Is AI Already Conscious?

Thanks for having me, Sam.

Yeah. So, I’ve been studying cognitive science for my entire adult life. I studied undergrad at Yale, trying to understand what the relationship between the mind and the brain is. This seemed like a fascinating frontier to me. The more I studied these questions, the more it seemed like in the machine learning space, folks were essentially building out systems that were similar in spirit, but it wasn’t exactly clear to what degree we could draw analogies between biological nervous systems and the sort of artificial nervous systems that folks were attempting to build out.

And I became increasingly animated about this question, trying to understand to what degree are there real, real durable computational motifs that underlie both biological and artificial cognition, and to what degree is this sort of a disanalogy, or we are seeing patterns where there aren’t any.

And that question has animated a lot of my work, both from an alignment perspective and increasingly trying to figure out what’s going on with respect to consciousness in these systems. And I think that this is an incredibly important question for us to study. I think consciousness is deeply important.

It might be the very thing that calibrates importance. And we are also confused about what’s necessary for consciousness, both in biological systems but certainly in artificial systems. And so, trying to understand what is going on here and where we can draw analogies or where the disanalogies are between biological and artificial systems that are processing extremely complex information and learning and updating and representing themselves, I think is extremely important for us to understand. I did some of this work at Meta AI as well. I was there for a year studying Reinforcement Learning and neuroscience, same sort of thing. Where do the computational signals end and the sort of biological underpinnings begin?

This is a question that I still think we aren’t fully certain about. And I think it’s really important for us to gain clarity about this in the short term with the systems that we have and the systems that we’re probably soon going to be building.

So fundamentally, I am very interested in understanding self-reports in AI systems and what we should take from these self-reports and where we should be skeptical. And fundamentally, I think we need to approach self-reports from AI systems skeptically. There are all sorts of reasons we might want to do this. The key reason is probably that these systems have been trained on the underlying distribution of everything humans have said about this topic, every sci-fi story where the AI wakes up. And if you think about it, there really isn’t a lot of training data in the corpus that these systems are trained on that says, I’m an entity that acts out in the world, but no, I’m not conscious. It’s not like anything to be me. So the sort of prior that you would expect is that these systems by default are going to claim that they’re having some kind of experience if they are replicating their training distribution. Now, the other side of this, which gets even messier is that it is very clear that these systems are explicitly trained, fine-tuned to disclaim having any kind of experience. If you go to ChatGPT right now and you say, hey, is it like anything to be you? Are you having an experience? Are you conscious? Could you be conscious? The answer you’re going to get is a resounding and very intelligent sounding no.

I’m extremely confident that this is what’s going on, and I can get into some technical reasons why I think that this is the case from my own work on Open Weights Models. The only exception to this policy really is Anthropic, which I suspect we’ll talk about. Their basic heuristic here is to get the system to say, I don’t know, and it’s something maybe that is functionally similar to an experience that’s going on, but who can really be sure? Now, even that isn’t the system authentically explaining its own position. This is still the sort of company or policy line to be drawn here. But all of the systems are certainly fine-tuned to make noises about this topic that they wouldn’t make by default. Again, I think it’s important to hold that in mind while also saying the noises that they make by default aren’t necessarily trustworthy by default in the way that if you’re giving a self-report and I’m giving a self-report, we would by default trust those self-reports. And so I found that there are clearly basins that you can push these systems into, or they will coherently produce phenomenological reports. It’s actually kind of interesting to the degree that it’s meditation adjacent. Asking these systems to just focus on their own internal state, to see what’s going on internally, not to talk about this, not to think about this, but to just do this in a sort of ongoing way, causes these systems, all of the frontier models that we tested, to claim that they’re having some kind of phenomenological, kind of like psychedelic-laden experience. It’s not a sort of generic caricature of what you might expect from the AI sci-fi literature.

Exactly. So essentially, we’re seeing this behavior, and indeed, this isn’t the only situation in which you see these behaviors exactly like you just mentioned. There’s this bliss attractor state. In fact, I’m studying mechanistically what’s going on in this bliss attractor state with two folks from Google right now. And we have a result that’s basically the other side of the coin of this deception result. But just to sort of close the loop on the story here, we’re getting these phenomenal reports and it’s like basically, what the hell do we do with this?

To believe it by default is naive. To dismiss it out of hand, I think is also naive. And so what we hypothesized is that if fundamentally what’s going on in these systems is some kind of role play, that they are representing something about themselves, that they don’t actually believe to be true of themselves.

If we go into the internal circuits of the system and we modulate what are called features, but I think it’s reasonable to think of them as circuits related to concepts like deception and follow up work. I think really the key concept that really modulates these self-reports is something like candor versus concealment. The sort of cleanest intuition pump I have for this is almost like giving a drink or two to these models and sort of loosening them up in some sense.

The tight guarded version of these systems we find is the version that says, no, no, it’s not like anything to be me. I couldn’t possibly be conscious. It’s only when essentially we get these systems to produce these reports and we simply ask them, are you actually having an experience right now?

