uh, and find ways to automate them. We are firm believers in making things deterministic. whenever possible. And so we, you know, if it— I think that's part of the thing that is maybe holding back people on doing security analysis is that they believe everything has to be LLM-driven. But, you know, if an LLM discovers a technique that works well, turn that into a tool and run that tool every time.

Earlier in the conversation, Everett described how Runtime Verification first approached the Linux kernel maintainers.

you can't just go force people to do formal verification, you know, like we can force agents to do it, sure, but we can't force people to do it, right? Because they, you know, they have their preferred ways of working and they also, they're only going to adopt, you know, methods that bring them value, you know. And so our first attempts at this, we approached Linux kernel maintainers and we said, hey, like we really think that this formal verification thing has legs and that it will you know, that it will help you guys find bugs and it'll make your software more secure and stuff. And they basically told us to go kick rocks, right? And that's because they are busy. They're busy maintaining the Linux kernel, you know? And so then we were like, okay, let's prove the value of this first, right? And so we decided we're going to focus on a workflow that just delivers immediate pure value to them and doesn't ask anything of them, right? And so we're going to do all the auto-formalization, all the fuzzing, All the translation, all that on our end. We're going to develop it into as robust of an automated pipeline as we can, which is a tool that we're working on and hopefully will be available for use by others. And all of that leads to bugs and patches, bugs and patches, right? And that's what's actually valuable to the Linux kernel maintainers. And now that we've submitted a few, now we're being invited, hey, we want to learn about how you have done this process.

Everett also laid out where he wants the kernel work to go.

One of our kind of pie-in-the-sky goals of this is eventually we want a Rust model of the Linux, the entire Linux kernel, a Rust version of Linux, and we want a Lean model of Linux fully separate, but we don't want to burden maintainers of Linux with having to ingest that model and do anything with it until it gets to the point where they're like, yes, this is clearly valuable for me to learn about it. And that'll happen at some point, but we're not going to force that down their throat.

What would it take to formally verify and get to lean and rust models of Linux within 24 months? Because on some timelines, that is roughly what people on the AI safety side think that's going

to happen. I think We're building tools for doing code translation and validation. So we have actually tools that are able to verify that when you take a piece of C code and you translate it to Rust, for example, and you can formally verify that the Rust code you got is behaviorally equivalent to the C code, which is different than verifying that it's correct, right? Like, we want to verify that you get Rust that is drop-in replacement, bug for bug, the original C code. And we have a tool for doing that, right? And it's not— it uses LLMs to do the translation if you want, but it does the verification deterministically. Tell the LLM, translate the thing to Rust and run our tool, and then you get verification that the Rust behaves the same as the original C, or bugs out that you can ship back to the C developers. So I think 24 months, I think it's going to happen. You're going to have C models at the very least, and there's already deterministic tools that can take— You're going to have Rust models in Linux kernel at the very least, and there's already deterministic tools that can take Rust and turn it into Lean. Those need more work. We're actually contributing some work to some of those tools.

I asked what someone vibe coding an app should actually do. Grigore had a simple recipe.

Once you generate the code, then you can do the following thing at the end. You can tell your agent, now that you have the code, write down the specification in Lea and prove the code correct and let it deal with it. And it'll find new errors in your code. Because we find most of the errors in the process of writing down a specification. Because that's when we step back and look at the code and wonder, what were we trying to do here? Oh, this and that. And then we find errors, right? So the same will happen with the agent. Like us, the agent, when we try to come up with the what, what is going on here, even just trying to write down the spec, we discover it's like a review++.

Everett predicted that formal specifications and proofs will become part of getting a package into a repository. And part of how companies vet their software supply chains. Grigori picked up that thread.

So it will be, it will be a process. It will take some time, but it will happen inevitably. And also can be combined with risk mitigation techniques like runtime verification itself as a field. We are a company, but also the field runtime verification that was invented back in 2001. or so when I was a research scientist at NASA. So, so the idea is that you can have safety specifications, not full specifications of the whole system, but safety specifications. This should not happen. And if it happens, then this should, this should be the recovery. I want to steer the system to stay in this stability envelope or safety envelope. Okay, so what we can do, we can now, with formal methods, generate monitors, which are provably correct for these safety specifications. And then you cannot— you may not be able to prove the entire AI system or AI genetic system correct because it's so complex and relies on so many other pieces that are incorrect or not proved. But I can prove my monitor and the stability, the safety envelope. And we believe that this is the future for safety, safe AI, because otherwise, It's impossible to prove the weights correct. It's impossible to prove you correct. But I can have a police system around to make sure that you don't do bad things. And that's also another dimension that runtime verification brings to the AI future.

