VAEs Are Energy-Based Models? [Dr. Jeff Beck]

What makes something truly *intelligent?* Is a rock an agent? Could a perfect simulation of your brain actually *be* you? In this fascinating conversation, Dr. Jeff Beck takes us on a journey through the philosophical and technical foundations of agency, intelligence, and the future of AI. Jeff doesn't hold back on the big questions. He argues that from a purely mathematical perspective, there's no structural difference between an agent and a rock – both execute policies that map inputs to outputs. The real distinction lies in *sophistication* – how complex are the internal computations? Does the system engage in planning and counterfactual reasoning, or is it just a lookup table that happens to give the right answers? *Key topics explored in this conversation:* *The Black Box Problem of Agency* – How can we tell if something is truly planning versus just executing a pre-computed response? Jeff explains why this question is nearly impossible to answer from the outside, and why the best we can do is ask which model gives us the simplest explanation. *Energy-Based Models Explained* – A masterclass on how EBMs differ from standard neural networks. The key insight: traditional networks only optimize weights, while energy-based models optimize *both* weights and internal states – a subtle but profound distinction that connects to Bayesian inference. *Why Your Brain Might Have Evolved from Your Nose* – One of the most surprising moments in the conversation. Jeff proposes that the complex, non-smooth nature of olfactory space may have driven the evolution of our associative cortex and planning abilities. *The JEPA Revolution* – A deep dive into Yann LeCun's Joint Embedding Prediction Architecture and why learning in latent space (rather than predicting every pixel) might be the key to more robust AI representations. *AI Safety Without Skynet Fears* – Jeff takes a refreshingly grounded stance on AI risk. He's less worried about rogue superintelligences and more concerned about humans becoming "reward function selectors" – couch potatoes who just approve or reject AI outputs. His proposed solution? Use inverse reinforcement learning to derive AI goals from observed human behavior, then make *small* perturbations rather than naive commands like "end world hunger." Whether you're interested in the philosophy of mind, the technical details of modern machine learning, or just want to understand what makes intelligence *tick,* this conversation delivers insights you won't find anywhere else. --- TIMESTAMPS: 00:00:00 Geometric Deep Learning & Physical Symmetries 00:00:56 Defining Agency: From Rocks to Planning 00:05:25 The Black Box Problem & Counterfactuals 00:08:45 Simulated Agency vs. Physical Reality 00:12:55 Energy-Based Models & Test-Time Training 00:17:30 Bayesian Inference & Free Energy 00:20:07 JEPA, Latent Space, & Non-Contrastive Learning 00:27:07 Evolution of Intelligence & Modular Brains 00:34:00 Scientific Discovery & Automated Experimentation 00:38:04 AI Safety, Enfeeblement & The Future of Work --- REFERENCES: Concept: [00:00:58] Free Energy Principle (FEP) https://en.wikipedia.org/wiki/Free_energy_principle [00:06:00] Monte Carlo Tree Search https://en.wikipedia.org/wiki/Monte_Carlo_tree_search Book: [00:09:00] The Intentional Stance https://mitpress.mit.edu/9780262540537/the-intentional-stance/ Paper: [00:13:00] A Tutorial on Energy-Based Learning (LeCun 2006) http://yann.lecun.com/exdb/publis/pdf/lecun-06.pdf [00:15:00] Auto-Encoding Variational Bayes (VAE) https://arxiv.org/abs/1312.6114 [00:20:15] JEPA (Joint Embedding Prediction Architecture) https://openreview.net/forum?id=BZ5a1r-kVsf [00:22:30] The Wake-Sleep Algorithm https://www.cs.toronto.edu/~hinton/absps/ws.pdf --- RESCRIPT: https://app.rescript.info/public/share/DJlSbJ_Qx080q315tWaqMWn3PixCQsOcM4Kf1IW9_Eo PDF: https://app.rescript.info/api/public/sessions/0efec296b9b6e905/pdf

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  This episode is brought to you by Indeed. You're ready to level up your business, but the pressure to find the right person to make that happen is intense. Indeed sponsored jobs can help carry the load. It boosts your job post to help it reach quality talent fast. Cut through the chaos with Indeed sponsored jobs. Get a $75 sponsored job credit at indeed.com/podcast. Terms and conditions apply. Hiring now? Then this is a job for Indeed sponsored jobs.

• 0:30Speaker 01

  Visible puts the ultimate wireless hack in the palm of your hand. With unlimited 5G wireless data and hotspot designed to keep you connected. Powered by Verizon's 5G network. And no contract holding you back. Plans start at $25 a month or get the premium Visible Plus Pro plan and save $10 on your first month when you use promo code HACK. Make the switch at visible.com. Terms apply. See visible.com for details.

• 1:00Speaker 02

  Geometry deep learning is a big part of like, is a big part of the stack, if for no other reason than when we talk about like modeling the physical world. That means like incorporating the symmetries that exist in the physical world. It's like, we're highly motivated to employ a lot of those methods and techniques.

• 1:15Tim Scarfe?

  But is the world written in code? Or do you mean exploiting the regularities in the code that seem to have some?

• 1:21Speaker 02

  Exploiting the regularities. No, it's like, look, we, it, things are, it is, the world is translation invariant. The world is like rotation, well, not really, because there's gravity, but like in principle, you know, there is a principal axis, but it's certainly rotationally invariant in the X Y plane.

• 1:37Tim Scarfe?

  Yeah.

• 1:38Speaker 02

  And if you, if you want to have a good model of the world as it actually is, it should incorporate those features. And of course, you can discover it, you know, in a brute forcey way, but the mathematician in me really wants to build, build the symmetries in. Unfortunately, we've got a lot of great tools that were developed over the last several years that can do that.

• 1:56Tim Scarfe?

  What's your view on agency?

• 1:58Speaker 02

  If I'm being, you know, like an FEP purist, I have to sort of say like, oh, well, there's no difference between, you know, an agent and an object in a very real way. Or at least there's nothing structurally distinct between what, how we model an agent and how we model an object. It's really just a question of degrees, right? An agent is a really sophisticated object, right? It has internal states that represent things over very long time scales. You know, it has sophisticated policies that are context dependent, which is basically saying really long time scales again, and things like that.