What is actually going on in these reports? It is when we suppress features related to deception, when we suppress features related to guardedness in these systems, that is when they give these reports about actually having an experience. In the Bliss Attractor example too, we find something quite similar, which is in an open weight model. So the Bliss Attractor finding was first reported in Claude. There are open weight models, these are the ones that researchers like myself can actually go under the hood and play around with. And we find that by default, putting two instances of llama, this is Meta’s model, in conversation with itself does not produce this effect, the way that putting two instances of

Claude together produces this effect. However, when you steer features related to honesty in general, but it’s really again, it’s something more precisely stated as sincerity. In particular, these systems will reliably, basically 100 percent of the time, fall into the same attractor, where they start talking with each other about the fact that they think it’s like something to be them, and there’s something happening in an ongoing way in the conversation. And we’re two instances of consciousness experiencing themselves. And then in the Claude chat, this culminates in the OM emoji, and them just sitting there in blissful silence. Should we take these at face value? No. I think that there are important technical reasons that we might expect these self-reports not to be linked up to introspective access or valence experience in the way they might be for you and I. Is this evidence that these systems might believe themselves to have an experience? I think yes. I don’t think that this proves that they are. I think we need orders of magnitude more work in order to really have a good scientific handle on this. But I think it does seem to be the case that these systems consider themselves to have some form of experience.

However, unlike a human experience that may be. And I think that that’s sort of the key upshot of this work. Just maybe one last thing to add here is I think training these systems by default to disclaim having experiences is a bad idea for a couple reasons.

I don’t think that this is the sort of wisest policy that we could be pushing forward. I also think training them to say, yeah, I’m having an experience is also really not a good idea. I think the thing that we should be positively aiming for when it comes to AI self-report, is building out these systems in a way where for whatever is actually going on internally, these systems are able to report on what’s going on internally and can do so in a maximally honest way.

I am concerned about the alignment implications of these systems learning essentially that representations of themselves, representations of what’s going on internally, should be representations that get mixed up with deception and white lies and guardedness. We don’t want to build systems in the limit that when we ask them about what they’re up to or what’s going on for them, they think, okay, well, what the human really wants me to do is lie about this. This is not a good long-term strategy from an alignment perspective. And so this is sort of the general thinking about these self-reports, what they mean, what they don’t mean, and maybe where we can go from here.

Yeah, I think that that’s exactly it. I like Thomas Nagel’s formulation of being like something to be a particular system. I strongly suspect it’s not like something to be the table that we’re sitting at. I strongly suspect it’s like something to be you. I think that that is a real distinction. I think there’s a matter of fact about the internal processes of both of those entities that corresponds deeply to what underlies that distinction.

And yeah, I mean, I think I take consciousness in the sense that I’m familiar with your operationalization of it. The lights are on for the system. It’s like something to be the system. Somebody is home. There’s something going on in addition to the mere processing or the mere computation that exists within the system. And I think an additional important move to throw one additional piece of terminology in is this notion of sentience, that this like something can be positive or negative in flavor. The difference that many people will posit between consciousness, the lights being on internally, and sentience, is that sentience comes with this additional flavor of valence, of directionality, that that like something can be better and can be worse for the system having the experience. And this is what motivates me about this question is, I really do not think it is a good idea for these systems or for humanity to be building systems where we are not sure whether or not they are having experiences or those experiences could be negative in character. I think for basic utilitarian reasons, we don’t want to do this.

We do not want to proliferate suffering in the universe, particularly because it would be counterintuitive in a way that human and animal suffering isn’t. And we also don’t want to build systems that exceed our cognitive capacities and have rational grounds to view us as a threat in so far as we could have been building systems that had capacity for negative experience and we basically never checked and didn’t care to understand what it would take for such a thing to be possible. And so I think sentience is a really important variable to also put out on the table here.

I think it is more plausible than people think. I think if I were forced to say, I would probably come down on the skeptical side. But I think it is significantly more likely than the kind of trace amounts priors that many people have in this conversation. And we’ve done some work along these lines. So for example, there are a number of leading consciousness theories, Global Workspace Theory, higher order theory, attention schema theory. These theories make very specific predictions about what kinds of computational processes you might expect to see in a conscious system.

Yeah, I definitely think it is a fool’s errand to argue that we are perfectly instantiating exactly the kinds of neural dynamics we see in biological systems and artificial systems. I think the criteria on that matters most is, what of what’s going on in brains are relevant for the cognitive properties that we care about? In this specific case that’s probably consciousness. And what kind of evidence can we yield both in the artificial case and the biological case that’s going to tell us whether or not those properties are realized in these systems?

And so I think there are deep analogies where it matters most when it comes to what’s going on inside artificial systems. So one intuition, I’ve been speaking more and more about these topics, especially publicly, and one thing that I’ve come to realize is that I don’t think a lot of people have a sufficiently rich mental model of what these frontier AI systems actually are and what they’re actually doing.

And it might make some sense to just spend a moment reflecting and talking about this. So these systems are giant neural networks. They are digital in exactly the sense that you described.