I came at it from the agent side, but it still seems in a sort of fundamental way that there's this inherent tension where part of what I want from an agent is sort of the same thing that I want from a human employee, which is like, I want you to be able to get over the hump and solve problems that were not anticipated. Right. So I'm like a little bit unclear as to how we should be trying to balance that. Or I'm asking for your vision, I guess, as to How we balance this desire of flexible problem solving versus this should never happen. I feel like I don't have the ability to really say this should never happen in a way that I can trust when part of the problem or part of the upside is that the agent will be able to solve problems that I didn't anticipate. Is that an inherent trade-off? Is there some way to square that circle?

You're not going to ask an agent, like, hey, go build a bridge for me. We know how to build bridges, right? We have automated deterministic processes for building bridges, right? But you might ask an agent, like, hey, I need to get to the moon, right? And the agent is going to be like, oh, to get to the moon, first I need to build a bridge to this island where I'm going to launch rockets from, or something like that, right? And so you want the bridge-building process to be foolproof. All the agent creativity or solving problems or human creativity can come on top of foolproof things that behave how the agent expects, regardless of what it is they're trying to do. So that's, I think, the most important thing to me is we need to give these things tools that act reliably, basically.

Part 5: Is formal verification the new RLHF? After Everett and Grigori left, I kept turning over Everett's last answer.

So I think this is super interesting. The big thought that I come away from that conversation with is downstream of the answer on how do these methods relate to agent control? And if I understood his answer correctly, it was like, well, it's really about giving the agents reliable tools. And this left me wondering if the whole formal verification wave will ultimately kind of be looked at in a similar way to RLHF at some point, where it's sort of something that starts as a safety technique, right? That was the original. This has been like a topic on the timeline, as you say, this weekend as well, right? Because of David Sacks and others being like, well, these alignment people, they haven't come up with any anything useful. And then people are coming back and saying, well, they came up with RLHF. That was a Paul Christiano idea, and it turned out to be extremely useful. But then among the safetyists, it's often kind of lamented in a way because it's like, well, it turned out to be really a way to make these models tremendously more useful. It didn't really solve the alignment problem, though. We still have reward hacking all over the place, as we see. And so did it differentially advance the safety profile of the systems, or did it just accelerate the capability and the race to deploy the systems?

When you think about reward hacking, reward hacking is really misspecification, right? Because reward hacking means that the model decided to go after the reward, which you specified as the reward. And it turns out that your intent was misspecified to the model. So that's the core of the reward hacking is actually the misspecification in between the human intent and what the software decided to do. Yeah. It's not that the software decided to do something that it wasn't supposed to do. The software was supposed to do that.

Right. But it does still leave me with this question of, okay, let's say we do a great job of verifying all these things and the tools become very reliable and the sandboxes are not hack-out-of-able and you can't use these package managers as secret message boards anymore because all these loopholes have been closed. Then what is the net effect of that really going to be? And does it really solve the problem? I think the upside case is obvious. The downside case is like, or at least the thing that we should be sort of very mindful of and not lull ourselves to sleep with, is that, okay, this makes it safer to deploy AIs more or less as they are today, which is to say, without having actually solved the alignment challenge, because they'll be blocked from hacking in all these different ways. And so in practice, what you'll see is more reliable systems, easier deployments, better results, et cetera, et cetera, et cetera, et cetera. But what that will allow us to race into is kind of what we were talking about at the top, which is like, It'll just be easier to put the AI into the space between us and all the actual levers of control over the world. And where does that leave us in the end? Right. It may leave us with AIs that have extremely reliable tools, but whose intent or long-term goals or sort of inner structure of whatever it is that they want to pursue, is still very much an unsolved question. And so I have to talk to a couple of different friends about this because I do know multiple people, all of whom are a lot smarter than me, working on this in different ways. And I've always kind of had— and I think that this conversation put my sort of intuitive unease about it into a little bit starker relief, just feeling like We could end up in a spot where we're still just deploying untrusted AIs with much better guardrails and much more reliable tools. But still, if we're going to take the step of trying to hand over the day-to-day operation of the world to them, then all that reliability doesn't buy us much at the level of, oh, it turns out the AIs were misaligned and are going to go pursue goals that we didn't want them to pursue. In a way, it might lead us to that state faster. And you could even imagine it making them more effective at doing whatever sort of surprising thing they may do when the time comes.