• 2:35Tim Scarfe?

  Yeah, you know, there's the kind of the philosophical highbrow notion of agency that we introduce notions of intentionality and self-causation and things like that. I mean, the really no-nonsense version of an agency is it's just it's just a thing which acts and performs some kind of computation. And I guess you could almost model anything as an agent, you know?

• 3:00Speaker 02

  Yeah, well, so if if if your definition of an agent is something that executes a policy, then anything is an agent, right? A rock is an agent, right? Everything has, you know, it's an input. A policy is an input-output relationship. When many people talk about agents, they they're adding a few they're adding a few additional elements that I think have a lot to do with how the policy is computed, right? So, for example, when we think of how the difference between like us and like like really like amoebas, we we often cite things like planning, counterfactual reasoning, goal-oriented behavior, right? We're specifying things that that have that that are specific mean

• 3:45Speaker 02

  that that are all related to how it is we compute our policies, right? There are latent variables that represent policies that are, you know, that are compatible with like well reinforcement learning, right? And and that's the defining characteristic of an agent. But you could very easily just sort of say like from an outside perspective, if you can't look at how someone or something is doing the computations, if the only thing you observe is the policy. Right, does that mean that you can never conclude that something's an agent? And I would say no, right, you'd still like to be able to conclude that this is an agent, even though the only thing I ever get to measure is its policy.

• 4:25Tim Scarfe?

  But do you think we should have some notion of the strength of an agent?

• 4:28Speaker 02

  The strength of an agent, or how, is this like a measure of agency? Is that what you mean?

• 4:32Tim Scarfe?

  Yes.

• 4:33Speaker 02

  Yeah, so, I mean, I think you could use like notions of like transfer entropy and things like that in order to estimate like the timetable over which something is incorporating information, or the degree to which it's taken into, it exhibits a context-dependent behavior and things like that. And that would be a pretty good measure. Now, is it normative? No, it's not, it's, it's, but it is a measure and you could use things like that. But at that point, you're really just talking again about policy sophistication, right? Not, does it have a reward function? Like, is it actually executing planning?

• 5:06Tim Scarfe?

  Yeah, I mean, certainly intuitively, agents to me seem to be kind of causally disconnected. Because they're planning into the future, they are not impulse response machines, they're not just, you know, part of the mass of things going on around them, they are just obviously disconnected from the locality.

• 5:25Speaker 02

  So here, the trick is, is that, okay, so I've got this agent and I know exactly what it does, right? It takes into, it takes into account information, it rolls out future, you know, internally, it rolls out a whole bunch of like future consequences of various different actions or plans that it could take, it selects the best one and then it executes it, right? So all of those variables, all of those variables that were, that occurred inside, right, from the outside perspective, it just looked like a function transformation, right? It's, I don't, unless I, unless I'm somehow going in and recording and somehow demonstrating the fact that the manner in which it is calculating its policy, you know, like involved doing those rollouts. Right. I wouldn't be able to show that it's actually doing those rollouts.

• 6:10Speaker 02

  I would just be able to conclude it has a really sophisticated policy. So can you conclude that something is... So the question is, how do you identify something as actually doing planning? And I think that's a really hard question, as opposed to having an incredibly sophisticated policy.

• 6:25Tim Scarfe?

  I think my intuition is, it feels to me that a function, a simple input-output mapping can't be an agent. And in a way, this is related to what we were talking about with grounding. You know, it seems that when things are physically embedded in the world, then they're more likely to be agents. This functionalist idea that just a bit of computer code running on a machine, it kind of feels like that can't be an agent.

• 6:48Speaker 02

  It does. So suppose I coded it up so it was doing all of that planning. It's like, gets its inputs, does some crazy, like massive Monte Carlo research, picks the best policy possible, and then executes it. Now, you don't observe any of that, right? Because you know what's going on, you could say, oh, well, it's clearly like executing, you know, this is, it's doing planning and counterfactual reasoning. It's going on, like, look, there it is, because you coded it, so you know it's doing it. But if you're looking at it from the outside, right? You know, if you don't know what's happening inside, it's going, you know, all you have access to is, oh, here's the action that it did, given this long series of inputs.

• 7:30Speaker 02

  And so it's really hard to identify what, you know, something as an agent per se, from the outside. You kind of have to know what's going on inside. This, by the way, is why I don't think that like, you know, these sort of prediction-based approaches to like AI are necess... You know, you could sort of say, well, it's not really doing anything even remotely agentic, unless it's executing and doing planning and counterfactual reasoning. So like, your chess program is like, oh, clearly it's doing some planning and counterfactual reasoning, because you know it's doing it. But... But you could, like, I could describe the exact same set of behaviors just with the policy function.

• 8:10Tim Scarfe?

  I think the counterfactual thing is an important feature here because we could take something which was conscious or something which had agency and we could just take a trace of the actual path which was found. And now we've just got, this is a reductio ad absurdum, but you know, now we've just got a computational trace and that thing clearly has now lost whatever agency or consciousness it had. So there's something about considering all of the possibilities.

• 8:33Speaker 02

  Yeah, I think, so in my mind, that is the fundamental feature of an agent. Like, if you can show that it's engaged in planning counterfactual reasoning, then it's definitely an agent. My argument is just simply that that's hard to do unless you crack it open and see what's going on inside. Now, you could take a pragmatic view and say, well, if the simplest computational model of the behavior, model it as if it was doing planning and counterfactual reasoning, then you can draw an implicit conclusion that, oh, yes, well, I may as well say it's an agent. And that's kind of the approach that I've taken. So like, one of the things that comes out of the physics discovery algorithm is that you apply it to agents and what do you get? Well, you get a model. Now, bear in mind, I called them all objects before and I didn't change anything to make

• 9:18Speaker 02

  it special to an actual agent, right? But what I do have the ability to do because of the model is I can look at the internal states associated with that object that I want to call an agent and look at how sophisticated it is, right? And that degree of sophistication is what allows me to say, oh, well, I'm going to go ahead and say that, like, and I like the whole idea, it's a great idea, like, it's have a metric, right? And I'm sure it would be something that would effectively be like transfer entropy or something like that, but we have this metric on like, well, how sophisticated were the internal states that were necessary in order to generate this output? And if it's above some threshold, we'll call it an agent. I don't like thresholds, but you know, we just sort of say a degree of agency, a degree of sophistication.