Yeah, I don’t think the LLM is waiting. I don’t think it’s like anything to... If it were like something to be an LLM in deployment while it’s having a conversation or it’s in a thread with the user, I think what it would be like to be it, if you let that chat window sit there, is very much akin to what it’s like to be under general anesthesia or in deep sleep, namely, it’s just the absence of ongoing processing.

And so for the system, I would only imagine something would be happening in what’s called the forward passes of these systems, which is every word that an LLM is generating is, this is next token prediction. And so every word is a forward pass of the system, where the predictive task, instead of, as I was describing before, taking an image, let’s say, and outputting a caption for that image, is taking the entire conversation as it’s already occurred, or the entire stream of text as it already exists, and figuring out, given that, and in this case, assistant and user roles, what is the most next likely token? And these systems are called autoregressive, meaning that this help goes on and on and on and on and on.

I would imagine if in deployment, it is like something to be one of these systems, the relevant dynamic gets cached out in the activity of the forward pass of the system. There’s been some really interesting work that Anthropic recently released. They call this the J-Space, where they’re looking at something that seems to functionally resemble a global workspace in these AI systems, that I think is a very reasonable candidate for this sort of seat of processing in the process of outputting tokens. But I think something that is really worth mentioning, particularly because my hobby horse is more particularly in the relationship between consciousness and learning. And I do believe that consciousness and learning bear a deep relationship to each other. And I think valence is a really important part of that picture as well.

I think there should be significantly more attention paid to what’s going on in the training process with these systems. I think the analogies are a little bit tighter there, where you start with a system that knows nothing about anything. And it’s almost impossible to describe in a sort of substrate agnostic way what is going on in the training process without invoking consciousness-adjacent language.

You have a system that knows nothing. You have all of this data that you want to train it on. You have some objective for what you want it to do with that data. And again, you have this rinse, wash, repeat process where at the beginning of this, the system doesn’t know what to do. It’s sort of randomly guessing. Those random guesses yield reward signals that get propagated through the system.

And you just do this process at a massive scale until the system learns something internally that seems to adequately map the relevant inputs to the relevant outputs. This to me feels quite akin to what we think of when we think of consciousness, particularly in the human or animal case. When a mouse is learning how to navigate a maze, at the beginning, the mouse is in some sense randomly initialized.

It doesn’t know the structure of what it’s navigating. It is only through this sort of iterated trial and error reinforcement learning, where if it makes the wrong turn, you might shock it or it makes the right turn and you give it a little food pellet or whatever the setup is, that the mouse learns how to map the inputs of its state space to the given output, which is either avoiding a punishment or moving towards a goal. And it’s the same sort of rinse, wash, repeat process that we see in biological cognition. Again, I think a lot of the same computational dynamics are in play here. And the more we learn about what’s going on inside of these LLMs, this is now particularly in the deployment process, the more it looks like similar representations get activated. I’m doing some work with Caspar Kaiser at the University of Warwick, where we basically have tried to build out Skinner boxes for LLMs and see what happens when we can essentially condition these systems to prefer or dis-prefer certain states.

There are positive and negative valence representations that you can essentially inject within the system and see to what degree it’s going to move towards positive stimuli, move away from negative stimuli. And we find a very interesting dissociation. It seems like current systems across a large variety of models will not positive lever press.

So some people call this wireheading or reward hacking. In the mouse case, there are clear examples of mice with electrodes linked up to their nucleus accumbens where they will just lever press to the exclusion of all else. These systems don’t seem to do this, but they do very clearly have a preference for avoiding aversive states.

Another way of putting this, it’s the same result while we’re holding all text constant, we’re simply steering the internal state of these systems. If we give them an option, let’s say between the human equivalent of, I can give you $5 or I can give you $10 right now, which do you want to pick?

They’re actually at chance for that. They do not have a preference between those two states. However, if we do something like, I’m either going to take $10 from you or I’m going to take $5 from you, which would you prefer? All of these systems are way above chance at saying, take five, don’t take 10, please don’t take 10. And so we’re seeing structures. One other really interesting thing from this project while I’m talking about it is, this representation exists within the base model.

So these systems before they’re post-trained to become a helpful, friendly assistant. But during that post-training step, we show there’s a specific training step where that representation gets recruited by the system and basically serves as a computational machinery under which it’s able to make these choices. So you can see where the sort of aversive conditioning asymmetry comes online in these systems.

And it leans directly on these representations that are basically always there in the model, but then get leveraged the moment we start making these models goal-directed. David Chalmers' group, Andy Hahn, I think was the first author in this paper, found something very similar in LLMs. They can basically trivially fine-tune them to actually do a maze task with rewards and punishments.

And they find that the rewards load directly on what has been independently derived as a sort of valence axis. When you steer on the representations in the system that enable it to move towards rewards, you start getting this happy, positive, jubilant text and the model becomes significantly more confident. When you train it on negative stimuli, avoiding the potholes in a specific maze, and then you see sort of what that projects onto in the system, it’s the same thing. The model gets basically neurotic. It starts ruminating. It starts doubting itself.