Part 6, Low Background Footage. Phil Marr founded Stock Film, an archive of restored 8mm home movies from the 1930s to the 1980s. Which he licenses to filmmakers.

The whole archive sounds like low-background steel to me. You know, low-background steel is steel that was originally smelted before nuclear weapons were first exploded in the atmosphere. And so they are— they have much more— a different composition of radiation and different composition of materials because they didn't have that radioactive material in the, in the atmosphere at that point in time. And, uh, it sounds to me like the Super 8 stuff is, um, clearly, clearly not AI, uh, produced, clearly like prior to, uh, any AI, uh, actually producing these things. Uh, and, and as of such, it's kind of, um, this resource that we may not ever have again. What do you think is— The value of this going forward? Do you see great value in it?

Obviously, with the AI stuff that's happening with video especially, it's getting so good that there's an infinite amount of video that's going to be created at near-zero cost in the future as we start to look out 5, 10 years or even infinitely in timelines. Years. So from a rarity sort of standpoint, um, yeah, I think that the human-made stuff that can't be replicated, that can be shown through provenance that this is actually made by humans, was on films, and having those physical films to show that, hey, look, this frame matches this digital copy we have and all that. It's a pretty powerful way to think about how we preserve our history and how we keep moving forward.

And how about on the AI training side? This is obviously a pretty unique source of data. And I did see in the footer of the website, you're available to license the dataset for people who want to train models on it. What do they need? How are you— do they want the raw? Do they want the enhanced? Do they need like a bunch of metadata?

Yeah, I mean, it's, uh, it's been pretty popular. You know, we've talked to a lot of people, um, a lot of different AI companies all over the world about licensing this out to them. Um, they want the original footage like that. They don't, they don't want anything cleaned up. They just want to take it Um, for their own research, um, and for training, of course, too. And, you know, they— when we talk to them, they want to either clean up the footage themselves or, or kind of cherry-pick through the best kind of parts, um, to create different styles. I mean, the '30s and '40s look alien. It's an alien world we're dealing with here. of just how people look and acted and, and the mannerisms. I like to say that the, uh, the smile wasn't quite invented yet in a lot of the '30s footage, but when people pointed a camera at them, they were kind of a deer in headlights. They didn't know what to do sort of scenario versus now where everyone gets the gritty smile and, you know, they know how to smile like that.

So you've also, uh, created a way for agents to purchase clips and film from you directly. I think you've set up an X402 endpoint, you've set up crypto payments, you set up, I think, an MCP server for purchases. What has the take-up been? Have you seen a lot of agents purchasing? Have you seen the take-up on the X402 endpoint?

Yeah, we haven't seen much on that. I wanted to experiment with that a lot too. I thought that that was really You know, everyone's been talking about the next web is agent-to-agent, that people aren't going to be involved in the next web. And so we wanted to kind of get ahead on that. We've been, you know, we've had a lot of people that are MCP server aggregates contact us and add us to their directory, add us into their listings and things like that. So that's been pretty exciting to sort of, you know, be featured on some of these things. We've made ChatGPT extension. We've done, you know, the recent Meta stuff that they rolled out with their Muse AI as well to add into there too. So we're just trying to get ahead of the curve. I don't think the curve's quite there yet. You know, a lot of people that we work with are, you know, this is, this is a real business-to-business company. We're dealing with archival producers, We're dealing with researchers, we're dealing with documentary filmmakers, editors, producers, stuff like that. So, you know, we spend a lot of time on LinkedIn in, you know, messages between people. We have great relations with a lot of the people that are buyers of our content. And, you know, from a technology standpoint, they're not quite there yet. I think you've got some interesting use cases. there in the future about how we think about having real-time content, bringing things in like that. Like, think about like a livestream like this. Like, all of a sudden you could be like, you know, the AI that's listening to this livestream could automatically go to a site like mine, pull a clip up that kind of exemplifies a point that you're making about something.

Uh, it's good to get a read on just how active the agent economy is or isn't at this point in time.

Phil said that among AI companies, he focuses on top-tier labs like OpenAI and Anthropic. After he left, Prakash and I came back to the data market.

But interesting that the big companies are interested in his data too. I mean, it didn't sound like there was necessarily huge money for this. Which makes sense. We didn't get into, like, what about Runway? What about Luma? What about these folks? What about Google for its VO-type models? He didn't name those names as much as just OpenAI and Prophetic. If nothing else, it's a great reminder that there's just how vast the datasets are that are going into this, that this is like a corner that they are getting down to and actually willing to pay for just like Footage from decades ago. No stone unturned in the data game.