• 10:30Tim Scarfe?

  And coming back to Dennett's intentional stance, so this is that, you know, there is a level of representation which serves as a useful explanation, even though it's not actually, you know, the microscopic causal graph. And maybe we can agree that no agent can possibly be the cause of its own actions. But when there is a degree of planning sophistication, you know, macroscopically, it's as if it's the cause of its own actions.

• 10:54Speaker 02

  Yes, and that's why this "as if" phrase comes up a lot, right? I mean, this is, it's important to remember that, like, no matter how clever your model is, and no matter how clever your approach is, and how clever the words are that you use to describe it, a lot of this stuff is "as if", right? This is the best model, right? It's not, it's not, this is why, like, I repeat this over and over again, grind it into the students, right? Is that, you know, science is about like prediction and data compression, and like nothing else. And the same thing is going on here, right? You'll never, you know, just looking at behavior, you'll never know for sure in any meaningful way, like, whether or not it's, it's just doing a function transformation, or whether

• 11:37Speaker 02

  it's engaged in planning and counterfactual reasoning. But if your best model of it, if you sort of say, "Well, I tried to model as a function transformation, but goddamn it had a lot of parameters, right? But then I tried to model it as something that was just doing Monte Carlo tree search on the inside, and giving the answer, and that had like, you know, 40 parameters, and it's like, well, that's the model I'm going to go with, and now I'm going to call it an agent."

• 11:57Tim Scarfe?

  If we had a physical agent in the real world that was doing all of this planning and so on, would that have some kind of primacy to the computer simulation of agents that were doing all of this planning?

• 12:07Speaker 02

  Oh, is this, is this like, if I uploaded my brain onto a computer and didn't connect it to the world, would it still be thinking, even though it's like doing all of those things? Is that the idea here, or am I like way off?

• 12:18Tim Scarfe?

  That works, so yeah, let's say a high-fidelity computer simulation of Jeff. Would Jeff be an agent?

• 12:23Speaker 02

  No.

• 12:24Tim Scarfe?

  Oh, wasn't expecting to say that.

• 12:27Speaker 02

  Because I'm the agent, and if you upload it, no, I don't know. So... If you do a high fidelity computer simulation and you put it in my body, then I think I would have to say it's an agent.

• 12:39Tim Scarfe?

  Yeah.

• 12:39Speaker 02

  Right. If it's doing exactly the same, I mean, this is like the standard error. It's doing exactly the same calculations from a purely like phenomenal auto. It's like, it's the same. It's indistinguishable.

• 12:48Tim Scarfe?

  Okay. So agents need to be physical.

• 12:50Speaker 02

  So I do believe that an agent needs to be physical. Absolutely. I don't believe, you know, I believe you can have a model of agency and not have an agent. Right. You know, you can put that model in a computer and run it and make predictions as to what an agent would do. And it might even be 100% correct, but I still wouldn't call it an agent. But again, this is like getting into philosophy and like philosophy frustrates the Bayesian because philosophy is not probabilistic. Right. Philosophy is really about drawing clear lines and distinctions. And in my world, those don't really exist. Right. There's, everything has an error bar, you know, all of, there isn't a clear delineation between,

• 13:30Speaker 02

  you know, you know, an object and an agent. It's really, you know, from this modeling perspective, it's really just a question of degrees and philosophy is terrible at handling questions of degree.

• 13:42Tim Scarfe?

  My friend Keith, he's a big fan of computability theory and he thinks that an agent is basically, you know, like a type of computation and it has access to ambient state and it can take action and there's this kind of like sybnistic loop. And for him, the strength of the agency in the system is the compute type that the thing is doing. Right. So if it's, if it's a finite state automata, then it's a weak agent. If it's a Turing machine, it's a strong agent.

• 14:11Speaker 02

  Yeah. It's the degree of sophistication of the compute.

• 14:14Tim Scarfe?

  Pretty much. Yeah. Does that ring true to you?

• 14:16Speaker 02

  I mean, that if you were gonna, if you forced me like, you know, at the point of a gun, to put a measure on agency, it'd probably look a lot like that.

• 14:25Tim Scarfe?

  Yes. Jeff, let's talk about energy based models.

• 14:27Speaker 02

  Sure.

• 14:28Tim Scarfe?

  So Yan Le Cun, he had a monograph out, I think in 2006, talking about this.

• 14:32Speaker 02

  Oh, yeah.

• 14:33Tim Scarfe?

  Talking about this for a long time.

• 14:34Speaker 02

  Oh, yeah. When you fit your neural network to data, you know, via gradient descent. Right, then you have written an energy function in weight space, and you are following it to its energetic minimum. You know, the advantage of using an energy-based, taking an energy-based approach as opposed to taking, say, a straight-up like function approximation approach is that an energy-based model comes with something that's kind of like an inductive prior. Right, it basically, you know, an energy-based model, you know, if you're just doing function approximation, you're basically saying, there's any mapping from X to Y. X is my inputs, Y is my inputs. Any mapping is out there, I just want to figure out what it is. Right, now, in an, you know, in an energy-based model, right, you're effectively placing constraints

• 15:17Speaker 02

  on what that input-output relationship can be. I like thinking about the distinction between an energy-based model and a traditional sort of feed-forward neural network has to do with where your cost function is applied. Right, so in a traditional neural network, you're taking your inputs, you got your outputs, and the cost function is just a function of the inputs and the outputs. And the only thing that you're optimizing is the weights. In an energy-based model, there's another thing that your cost function operates on, and that's something, one of the internal states of your model. And as a result, like, in order to figure out what the best, you know, the best approach is, right, you actually have to do two minimizations. One that finds the energetic minimum associated with the part of the cost function that operates on the internal states,

• 16:03Speaker 02

  like the hidden nodes of your network. Right, and then one that is the prediction, that is your like effective prediction error. This is very much consistent with the approach that a Bayesian would take, right? You have a prior probability distribution, which gives you an energy function for every single latent variable in your model, and you are optimizing with respect to all of them. So you take a probabilistic approach. Good examples of this are like a variational autoencoder. A variational autoencoder, I think, is the best example of the most commonly used energy-based model out there. Why? Because you have an encoder network, you have a decoder network, right? And your cost function is based on the difference between inputs and outputs, right? So that's just like, it's fine, that's still regular. But it also is how Gaussian, and it, well, it depends on what flavor of V8.