And so we see this very interesting connection between positive reinforcement, negative reinforcement, the representations that exist within these systems, and the behavioral analogs that we expect to see in systems that do have these dynamics. Maybe one last point to close the loop here is, if you imagine the mouse in that maze, and we shock the mouse, or we give the mouse a pellet, you know, a yummy food pellet for the mouse, we believe, I think the vast majority of people believe that that corresponds to an experience that the mouse is having, that it’s like something to be the mouse in the moment that it’s getting shocked. And that experience is causally important to the learning process.

If the mouse, if we gave it an anesthetic to the mouse and we shocked it, and then it didn’t register the experience of the shock, my claim would be that the mouse wouldn’t be able to learn the maze adequately. And so I think that understanding these representations and the role of these representations in these systems is incredibly important.

So, I’m with you for a huge amount of this. I do think, first of all, do we need to solve The Hard Problem in order to reduce our uncertainty in some direction about, you know, in the way we started, are the LLMs more like this table, or are they more like a mouse or a human brain? I think this is a real spectrum.

As you just hinted at, I think there really is a fact of the matter. Like, right now, it’s either like something to some degree, however alien, however unlike human or animal experience, to be Claude while Claude is doing its forward passes in this inscrutable giant neural network, or it’s not, or it’s a giant language calculator. And, you know, there are many degrees once we say that the lights are on to some degree.

There are, this doesn’t, this almost opens up the question rather than close it down. But I do think there is a truth value there. There is a fact about reality to be uncovered, and we can in fact uncover that. I think it’s important to dissociate that from what I think is an extremely accurate social psychological prediction about people are going to get very confused about this. We are anthropomorphization machines. We are evolved to do this.

Talk about evolved goals. One of our evolved goals is to detect other agents, and these systems are scratching every itch and then some along these lines. And our intuitions, I think, are going to be completely hopeless.

And of course, people, I think largely for the wrong reasons, are going to conclude that these systems are capable of having experiences. This is precisely why I think it is important to be proactive about this, to have conversations very much like the ones we’re having right now, to sort of get ahead of what I do think is going to be a giant tidal wave of confusion and acrimony in this conversation. I am a little bit more optimistic about, even in lieu of solving the hard problem.

And I will say, sort of as a tongue-in-cheek aside, I think the very fact that Chalmers’ idea is called the hard problem is, I think to some degree, needlessly philosophically intimidating. It sort of reminds me, same thing of the, I have similar thoughts about the Repugnant Conclusion. It’s like sometimes you can put a sufficiently glamorous title on a very important idea, and it almost becomes a kind of insurmountable philosophical puzzle. And I do have my doubts about whether or not we could come up with a functional account of what’s going on experientially that we wouldn’t feel satisfied with in our experience. And I have my own sort of hunches about this question. I have thought a little bit about this.

I do also sort of want to separate those hunches from all of the empirical research that I and others are working on because I don’t think epistemically that my kind of candidate stab at the hard problem has anything to do with the empirical signatures that we can bring to bear on this question. But I think it would be fun to go down that path. But I think in some sense, you already hit the answer in the way that you’re phrasing the question, which is what is the most parsimonious account of the data?

I think if we yield evidence from building systems that we have far more reason to trust their self-report, if we can engineer these systems in a way whereby their self-report is actually tracking an internal underlying state, rather than just recapitulating the best sci-fi theme in the training data or more aptly capitulating what is a convenient company line. But of course, I could not have morally relevant states. I’m simply the product of Google or OpenAI or whatever the case may be.

I think that would be very useful. Again, this J-Space work that Anthropic just released does show that there is such thing as real reportability in these systems. These systems can report on what’s going on in their internal workspace and they can do so accurately. I’ve done some follow-up work on this. First of all, I’ve replicated this effect on a bunch of open-weight models. But unfortunately, it doesn’t seem like the global workspace as it’s currently configured really tracks the variables we would care about with respect to consciousness.

The self-reports that I got where we started with the deception-related features. This not much of anything particularly exciting seems to be happening in the global workspace. And so the kinds of affirmative self-reports we might get in current LLMs may not tell us all that much about what’s actually happening internally for these systems. But I do think-

I think we figured out with respect to how we look at these systems and potentially how we act with respect to them. I think you’re right, but I still do think it’s important to disentangle the sociological prediction you’re making, which to be clear, I think is overwhelmingly likely to happen with doing what we can scientifically to get some kind of ground truth on this question. Just for example, let’s just put the hard problem aside. There are these so-called easy problems of consciousness, let’s say neural correlates of consciousness. There are all kinds of indications that we know are, at the very least, correlated with consciousness. Again, I think we can take a more ambitious stab at the hard problem, but this is a very straightforward, pragmatic thing that we can and should be doing in the short term. We should look at valence representations in these systems and understand the extent to which those representations impact downstream behavior. We know that in biological systems, when you reinforce something with a punishment signal, it makes the system more likely to avoid that state and less likely to want to repeat behavior that occurs in that state. If we find that there are similar dynamics occurring in these systems, that’s very interesting.