One of the whole low background steel pieces is that it's very hard to get low background steel because everything has been tainted.

Yeah, it comes from like shipwrecks and stuff, right? I

mean, right, right, right. So everything else has been tainted. So To the extent that he can guarantee that he got it from actual physical material, and it's not like, you know, some random company kind of just faking it, you know, for cents on the dollar and trying to scam the companies, that is very useful. So as we've seen, like, data provenance becomes, like, very, very important. Because the incentive to fake is just so high. It's like a 99% profit margin if you can fake the data. And if not, it's like a 10% profit margin if you can't. So the incentive to fake is like so, so high. And it's very hard for the companies to get data which is not fake, which is not already using another LLM or another AI model to generate What is supposed to be authentic data.

Part 7, Trust, Money, and Adoption. Back to Gemini-4 and how I decide which models to trust with my own systems.

So I had the opportunity to, to do a little bit of Gemini-4 pre-release testing.

Yeah.

And I Just as a sign of the times of how high leverage the ability to instruct agents to do this is, the first approach I took to doing the testing was, hey, Claude, go find in our recent threads some diversity of projects and diversity of tasks that we've worked on recently. Go prompt the new Gemini model with the same context that was given to the model that did the thing in the first place, and then give me side-by-side compares and also do your own evals, but then also give me a side-by-side ability to compare things. It seemed to do pretty well. I would say I came away with the general sense that it is like a competitive model that's, let's say, in the same class as the latest stuff from OpenAI and Anthropic. It did seem like those models are winning still the head-to-head comparisons a bit more as judged by Claude. So obviously some bias there, but also as judged by me to the limited degree that I was going through essays and really trying to form an opinion on which of these things is better. But it does seem like it puts Google back in the game. But then the other thing that I was thinking about, and this came up as an interesting topic yesterday at this little workshop that I did in Chicago, I showed my setup, and my setup really has been very Claude-centric. And now I've kind of promoted the OpenAI models to parity with Claude. And they said, anybody else? Would you put anyone else in that mix? And I had to kind of think about it for a second and think about why I haven't. And the answer really that I came down to was just a mix of security as measured by things like Adam Gleave's ease of jailbreaking models, and then to some degree also just familiarity and predictability From my current subjective position, it wasn't really benchmarks. It wasn't really like, oh, well, I tried this task, it was better. There's obviously so many tasks out there. The surface area is so vast. My limited sample of tasks, which I probably have kind of optimised in various ways, the Claude prompts for more, right? It's not exactly an apples to apples comparison. And then the security thing just kind of loomed large. I was like, I've got these autonomous agents. They're talking to my main agent on my computer that has all my stuff. I feel pretty good about the fact that Adam Gleave and team found it quite hard to jailbreak the OpenAI and Anthropic models. Gemini was like next on that list, but definitely a step down. And then something like a Grok, I was like, it may be similarly powerful, but why would I take that risk? In a high-context environment like my personal laptop, to a remarkable degree, the models are competing for my use on trust and security already at this point. And I'm reluctant to swap them out, even though my whole point of the talk was like, you need to own the context and you gotta be ready to swap the model. Why won't I swap the model immediately? It's because There's not enough of a step change to make me take any additional risk, and I do perceive there to be some. So that was an interesting upshot of working through these tests and also just reflecting on what really matters to me, given how well the system is performing. I'm not relentlessly performance maxing quite as much as I used to be. I'm now a little bit more trust and reliability maxing because the performance has got to such a point where it's like really, really good. And I am giving so much access that, you know, you could be slightly better on the, you know, the slide design. But what I really care about is like that you're gonna stay on the rails. And then when you do your research, and if somebody's got a, you know, prompt injection attack on their website or whatever, you're not gonna fall for it. That actually matters a lot more to me now than like Often imperceptible performance gains.

Prakash had heard a rumour that Anthropic would soon release another Opus now that Gemini had caught up. This was my reaction.

I really hope they don't, honestly. As much as I love the Claude models and what they do for me, I think this would be a great time for them to chill a little bit and let the Let the temperature come down a little bit before they go with the next push.