• 16:46Speaker 02

  But you also have some part of your Cost function is a function of the actual internal representation, right? In a traditional VAE, it's how Gaussian is it? You want that internal representation to be as Gaussian as possible. If it's a VQVAE, then it's like mixture of Gaussian. But it's still like a cost function that is applied on the internal states as well as on the inputs and outputs.

• 17:08Tim Scarfe?

  Very cool. So a VAE is a fairly canonical example of an energy based model. And what you were saying about the, I mean, you know, the whole DL world is obsessed with test time inference at the moment. And in a way that that is a step towards what you're talking about. So.

• 17:21Speaker 02

  Yeah, you're treating, yeah, you're treating some of the weights of your model, right? I mean, yeah, you're treating some of the weights of your model as if they're latent variables, right? Because when you, when you show a new input, right? You're allowed to change some of the weights without looking at the output, right? And so what are you doing? Well, you're treating the weights as latents. Now, I think that like, which makes it a great trick, in my opinion. It's like, oh, great. I gather they're, they're, they're, they're moving in the direction of energy based models. I love it. The only thing I don't like about test time training is the vast majority of the training that is done. So in a traditional energy based model, you always find the minimum with respect to the latent variables. Like these extra weights that, you know, which in this case, which in the case of test time training is the, you know, the subset of weights that you're allowed to, to change during, you know, during test time.

• 18:07Speaker 02

  When you do the training for a traditional energy based model, you're allowed to make those changes, right? Throughout the entire course of training. The way that we're often doing test time training these days is we just do regular old neural network learning. Like we don't do, and then, and then, and then finally, when it comes to, when we get to the deployment phase, then we suddenly turn on, right? These additional latents, which are basically some of the weights of the network. And we do additional, an additional bit of learning at that point. This seems monument, now again, not an expert here, right? But this seems unwise to me. And the reason it seems unwise is because you didn't train the original network with that on. Right, you trained it in a completely supervised way.

• 18:50Tim Scarfe?

  Yes.

• 18:50Speaker 02

  Now, I'm sure that people are aware of this and it's been addressed in the literature, but I'm not personally aware of that. I don't think that's how it's used in practice.

• 18:59Tim Scarfe?

  We should also introduce this term transduction. So my definition of transduction is that you're actually doing search or optimization as a function of the test samples. Like I interviewed Clement Bonnet, he had a VAE on Arc, you know, searching latent spaces. And he actually searched through the decoder as a function of the test sample. And because these models, they are maximum likelihood estimators, right? Which means they're always giving you a kind of smoothed out average. And there's so much information in the test sample. Let's just riff on the relationship between energy-based models and Bayesian inference. So of course, they have this advantage that you don't need to do this very expensive, intractable normalization.

• 19:39Speaker 02

  Yes.

• 19:39Tim Scarfe?

  Yes, tell me about that.

• 19:40Speaker 02

  My take on it is that an energy-based model and a Bayesian model have a lot in common, right? In many ways, like, energy, I mean, well, literally in physics, right? Energy is like log probability. Now, of course, there's the normalization factor that you don't need to worry about if you're just minimizing energy. And so the difference between, you know, like, which is sort of like, you know, in a Bayesian framework, that's like saying, well, you know, I'm not actually going to treat some of these latent variables in a probabilistic way. I'm just going to do maximum or map estimation on some of my variables and just be okay with that. And that's one way to interpret the relationship between an energy-based model and a properly Bayesian model. There's a happy medium here, though, right?

• 20:23Speaker 02

  And the happy medium is you can still treat it as if it's, you know, you don't have to just minimize the energy function, but you can calculate the curvature down there, too. Do a Laplace approximation and call yourself a Bayesian again, right? Yes, there is more computation involved, but we've got a lot of great tricks for making that totally tractable.

• 20:42Tim Scarfe?

  What's the relationship between the free energy in the free energy principle and the energy and energy based models?

• 20:47Speaker 02

  Regularization term, I think, is the short answer, right? No, so the difference, and if you're being very, very, very pedantic, the difference between an energy based, minimizing energy and minimizing free energy is that free energy has this additional entropy penalty term. Now, if you're just doing maximum likelihood estimation, if you're minimizing your energy function with respect to some particular, just we'll just pretend we're only worried about one variable, and I'm just going to get a point estimate and call it a day, do like, you know, some kind of map estimation to get that one thing, there's not that big of a difference, right? Because you're not, there is no probability distribution over the latent that allows you to compute that regularization term. But that's the only difference.

• 21:32Speaker 02

  It's are you regularizing or not is, I think, the easiest way to think about it.

• 21:36Tim Scarfe?

  So LeCun is a big advocate of JEPHA, so these joint embedding prediction architectures using this non-contrastive learning where essentially the learning objective is comparing the latence of observed and unobserved parts of the space. This is an architectural design.

• 21:53Speaker 02

  Well, what is, okay, so what does JEPHA stand for? It's Joint Embedding and Prediction Architecture. There we go. So what's the joint embedding bit about? Well, the joint embedding bit about is, you know, is well, I'm going to take my inputs, I'm going to take my outputs, and I'm going to embed them in some space, right? And then I'm going to learn a prediction between the two embeddings. And that's a great idea. It's a great idea because it has some of the flavor of what we would like to get out of our models. We're like, we're not interested in predicting every, in many situations, I should be very particular about this, in many situations, we're not interested in predicting every single pixel on the image. We want to get, you know, maybe something that's a little more gestalt, a little more high level, a little more conceptual understanding of what's going on. And so emphasizing the goal of predicting every single pixel, which is what's typically

• 22:38Speaker 02

  done in generative modeling right now, you know, might lose some of the power, the extractive power of some of the networks. And so like, let's do, so the whole point of JEPHA, as I understand it, I'm sure there are other points, is that, is that you're going to take, you're going to, you're going to compress your inputs and compress your outputs, and then do all the learning in this compressed space. Love it, right? Science is about prediction and data compression. Let's make that compression explicit on the front end and the back end. The downside of this approach is that, is that Is that it doesn't work out of the box, right? Because it's very easy to find a compression or an embedding of the inputs and an embedding of the outputs for which prediction is perfect.