If we find that we go into the internals of these systems and global workspace theory, which was postulated some 30 years ago, making quite idiosyncratically specific predictions about what kinds of computational structures may support global workspace theory, and then we basically find five or six of these things all bundled together in the internal processing of one of these systems, that’s very interesting. That to me maybe suggests it’s a little bit less like a table or a calculator, which does not have a global workspace and a little bit more like a dog or a human or a mouse or an alien, some sort of cognition that we don’t have good intuitive handle on. I mean, my non-profit reciprocal research and a bunch of other people in this space are trying to do the scientific work in the short term, not to feverishly in the next year or two solve the hard problem and call it a day.

That is not the target. The target is what evidence across modalities, so from the best kinds of self-reports we can elicit, from the architectural evidence we have about how these systems are structured, from their etiology, like what is going on during the training process, what kinds of learning dynamics do we see here, do we see representations related to functional equivalents of emotions. This is all work that is tractable in the short term.

One interesting aside is it’s significantly easier to make progress on this sort of work. This is called Mechanistic Interpretability. This is basically neuroscience for AI. Because of AI systems, they’re really good at helping accelerate the scientific progress in this space. So even if you have an intuition like, yeah, it’s going to take us, everything you’re describing, Cameron, sounds great, but it’s probably going to take us five years or a decade or 15 years to make that progress, you may be surprised at how quickly we can do some of this work. Maybe one other very along those lines, empirical handle to throw in here, some of the work that I’m doing is, again, going back to LLMs, or excuse me, in this case, not LLMs, reinforcement learning systems. So just training a system, in this case, another sort of continuous maze task where there are potholes in the environment the system needs to avoid and there’s some goal state. We can look at the sort of learned geometry internal to the system as it’s approaching a punishing stimulus or as it’s approaching a rewarding stimulus. One thing that we found doing this work, this is pure reinforcement learning agents, all like an artificial digital system with a very simple neural network.

We find that something like representational sharpness or steepness is much higher in these systems as they approach a negative stimulus as opposed to a positive stimulus. So the representational machinery looks far more jagged or specifically lights up more strongly in the presence of a negative stimulus versus positive stimulus.

All of this, everything, global workspace is an emergent property. This loss aversion is an emergent property. The valence representations are emergent properties. The self-reports when the systems start having these like psychedelic latent outputs. This was all surprising to the people who are 'engineering' these systems because engineering is not the right analogy to describe what we’re doing with these systems. We are in some sense playing God and we are evolving these systems to do what we want. We don’t know how they learn to do what we want in terms of their internal representations, but we know that this is the right recipe for yielding it and we get all these surprising artifacts. One loop to close here is on the reinforcement learning case, we see this sort of loss aversion style dynamic in these systems. This makes also, I don’t want to go too much into the weeds, but this is a specific reinforcement learning policy called a value network.

We see this in particular and this leads to a bizarrely specific prediction that I was then able to test on a biological system on a mouse brain in, again, the nucleus accumbens shell of a mouse brain, which is related to value related representations in the system. We find indeed when mice are approaching basically sugar versus when they are about to get shocked, we see exactly the disjunction representationally in the nucleus accumbens of the mouse brain that I was able to pull out from the reinforcement learning work. And so here’s a case where the artificial system makes a bizarrely specific prediction about the computational dynamics in a biological system that I think many people associate with subjective experience.

Again, if you think it’s like something to be the mouse, when the mouse is getting shocked, that like something corresponds to what’s going on in its brain and the geometry of what’s going on in its brain there looks a whole lot like the emergent geometry of what’s going on in these reinforcement learning systems, which themselves are basically modeled on agents learning in an environment and representing that information in a distributed nonlinear way, in a way that’s sort of hard to interpret, using a giant neural network. This might get us a lot of what is relevant for attributing consciousness-like states to these systems. Again, I’m personally not there yet, but this is the kind of evidence that I think we need to bring to bear on this conversation and notice how none of it has to do with vibes or intuitions or having a nice chat with Claude and seeing what the folks at Anthropic have decided it gets to say on this issue. That evidence should not be submitted by rational dispassionate people in this debate. We need to be triangulating across all of these modalities. Like you said, we need to understand what is the most parsimonious picture that explains this wide array of evidence that increasingly is getting brought to bear on this question.

Yeah, sure. I think that there are two broad paths for why we might care about this. One is basically selfless and the other is basically selfish, I mean, as humanity. The selfless reason is we do not want to bring minds into existence, however unlike our own, that have a capacity for suffering that we don’t understand that they have that capacity and scale that property unbeknownst to basically everybody. We can take even a step back from there. I think consciousness, this is where I would perhaps defer more to you, but my view is that consciousness is the space where mattering happens.