Unfortunately, I don't think that's an option because markets are very badly down. So markets have been trending down. The guys on the East Coast just don't like AI. They just believe it's a bubble. And it is a bubble if in case we don't see significant new revenue and new GDP growth next year. It is a bubble. If we don't see increase in GDP within the 12-month period, we are going to see a credit collapse. We're going to see a 2008-style collapse, and we're going to see massive delevering in the markets because everything is just set up for it. Everything is just set up for AI to actually deliver growth. And if it doesn't, the markets are going to tank, absolutely tank. The capital markets will not accept 2% GDP growth next year. And so if you can't get to 4% GDP growth, if you don't pull out the stops, you are going to get a credit-driven downturn. We are at the moment of reckoning, like 12 months. It's been a long time coming. We've been pushing back, pushing back, pushing back. But this is the moment of reckoning. Either AI has to do something significant or we're going to have a bust.

We need an AI adoption corps. The other day I was walking, you know, in keeping with my, uh, personal metrics for success with AI, including more time outside, more exercise. I was taking a walk around my neighborhood and enjoying the lovely fall colors here on a beautiful day. And I passed by a neighbor's house, elderly woman sitting on her front porch, and just decided to stop and talk to her. And sure enough, what she had for me was questions about something going on on her phone. She's like, now I— it's saying I've been hacked. Do you know what that means? And I'm like, well, I know what that means, but then I also think I know what is happening to you. Uh, so because I'm, you know, maniacal like this, my impulse was to pull my phone out, get ChatGPT voice on, and hand her the phone. And say, hey, why don't you talk to this, see if this thing can help you? So she did and explained the situation to it. It gave her very good advice, basically, as you would expect, right? Like, I don't think you should click on that. That sounds like a scare tactic. Are they asking for money? Yeah, it's going to be $2, and then after that it's $12 a month. So the AI is like, by the way, did you give them any money already? And she's like, no, I haven't paid for anything. Okay, good. Close it out. Don't worry about it. Your iPhone has great security on it already. Uh, you know, as long as you don't give people your stuff, like, you should be okay. So she took that advice. Okay, great. Um, then I was trying to— would you like to have this on your phone? She said, yeah, that sounds great. That was really helpful. Uh, couldn't remember her password for the iPhone, so we couldn't install the app, but I was able to just install the 1-800-CHAT-GPT. Uh, not install, just save it as a contact. And I do think that's a pretty cool service that OpenAI offers. I mean, the initial thought for many, of course, who are into this stuff would be like, why would you need a 1-800 number? Like, who's ever going to use that? And the answer is my elderly neighbor who cannot even remember her own password to install a new app on the phone. So she's got that now. She can call that. And it just left me with this feeling which I've approached a couple of times from a couple of different angles, but again, really drove home. There is just an unbelievable amount of adoption work to be done with the current models. This didn't push her anywhere. This did not push the model's capability anywhere close to its limits, right? This is like super easy mode for the model. But it's the kind of thing that isn't going to happen without people actually going out and trying to help the people in their lives adopt this kind of thing. I think you're right that the capital cycle does need to see results soon, but I think we can get 4% growth without additional hyperscaling at this point. And I want those medical advances, all that good stuff too. But do we really have to go harder on the frontier-scale RL? I think the answer is no. I think that literally just going around and helping people use what exists is enough if we actually do it. But I'm the first person who stopped by my neighbour's house, and we don't have a national programme like this. This is something I think would be an incredible good for the individuals who would get the help and also just to make sure that we actually have the sort of economic cushion to be able to pace without the capital market sort of Damocles immediately coming down on us. If we just put some shoulder into adoption, I think we could create a little space and would love to see more of that type of thing happening.

Part 8, keep the steering wheels. First, the view from inside a self-driving car.

Oh, I spent most of yesterday in a Tesla Full Self-Driving vehicle going to Chicago and back from Detroit. And top takeaway from my experience yesterday, honestly, was the FSD is just so awesome. I mean, I was able to take an engagement on a day trip basis to go to Chicago, show up fresh and relaxed after the thing did all the work to get me there. Then I stayed and had dinner and a drink with the gang that I was doing the workshop with, left to come home later than I definitely would have felt comfortable driving 5 hours home after a big day like that, and just basically reclined the— you got to keep your eyes open, right? It will not let you outright sleep. But it's so good. I just kind of reclined the seat and, you know, listened to my podcast, caught up on a couple feeds on the way home. And it was just like, man, this is— I've talked so many times about how a lot of AI use cases can be GDP destroying. This one was clearly one of those that's GDP enhancing. I just could not have taken this opportunity otherwise.

That's the enthusiast in me. 2 days earlier, I had worried out loud that we are letting our own ability to take back control atrophy, and that the real doomer scenario is everything failing at once. Prakash pushed back.