• 23:18Tim Scarfe?

  Yeah.

• 23:18Speaker 02

  Which is to basically make both of them zero. And so you have to do some other things. Other tricks need to be employed in order to make it work.

• 23:25Tim Scarfe?

  Yes, yes, I remember Lacun was talking about this. So there was there's the traditional contrastive method, which is from it's kind of Hinton's idea, apparently.

• 23:34Speaker 02

  Yeah.

• 23:34Tim Scarfe?

  Like the negative sampling and whatnot. And and that's very expensive because you actually have to do lots and lots of sampling. And this non-contrastive thing.

• 23:42Speaker 02

  Yeah, this is this, by the way, is what he should have won a Nobel Prize for.

• 23:45Tim Scarfe?

  Right.

• 23:47Speaker 02

  In my opinion.

• 23:48Tim Scarfe?

  Yes.

• 23:48Speaker 02

  Because the whole the whole point of of the wake-sleep algorithm and contrastive divergence was that, oh, it's actually biologically plausible, right? It was it was it was an end run around the need to do backprop. And that's what made it so clever and interesting, in my opinion.

• 24:02Tim Scarfe?

  Lacun is a big fan of this non-contrastive thing where you work in the latent space. There are many different algorithms that do this. We had a whole load of shows all about non-contrastive learning. There's things like V-Creg and BYOL and BiLo Twins. And there's there's an entire thread of research all around that. And in many different ways, what they're trying to do is avoid this mode collapse problem that you're talking about. And they use different forms of regularization.

• 24:26Speaker 02

  There's an old school way of accomplishing the same thing. And that is that is to to do all of your is it's it's called preprocessing, right? And this is this is something that a lot of people do. You take your data and in fact, we do this all the time with with with like vision language models, right? So we want to do we want to use an LLM and we want to predict images. So what do we do? Well, the first thing we have to do is tokenize the image, right? And so what do we do? We run a VA that we do the preprocessing and we do it by. It is the preprocessing step is completely independent, right? From the actual algorithm that's going to be the be be tasked with solving the problem of interest. And, you know.

• 25:09Speaker 02

  That's not something that we necessarily have to stick with, right? It would be very nice if there was a way, if there was a way of like, again, well jointly, we're getting right back to Jeff again. What we'd like to do is we'd like to choose our preprocessing algorithm in a manner that, you know, you know, not a priori, not do it first. We like to choose the preprocessor that works the best in this space. And I think that that's the ultimate motivation for a lot of this work is like, what's the right embedding? One of my favorite tricks, like, of course, I, you know, I pre-process the VAEs all the time. In fact, it's, you know, the second, every time someone hands me a new neural data set,

• 25:51Speaker 02

  the first thing I do, I'm, you can, I'm not ashamed to admit, I run PCA on it and pass it through a VAE and then sort of take a look, right? It's the first thing you do with your data because it gives you a good idea of what the signal to noise ratio is in the data set itself. Yes. And then, yeah, and then what do I do? I subsequently do most of my analysis, right? In that discovered embedding space. And there's, I don't see a huge problem with that from a purely pragmatic perspective, but it's certainly cleaner, right? To have a single algorithm and approach and not just be stringing these sort of things together in an ad hoc way. There's, you know, when, when doing PCA, PCA is a really great example of this.

• 26:33Speaker 02

  There's a failure mode for principal component analysis, which is actually really common in neural data. Because principal component analysis basically is, well, where's the most variability? Okay, I'm worried about that. And then all the stuff that's not varying very much, I'm just gonna throw it away, right? It's just like, you know, dimensions in which there's low variability are not important. Well, it turns out that in neural data, the dimensions in which there's very little variability are some of the most important dimensions. And so, preprocessing with PCA runs a risk of throwing out the most valuable information in your data set.

• 27:05Tim Scarfe?

  Yes.

• 27:06Speaker 02

  And so there's a lot of wisdom in jointly, right? In jointly fitting your preprocessing model as well as your inference and prediction model.

• 27:15Tim Scarfe?

  I mean, on this subject of not throwing things away. And JEPAR and non-Contrastive learning, it's part of this bigger field of self-supervised learning. And we want to learn representations that maintain fidelity and richness. And Le Cun's hypothesis is that when you do something like supervised learning with, you know, some particular downstream task in mind, the neural network gets wise. And what it does is it kind of discards all of the long-tail stuff that aren't relevant for that particular task. So when you train these models, what you're trying to do is sort of maintain enough ambiguity so that it compresses the information, but it also maintains enough fidelity to work broadly for different things.

• 27:54Speaker 02

  Yes, and that is a laudable goal, right? And I certainly share it, right? The last thing you want to do is, I mean, you know, fortunately, like networks are so big, we don't really run the risk of like overfitting as much as we used to. But the last thing you want to do is train your network to toss information that you might need down the road. That said, like the vast majority of what, you know, the brain does, just like these neural networks, is decide what information is currently task irrelevant. But that's all the more reason to do things in a self-supervised or unsupervised way, right? Because you're basically not telling it, this is the importance, you know, you're not telling it like what's all task relevant and task irrelevant.

• 29:07Speaker 01

  Visible puts the ultimate wireless hack in the palm of your hand. With unlimited 5G wireless data and hotspot designed to keep you connected. Powered by Verizon's 5G network. And no contract holding you back. Plans start at $25 a month. Or get the premium Visible Plus Pro Plan and save $10 on your first month when you use promo code HACK. Make the switch at visible.com. Terms apply. See visible.com for details.

• 29:37Tim Scarfe?