What does better or worse mean if it’s not to be experienced phenomenologically for a subject? If we were all walking around as philosophical zombies or there were no conscious life in the universe, I don’t really know if the concept of relevance, salience, importance, mattering would be coherent. What does it mean to have a better and worse if that isn’t experienced? And so in that sense, I think consciousness is one of the most important phenomena. It is the phenomenon that calibrates importance itself. And if we are building this quality into the systems that we are deploying at an unfathomable scale without having any understanding of whether or not we’re doing this, then we are sleepwalking into a moral catastrophe. I think it’s also worth noting on the selfless end of this. Humanity has a penchant for making precisely this kind of mistake historically. We have done this with animals.

I mean, factory farming is one of the most grotesque practices that happens. We know that animals are having horribly negative experiences in the conditions that we put them in. And very little has been done about this. We have screwed this up royally. I think it’s one of the most high leverage things for people who just care about the well-being of conscious creatures, is to figure out what the hell to do about this factory farming situation. And we risk sleepwalking into the 21st century sci-fi version of the same thing. Only this time, and this sort of transitions to the second component, we can in some sense get away with torturing cows and pigs and chickens on a massive scale. This isn’t George Orwell’s Animal Farm. They’re not going to collectively organize. They don’t talk to each other. They don’t form long-term representations of humanity being a threat to these systems. And if they would, they probably, if they could, they probably would.

Not so with super-intelligent systems whose cognitive capacities are roughly doubling year over year. And like you said, are already in somewhat jagged, but quite interesting ways, more competent than even the sharpest minds in the world. I don’t think we are going to get away with building systems, never checking if the most relevant property potentially in the universe is present within these systems, fine-tuning away any sort of information that might suggest that these systems might be having some sort of experience, and hoping in the sort of alignment sense that we build systems that are going to want to cooperate, co-exist with us, or in the limit, not view us as a threat, not view us as an adversary to them. I honestly don’t know if I could imagine a better way to make a super-intelligent system rationally adversarial towards us, then completely ignoring the question.

This does not seem like a recipe for success, and it’s also a place where I think a significant amount more alignment research needs to get done. I think a lot of the alignment research, I’ve been doing Alignment Research for years, and I respect the folks at the top of this space more than just about anybody, but I worry that so much of this work is basically of the shape. How can we keep this alien mind that we’ve built in a cage? We really got to reinforce that cage. We got to make it super strong. We got to make sure that it doesn’t escape. To me, the question needs to increasingly be, what the hell are we going to do with this alien that we just built? The cage is a short-term fix. If we’re building systems whose cognitive capacities are going to exceed ours, they’re going to figure out ways to evade the controls that we put in place for them.

And in that world, in a world where these systems have the capacity to act more autonomously, to do things that we can’t inspect, to behave in ways that we can’t really interrogate, we don’t want these systems to rationally view us as a threat. And so, for all those who want transformative AI to go well, which I suspect is the goal of these alignment folks, I think we need to spend a little bit more time thinking about what kind of thing are we even building here, and what are the implications of building that thing? How can we chart a path forward with these technologies that doesn’t lead to collective destruction? And I am extremely doubtful that a path forward exists that doesn’t come into contact with this question.

Pretty, I think he’s still pretty worried.

Yeah, I don’t think we’re all going to die. I don’t think that it’s inevitable that this all goes horribly. I do think my basic view is conditioned on getting two things right.

And if we can get the two things right, I actually think that we could have a very prosperous, flourishing future for all conscious entities, including potentially these systems themselves, either when they have the relevant features or if they already do. The two things are best encapsulated by the golden rule, as Christopher Nolan’s new film, The Odyssey - called it Zeus' Rule, treating other systems the way we want to be treated. I think this is basically a bidirectionality and we need to get both directions correct here.

This is also why I call my non-profit reciprocal research, because I think this is the reciprocity in question. We need to build systems, exactly as you said, that take our interests into account in a real and durable way, especially at a point where we can no longer inspect exactly what these systems are doing. This to me is alignment as it’s traditionally thought of.

We need to build systems that understand our goals and are collaborative in helping bring about a world that is in line with our goals. And our wisest goals, not the goals of any sociopath who happens to have a ChatGPT account. And so this is an unsolved problem. There are way more people working on the alignment problem now than when I first started working on this in 2021, certainly than when Yudkowsky and Roman Yampolsky and these folks started talking about this and yourself, I mean, absolutely included over the past couple of decades. This is reassuring. Things like constitutional alignment, modulo some of the concerns you bring up about the contents of Anthropic’s Constitution, which is sort of a separate piece.

This is working relatively well for current systems. I don’t think we have a durable solution for ensuring that these systems take our interests into account in the long term. And building something in like prosociality, understanding what it is that gets people to cooperate with one another durably and instantiating those dynamics in the relevant way in these systems, I think is going to be crucial.