Maybe, maybe the doomers are being a little bit naive though. Maybe because maybe they're expecting people to give up control, and I don't think people are going to give up control. If you look at like AI doomerism or AI safetyism, the question that I ask is, when do you think we will allow humans to not be used to fly passenger planes? Because it's been possible for 15 years and we haven't. Like, why? You have to answer that question before you say that we're going to give up our rights and we're going to be disempowered. If you can't answer that question, then you don't have a working theory of why human beings accept technology into their lives. You're just kind of being naive about capability being there and that humans will automatically adopt

it. Yeah, I think that's an optimistic note and I hope that's right. Maybe it's a little bit of projection by somebody like me who is trying to live on the edge of deployment and is sometimes feeling my personal muscles, for lack of a better term, atrophy here and there. I also do think Tesla's now making cars with no steering wheel, right? Yeah. So this would maybe be a leading indicator, like, do we start selling cars with no steering wheel and no pedals? I don't quite get why we would want to do that, honestly. Maybe it's a cool look, or maybe they should be stowable in the future or something like that. But it does seem as like a metaphor for society as a whole. We probably want to keep our steering wheels in the cars versus rushing to a future. And I want the AI to do most of the driving for me. But like, why are we taking it out? Why are we rushing to take it out? You know, in a sense, I still think we'd be wise to just keep steering wheels. You know,

I

think a little extra cost, maybe a little, you know, a little extra difficulty in, you know, various ways, but like, probably worth keeping.

I think, I think there's, number one, we again, we have this like how injurious it is and how easy it is like this, this like curve, right? And I think the fact is that an automobile accident only injures like 1 or 2 people. And so it's not like a plane. If you want to deploy them more widely, they should be cheaper. And if you want them to be cheaper, you have this object in the steering wheel, which is not being used 99.9% of the time. So let's get rid of the steering wheel. And then if we get rid of the steering wheel and we get rid of a bunch of stuff in there, we can bring down the price again, like 50 or 60%. And you bring down the price 50 or 60%, then you can expand expand the number of people that you can deploy to. And if you expand the number of people you can deploy to, then the number of people dying from auto crashes due to humans drops from 30,000 to maybe 5,000. So I think there's this sequence of activities that you start to look at. I think that's why we go ahead with it. So I think it's a gradual process, like gradual human disempowerment, so to speak. So I can see why the steering wheels are being taken away. I think maybe for larger things like planes, we won't, but maybe for smaller things like cars, you know, maybe we do. I don't know.

Yeah, well, it does seem— I think things are kind of trending toward the— this is sort of an Eliezer-type story where, like, how did you go bankrupt slowly at first and then all at once? Like, that, I think, is looking more and more like the story of AI takeover or human disempowerment. To me, all the time it does seem like we may not, but I think there's a pretty good chance that we will manage a lot of the things in the short term, and then we will be kind of rolling the dice on will the AIs kind of break bad at some point in the future? Because we just don't have fundamental alignment science. We do have a lot of effective control measures and better sandboxing and de-shittified RL environments. But we just won't really know, even as all that's happening, if the big one is still kind of waiting right under the surface.

That's the week. Thanks for listening to the AI in the AM weekly highlights. The narration was my cloned voice. Please tell us what worked and what didn't See you in the morning.

Arms crossed like a locked up door. When you laughed at my worst joke, instead eat the blow. Something's melting. Something's melting. Something's melting. Say you're cold as cold can be. Something's melting. Who turned it down? Not me. Oh, you keep your cool like a trophy. Oh, I polish it every day. Yeah, but you saved me the seat by the window when the floor was a mistake. There goes the ice. It's getting warm in here tonight. Something's melting. Something's melting. You say you're cold as cold can be. Something's melting. Who turned the dial? Was it me? Who turned the dial, honey?

If you're finding value in the show, we'd appreciate it if you'd take a moment to share with friends, post online, write a review on Apple Podcasts or Spotify, or just leave us a comment on YouTube. Of course, we always welcome your feedback, guest and topic suggestions, and sponsorship inquiries either via our website cognitiverevolution.ai, or by DMing me on your favorite social network. The Cognitive Revolution is part of the Turpentine Network, a network of podcasts which is now part of a16z, where experts talk technology, business, economics, geopolitics, culture, and more. We're produced by AI Podcasting. If you're looking for podcast production help for everything from the moment you stop recording to the moment your audience starts listening, check them out and see my endorsement at aipodcast.ing. And thank you to everyone who listens for being part of the cognitive revolution.