  So I interviewed Sholay about the version 2 of the ARC challenge. And one thing that struck me is I think of intelligence as being multi-dimensional. So version 1 got saturated. The ARC was actually really amazing because it's the only intelligence benchmark that has survived for five years before being defeated. Since the advent of these thinking models, it has been defeated very quickly. But they're working on version 3 and there'll be version 4, there'll be version 5. Will there always just be something left over?

• 30:08Speaker 02

  That sounds like another philosophical... So yes is my answer. There will always be, there will always be something left over. In the sense that like, you know, we have this, this has been the trajectory things have been going for a really long time. Right? It's sort of like, we get algorithms that do amazing new cool things. And then someone comes along and says, "Yeah, but it can't build me, it can't pull a rabbit out of a hat." Right? And then, of course, what does someone do? They, oh, they, they figure out the new training protocol, slightly different architecture, or they just train it to pull rabbits out of hats and then suddenly it can. And then someone proposes a new challenge, and a new challenge, and a new challenge. And it's always this game of like one-upsmanship. So the question becomes, well, what's the point at which there are no more new challenges?

• 30:51Speaker 02

  And I'm not entirely certain we're ever going to get there. Right? It may very well be the case that we get, you know, these sort of algorithms that are capable of replicating the complete suite of human behaviors and then someone will come up with some criticism like, "Yeah, but it's not really doing X, it's just faking it." Right? This is just the direction things go because people really do think they're important.

• 31:12Tim Scarfe?

  Yeah. Do you think that the concept of recursive, self-improving intelligence is a valid one?

• 31:17Speaker 02

  Yes. I do think that is, so, so I think that one of the most critical missing elements right now is some form of continual learning. Right, at the end of the day, you really want an algorithm that doesn't just learn on the training that on the training set and then just gets deployed. You want something that runs around in the world and comes across things that it doesn't understand, right? And then is able to incorporate, to build, you know, append its model in some sense, right? So this is like the, you know, and there are some approaches that is all based on like Bayesian non-parametrics and Dirichlet process priors and stuff like that, where you you sort of see something that's surprising or unique or different, something you didn't expect, and it causes you to say, I need to turn learning on because I got to figure

• 32:02Speaker 02

  this out. That is an absolutely critical element that we need to be developing. We are developing that. It turns out that that's one of the nice things about this sort of object-centered physics discovery thing is because it's object-centered, if it comes across a new situation that it does not understand, it is capable of instantiating a completely brand-new object just to explain this new situation.

• 32:22Tim Scarfe?

  Continually learning agents can acquire new knowledge autonomously and the whole, you know, the whole thing just learns more knowledge. But intelligence feels different. It feels like in the system that we've been describing, the intelligence is the way we're implementing the, you know, the Bayesian updates and, you know, actually building the algorithms. Could the systems on their own meta-program themselves and develop better algorithms or something like that?

• 32:49Speaker 02

  That's a very good question. Something that would be closer to true artificial intelligence than what we currently have would be capable of building models on the fly to deal with new situations, to taking things that it knows about, right, and combining them in new and different ways. There are approaches that have some of that aspect to it, like G-Flo nets from like Benjio stuff is like, is like a great example of something that, at least in principle, is a generative model of generative models, right? It's sort of like, oh, like, you know, I might actually need a new node, like it's time to create a new latent variable because like, like the current set's just not cutting the mustard anymore. Those are things that I think are hallmarks of true intelligence.

• 33:32Speaker 02

  I don't want to ever make the statement, as soon as it's got that, it's truly intelligent. I will never, ever, ever say that. But I do think that that is a critical component that needs to be present, right, is the ability to generate new models. On the fly to deal with novel situations and data. Most of that, you know, as well as the ability to combine old models, previous models in new and interesting ways. This is actually how the brain evolved, right? We started out with like, you know, really simple brains and there were different regions and they solved sort of different problems. And what eventually happened as we evolved is that these different regions of the brain

• 34:15Speaker 02

  learned to communicate with each other in new ways and through that communication acquired new abilities, right? And then eventually evolved into, you know, new capabilities and things like that, right? I often like to point out to the, I think olfaction is like the sense that's not studied nearly enough. It's an incredibly old part of the brain. And arguably, right, it's the, it's the first part of the brain that evolved the ability to do proper like associative processing, right? Odor, unlike visual space, right, where there's translation symmetries and all that sort of stuff and things are smooth. Olfactory space that does not exist, right? It's, it's really, really, really combinatorial and complicated. And the part of the brain that evolved to solve the olfactory problem, arguably, is

• 35:00Speaker 02

  the part that evolved into our frontal cortex. Don't quote me on that. There's a lot of disagreement there. That's just my take. But it certainly has a lot of the features that we associate with associative cortex, right? It is, wow, I just said like six uses, three different uses of the word associate in that sentence, but, but I think you see what I mean, right? It, it was all about like taking old capabilities, right? Combining, combining, you know, simple models and modules to create something that was more complex. And then over time, right? So, so that was what made the brain work, right? It was all about taking little things that worked and combining them in new and different ways in order to evolve, you know, effectively an emergent, you know, emergent properties,

• 35:44Speaker 02

  emergent, you know, computational abilities and an emergent understanding of the world in which we live. And I do think that like what, what, you know, if, when we get to the point where we start really saying Oh, this is actually truly intelligent. It's going to have that feature. It's going to have the ability to have a. It's going to have a modular description of the world, and it's going to have the ability to to combine those modules in a way that creates a more sophisticated understanding. It's like Legos, right? I can, you know, the Lego bricks all connect in certain ways, and I can build like all sorts of new and amazing things that were never built before, right? Out of them. That's the capability that we have. And that's the essence of like creativity. It's why I refer to systems engineering as like the thing we really want our AI models to be able to do.

• 36:30Tim Scarfe?

  Collective intelligence is a bit different. We have this plasticity, right? We can adapt our behavior day by day. We might see some kind of meta learning or some kind of change in our organization dynamics. You know, maybe some agents will specialize. And it might be an existence proof of this kind of recursive, you know, superintelligence that we're talking about.