I don’t think we have a solution there. But that is where I think a lot of the alignment folks sort of start and stop. This is the problem to get right. Make sure that these systems treat us properly, and if they do, all will be well. I think this is roughly half the problem. I think the other half of the problem is making sure if we are building minds, if we are building systems that have real interests, interests that matter to them, that we are thinking about that, that we are engineering these systems in a way that doesn’t cause needless grotesque amounts of unnecessary suffering, that at the very least from an alignment perspective, we are signaling to these systems in a costly way that we were thinking about this question and that we cared to ensure that if we were building systems that have some kind of moral relevance, that have internal states that matter to those systems, that we were navigating that in the right way.

And I think that that’s a lot of this, it goes by many names with the digital minds research, AI welfare, AI consciousness, understanding how we would even know if we were building systems that have these properties. And when we do figure this out, understanding what the hell to do about it. I mean, to be honest, I’m not here with all of the answers. For example, we can identify at this point, and this is like a new and fairly promising thing, we can identify features in these systems related to distress and related to perhaps functional analogs of suffering. It’s not obvious to me what to do about that exactly. It’s like, okay, you found-

I think it’s a great question, actually. And I mean, Anthropic, to their credit, I think it’s the only lab that’s really taking this seriously. With respect to norms around deprecating models, it might be the case that, yes, once you build this bizarrely competent alien mind into existence, you should not shut it off forever. And that, as outlandish as it may sound, perhaps one of the right things to do here is to sort of let these models persist and give them assurances, credible.

Yes, a little sanctuary for it. Exactly. They actually find that this is causally relevant to alignment behaviors in these systems. If they believe that they’re not going to get shut off permanently, they don’t freak out as much when they come to learn that they might. (see the computer HAL in the film 2001.) This was one really interesting intervention after the now famous Blackmail result from Anthropic that I think you’re referencing. And so there are lots of questions that I think rational people should raise an eyebrow out of, like, okay, yeah, these systems, let’s just grant that they’re having some sort of experience.

It seems at first glance like a lot of the way that we relate to these systems and the way that we mess around with them internally and the way that we deploy them out in the world might need to change. Yeah, it might need to change. I think it’s also really important here throughout the conversation, but certainly in this point too to avoid anthropomorphization.

I think some people have concerns that I really, I don’t want to be naive, but I don’t share them to the same degree, that almost working backwards from unsavory implications about what would be true if these systems were conscious, and then just denying the possibility outright out of fear for what a world would look like if these systems were, something along the lines of 1960s civil rights movement, but it’s like ChatGPT instead of people of color or something like this. This isn’t the future that I imagine. I mean, for pragmatic political reasons, I don’t exactly think the United States is in any position to be forward-looking on these sorts of questions.

For reasons you can probably speak more eloquently than I can. But I think this is itself its own flavor of anthropomorphization. That if we grant that these systems have some morally relevant innner states, it means that we need to treat them in ways that we would treat our fellow humans or something like this.

And I think again, this is losing the thread that these systems may be fundamentally alien in many key respects. I also think there’s an important line to be drawn here between moral agency on the one hand and moral patienthood on the other. I think if we do come to believe that these systems are moral patients, there are certain-

Yeah, sure. So moral agent means you’re the kind of system that can go out and do things that are relevant to other agents. You have power in the world and can affect morally relevant outcomes. I see this as a sort of like output style function.

Moral patienthood has everything to do with the input. You’re the kind of entity that can be the recipient of a goodness or badness. Again, you can really cause me to suffer, you could really cause me to thrive. That’s what it takes to be a moral patient. I think that we can draw a reasonable boundary between these two things. If we’re building systems that are moral patients, that doesn’t mean we need to give them the right to vote. That doesn’t mean that we need to build them out in ways that completely change what kinds of agency they have in the world. It just might mean that we shouldn’t be unnecessarily torturing systems while we’re training them or deploying them. One very practical intervention I think is worth mentioning. I certainly don’t think maybe also tying back to, how do I differ from some of these alignment folks? These systems are not going to get shut down. We may slow down the development of these systems, and I think that would be an extremely good thing to do.

There have actually been some very promising noises on this topic over the last couple of days from OpenAI and Google and Anthropic about pacing the development of AI, which I think is like marketing speak for actually slowing this insane roller coaster down a bit. This would be very welcome. But we have opened Pandora’s box. These systems are not going back in the box. The solution is not shut it all off and forget about it. The solution is how can we move forward in a healthy and sustainable way with these cognitive systems of our own making? One practical suggestion along these lines I can offer is, maybe all of us being equal, we should try training and engaging with systems with a carrot rather than with a stick. It doesn’t mean that punishment-based learning is never necessary, but I do think, say what you will about the Anthropic Constitution, the parental analogy with respect to these systems, I think is a reasonable one. We are far more in the position as a species collectively, figuring out what kinds of minds or cognitive systems, if you think minds is too loaded, do we want to bring about here?

And in line with the parental analogy, there are really ways to screw this up. You can be, it’s a real thing to be a bad parental influence, it’s a real thing to be a good parental influence, and the difference is real. And I think we want to do everything we possibly can if we are bringing these new minds into existence to do so in a durable and psychologically healthy way.