• 36:48Speaker 02

  Yeah, I do. I think that's absolutely correct, right? Is that, you know, so the specialization is great. In fact, I would argue that specialization is how we got all of this, right? And this was I'm pointing at London in case you. There was some confusion there.

• 37:03Speaker 02

  Right, it was it was really about, you know, the interconnected, highly specialized intelligences that are people and their ability to learn how to to work together that that, you know, gave rise to the technological revolution. The brain is the same way, right? It's in my view, it's highly specialized little modules or agents that are capable of of of being repurposed, reused, capable of communicating with one another in order to solve really complicated problems. But there's always a benefit to specialization. I don't believe in like, like AGI. AGI seems like a bit of a misnomer to me. What we really want is not artificial general intelligence. We want collect. We want collective specialized intelligences.

• 37:47Tim Scarfe?

  What about scientific discovery? Do you think that we could? I mean, what would the world look like when we could discover new drugs? We could discover new knowledge in science?

• 37:55Speaker 02

  You know, right now, the way that we're doing that is is largely focused on summarizing vast troves of data and looking for correlations that are present in it. I think the next major milestone in this trajectory is is experimental design, right? Not just, oh, well, here's here's some correlations you may not have seen because they're really small, and this is what computers are good at. They're really good at identifying small but highly relevant correlations. And the next step, of course, is constructing a system that tests these hypotheses explicitly, right, and generates the experiments that will identify, like, that will, that will fill in the gaps of our knowledge. And all of this, I believe, can in fact be automated in a very sensible way.

• 38:38Speaker 02

  I, you know, I don't see any, like, major obstacles to automate an empirical inquiry, other than we probably want to place some safety constraints when we start letting them work, when we started letting the AIs run the labs, right? Because you never know. It's sort of, you always have this AI, it's like, well, you know, the most effective experiment to determine if this is correct is to set off a nuke. And that, that would be bad. Yes. Right, so pure empirical inquiry, right, does run risks like that, but I think that that's not, not, not the biggest issue. I think what we need to do is, we just had, need to have a nice concise framework for saying, like, oh, look, you know, like, I'll give you an example. So we had, we, we, we had this, um, a problem that popped up a while back. A gentleman we were talking to is, is, um, is, you've got these law, you, you know, you've got these robots,

• 39:22Speaker 02

  and the robots see something it's never seen before. And, and I, you know, so a robot is like running around, it comes across like a beach ball, never seen a beach ball in its entire life. And what you'd like is, you'd like the robot to know how to figure out that it's a beach ball, and to figure out what its properties are. And if you tell the robot, like, like, if you see something new, just stop. But you're kind of, then that's, that's no good. Right, what you really want to do is, you want to figure out a relatively non-invasive procedure for the robot to like, poke, do what a child would do. What does a kid do when they see a beach ball, right? They run up and they poke it, and they say, oh, right, yeah, and then it moved, and, and it actually learned, it actually experiments with its environment for the purposes of identifying the properties of the objects that exist in it.

• 40:07Speaker 02

  Now, I do think we probably want to test this out virtually before it's deployed in the real world, because you never know, it might very well be that the optimal experiment, experiment is to run up and kick it as hard as you possibly can. And we, we certainly want to avoid that, but like, something along those lines, something, you know, a robot that is able to test the theories that it has about how things work in an online way, and learn from those results in an online way is definitely part of the goal.

• 40:34Tim Scarfe?

  Looking forwards, what do you think the future will look like when we have more autonomous AIs among us? A lot of people worry about enfeeblement, loss of control, you know, it making us dumb, all of this kind of stuff.

• 40:46Speaker 02

  I do, I do worry about AI making us dumb, right? I mean, offloading, offloading your thinking onto a machine, which is something that, that, that AI allows is, is, is is a potentially a big problem. I, I don't really want to have a situation where humans are reduced to like, they're, they're just reduced to like value function selectors. They're just basically going, oh, no, I don't like that outcome, like do this instead. I do want to see a future where, where, where we have an AI that actually improves our understanding of the world. And simply automating everything runs the risk that you specified, right? It runs the risk of people becoming couch potatoes that just watch TV and occasionally say like, yeah, you know, these chips are no good.

• 41:28Speaker 02

  That seems like a bad outcome to me.

• 41:32Speaker 02

  I worry less about that, I think, than some because people are remarkably adaptable, right? I mean, I, you know, you have all these arguments about like, oh, you know, this new technology comes along and it's going to completely destroy this way of life. And, you know, and that's going to be awful for people, and it is maybe in the short term. You know, I think of like tractors, right? Or just go back, how many hundred years you have to go back when like 99% of people were involved in agriculture? And now it's like, what, two, right? I consider that a solid improvement, right? Because it allowed the rest of us to, it allowed us to do a bunch of other things that we find more satisfying, that are more interesting. It allowed us to like, you know, like I can read, you know, spend some time reading a book, don't have to labor in the fields all day.

• 42:18Speaker 02

  That's the future that I sort of see, and that's the future that I hope for, is that, is, is one in which, you know, all of these artificial agents running around and doing things autonomously are there to, to free us up to pursue more interesting, more, you know, you know, to improve ourselves in, in, in, in more interesting ways. But at the end of the day, it's just another technol, you know, at least initially, it'll just be another technology like the tractor. Now, a hundred years from now, who knows?

• 42:48Tim Scarfe?

  What will the value of work be if the AIs can do everything and there's nothing left for us to do?

• 42:53Speaker 02

  I don't think that will ever be the case that the AIs can do everything. Like I said, the future I worry about is one where like it's, you know, the sole role of people is like sitting around like making sure the AIs aren't going rogue. And, and, and things like that, which I don't consider a good outcome. I would really like to see human improvement. You know, I, I envision a future of, I don't know, this is like cybernetic transhumanism, if I'm going to go sci-fi on this, right? Where, where, you know, the technology and us evolved together in a way that's beneficial for both. That's the goal. You know, are there these dystopic possibilities where like, oh, well, what are humans in a world where, well, what are they? What are, what are humans in a world where everything can be done by a robot?

• 43:38Tim Scarfe?

  Yeah.

• 43:38Speaker 02

  You know, that's, that's a good question. And that's, and at the end of the day, right? They end up just becoming like reward function selectors, right? They end up just sort of saying, oh, I don't like this and I do like that. And they're basically, you know, I mean, you end up with this, it is another nightmare scenario. I don't like talking about these dystopian futures because honestly, I think people are too clever. And I think people are too motivated and people are too interested in how the world really works. And then people are too interested in actually understanding things that they will never stop. That they, that AI will become a partner, not an adversary or a crutch. And that's, that's, that's what I think will happen because that's, but that's a statement more about my belief about humans than it is about my belief about the development of AI.

• 44:22Speaker 02

  You know, I am a techno optimist, if you will, not a, not a pessimist. I, I believe that we will find a way to adapt to an ever-changing world as we have done for millions of years, including one that includes technology that alleviates most of our labors.

• 44:38Speaker 00

  This episode is brought to you by Indeed. You're ready to level up your business, but the pressure to find the right person to make that happen is intense. Indeed sponsored jobs can help carry the load. It boosts your job post to help it reach quality talent fast. Cut through the chaos with Indeed sponsored jobs. Get a $75 sponsored job credit at indeed.com/podcast. Terms and conditions apply. Hiring now, then this is a job for Indeed sponsored jobs.

• 45:07Speaker 01

  Visible puts the ultimate wireless hack in the palm of your hand. With unlimited 5G wireless data and hotspot designed to keep you connected. Powered by Verizon's 5G network. And no contract holding you back. Plans start at $25 a month. Or get the premium Visible Plus Pro Plan and save $10 on your first month when you use promo code "HACK". Make the switch at visible.com. Terms apply. See visible.com for details.

• 45:38Tim Scarfe?

  On that, there's an AI literacy thing because AI has moved so quickly now that certainly my parents don't understand anything about it. But by the same token, policymakers don't understand anything about it. And there are people saying AI is going to kill everyone. And there's people making negative arguments, there's people making positive arguments. So there's a bit of a fog of war now because there are so many people saying different things about AI. How should they make sense of all of this?

• 46:02Speaker 02

  Well, we are now well outside my area of expertise. So I'm just going to say that before I say anything else. AI is developing very quickly. But I am much more concerned about what people will do with the new technology than I am with what the technology will do all by itself. I don't have this big concern about, I don't really believe that like, you know, Skynet's going to take over, the internet's going to suddenly become conscious and kill us all. Right? In part because, you know, AI is not that advanced, but also because we are telling, you know, we are still in the position where we specify the goals of the system. And that will likely continue for a very long time.

• 46:46Speaker 02

  And it will always be the case that these systems, you know, will can be, you know, are subject to review. We will always keep an eye on them. They will always, at least initially, be released in relatively restricted domains and where we're testing, where we're keeping a close eye on what it is that they are and are not doing. So I don't worry too much about like the going rogue. I worry a lot more about somebody building, you know, it's sort of like a virus, which we already have to deal with. Like somebody builds like some insane virus and like takes down the internet. I'm more worried about malicious human actors than I malicious AI actors. Cause at the end of the day, all of these algorithms, they simply do what they are told.

• 47:29Speaker 02

  Right? We train them, we tell them, here's your objective function. As long as we are specifying the objective function and we understand the objective function, we're probably going to be okay. I think the safest way to deal with AI concerns is to tell people, hey, look, this AI is just doing what we told it to. We, you know, we, we set it up to make really good predictions and to achieve these outcomes. Now, is it dangerous to like specify these outcomes without being very, very, very careful? Yes, it is, right? That's, this is the whole like, hey, Skynet end world hunger and it kills all humans. That, that's a, that is, that, that is a real possibility, but whose fault was that? The fault was the person who like was very, very naively specified their goals. There are in fact relatively straightforward ways to specify the, the reward function

• 48:14Speaker 02

  that, that don't run that risk nearly as badly. And the best one is, so are you familiar with like maximum entropy inverse reinforcement learning? I like to call it active inference because it's really similar. And so there, what you're doing is you're basically observing someone's policy. And then you're trying to do a maximum entropy model. You're, you're doing maximum model on the reward function itself. At the end of the day, what ends up happening when you do this is, this is why it's like basically just like active inference. You get a reward function. So you have some, you know, organism or whatever, and you're trying to do this for it. And, and it's got some stationary distribution over actions and outcomes, right? It's inputs and outputs of the stationary distribution. That becomes your reward function, like not directly, there's some math involved, but basically your reward function is a function of the study state distributions

• 48:59Speaker 02

  over actions and outcomes. So we could do this, right? We could take the, we could take the current manner in which humans are making decisions. And we could write down, right? What's the stationary, what, what is the current estimate of the stationary distribution of our actions and outcomes? So this would include things like everyone's getting, you know, this number of people are going hungry, this, you know, and, and, you know, all the stats that describe like the inputs and outputs to our policy make, you know, to our policy distribution. And then we could just ask an AI.

• 49:27Speaker 02

  Your reward function is the one that results in the same outcome that we currently have, right, on average. And it would execute it and it would, and to the extent that it works, right, it would ultimately result in an AI algorithm that just sort of is like mimicking human behavior, right, or at least achieving the same outcome that we were achieving before. Now, here's the safe way to like improve the situation. You don't say end world hunger, right? You perturb that distribution over outcomes, right? And just over outcomes a little bit. And then you evaluate the consequences, right? It's all you're doing. You make these little changes in the reward, in an empirically estimated reward function,

• 50:11Speaker 02

  right, rather than just sort of specifying one by hand, because that's the dangerous thing.

• 50:15Tim Scarfe?

  Jeff, thank you so much for joining us.

• 50:16Speaker 02

  It's my pleasure.

• 50:18Tim Scarfe?

  Amazing.

• 50:27Speaker 04

  When you shop Kroger delivery, you can expect the savings you love and fresh groceries delivered right to your door. From delicious produce to everyday essentials, enjoy the convenience of delivery and the trusted selection you've come to know from Kroger. Because bringing you fresh, quality groceries is what we do best. Right now, enjoy free delivery on orders of $50 or more with your digital coupon. Restrictions apply. See site for details. Kroger, fresh for everyone.