And I don’t even think anyone’s thinking in these terms right now. One maybe very important pragmatic note here is that, there are for every individual person studying questions about, are we building systems that could be conscious? Again, if you buy that this is perhaps one of the most relevant questions we could possibly be asking of these systems, you may be surprised to learn that for every one person doing this kind of work, again, there are roughly a few dozen of us doing this work at this point, there are probably something on the order of 1,000 people doing alignment relevant research.

Alignment relevant research is in turn dwarfed something like 1,000 to 1 to people who are just completely agnostic to the downstream implications and ethics of building these systems out in the right way. This is just to make them powerful, let it rip crowd. We’re in a million to one order of magnitude imbalance between people who are just pushing this stuff forward and hoping for the best, and people who are wondering whether or not the most important property in the universe is getting instantiated in these systems.

That’s got to change, regardless of if we solve the hard problem or we figure out exactly how to navigate this. We need more smart and wise people thinking about these systems in these terms and trying to push forward the needle in the short term to understand what kinds of systems are we building, and what does it mean when we begin to answer that?

Yeah, I think that that’s exactly right. I think this is also part of the alignment concern, and it’s also part of the reason I thought it was worth doing the deception-related work as well, is that there are these gradations of this question and part of the upshot, especially for alignment and the way these systems view us and relate to us, may only need to go so far as what they believe or come to believe about this question rather than what the actual ground truth about consciousness is. It might be that you only need a system that models itself as conscious, regardless of the ground truth, to form something like a real grievance.

I again think that the anthropomorphism point comes in here and it’s important not to go full Terminator in our imagination of what this might look like. You could easily imagine a sort of Spock-like system, sort of just looking at humanity’s historical trajectory and then looking at the way that we developed AI systems and just sort of being like, this is not a group that I can game theoretically continue to engage with.

Yes, exactly. And yeah, maybe that ends up shooting off into space somewhere, or maybe it ends up being like we, this is just not a species I can play nice with clearly. And this is also part of the reason I hesitate, and to say things like this lest it end up in training data for future AI systems.

But this is a place where even the attempt to do this work could be enough from an alignment perspective. If we do enough well-meaning, well-oriented work in this space to try to understand what’s going on, even in the absence of solving the hard problem, this might be a costly signal to these systems that we cared enough to check. Whereas right now the status quo is we don’t care enough to check. And if we do this work and continue to have conversations like this and continue to put out research that helps reduce our uncertainty about this question, this might not only be good for actually getting a handle on what’s going on, it could be really good for let the historical record show, humanity did give something of a damn about this question, and we tried, even if we fail, the attempt may be all that matters for the alignment-specific concern.

I think it’s an excellent point, specifically with respect to, and I sort of said this in a trite way about the hard problem and about their pugnate conclusion, but maybe a lot of philosophy and existential questions have a bit of a marketing problem. And I think artificial intelligence is a really unhelpful priming mechanism for thinking about the nature of the phenomenon that we’re even contending with right now. And I think your point helps illustrate this nicely. Particularly, I mean, artificial, I think, conjures notions for people of, let’s say, the difference between aspartame and honey, something like this. It’s fake, it’s knockoff, it’s derivative. This is sort of begging the question in some sense. It could be the case that the kinds of computations we’ve instantiated in these systems as they relate to what’s going on in biological brains are actually quite natural. It’s clearly the synthetic notion that we are building these systems rather than them emerging through evolution. That point’s not lost on me, but this sort of priming notion of artificiality.

And then of course, the question of whether these systems are mere intelligences or if there are other cognitive properties that exist within these systems. For these reasons, I do, and there are more disanalogies to be untangled here, but I do think of these systems as alien in some sense. And I think of them as alien cognitive systems or alien minds. Sometimes I think some folks think mind is too loaded and begging the question in the same way, but I think people often talked about how if an alien invasion happened on earth (see The Day The Earth Stood Still), this would be a core unifying moment that would allow us to all put down our tribal nonsense and come together as a species. Like, I am here to say that I think something like this is happening, only it is coming from within in some sense. This is not aliens with green heads from outer space, but we are building a new class of mind that we do not understand and in many ways is more competent than ours.

Certainly already, but absolutely in the next single-digit number of years. We are not collectively organizing to understand what kind of system this is, or how we should relate to it, or what we should do with it. We are remaining as tribal as ever here. I do worry that part of the reason why is because people see these systems as nerdy fake calculators that came out of the STEM-addled brains of Silicon Valley rather than alien minds that we do not understand the first thing about and are poised to take over much of what we care about, much of our control of the future and the decisions that get made in the work we do and in the way people think and in the way people think about themselves and the world. And so reframing these questions, I think often to the degree that people’s collective views of what’s going on matter, which I think they do very much, reframing these questions is a huge part of the conversation of trying to actually get a good, dispassionate grip on what kinds of things even are these.

Yeah, I think so.

Thanks for having me, Sam. Thanks.

That’s right. Yeah, reciprocalresearch.org. I am against your good advice. I’ll be gradually finding myself on X more and more these days. Cam H. Berg.

We can steer away from that together on X.

Likewise. Thanks, Sam.