Ian Goodfellow: Generative Adversarial Networks (GANs)

Ian Goodfellow is the author of the popular textbook on deep learning (simply titled “Deep Learning”). He coined the term Generative Adversarial Networks (GANs) and with his 2014 paper is responsible for launching the incredible growth of research on GANs. Video version is available on YouTube. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter https://twitter.com/lexfridman, LinkedIn https://www.linkedin.com/in/lexfridman/, Facebook https://www.facebook.com/lexfridman, Medium https://medium.com/@lexfridman, or YouTube https://www.youtube.com/lexfridman where you can watch the video versions of these conversations.

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

• 0:00Lex Fridman?

  The following is a conversation with Ian Goodfellow. He's the author of the popular textbook on deep learning simply titled Deep Learning. He coined the term of generative adversarial networks, otherwise known as GANs. And with his 2014 paper is responsible for launching the incredible growth of research and innovation in this subfield of deep learning. He got his BS and MS at Stanford, his PhD at University of Montreal with Yashua Benjo and Aaron Kurville. He held several research positions, including at OpenAI, Google Brain, and now at Apple as the Director of Machine Learning. This recording happened while Ian was still a Google Brain.

• 0:45Lex Fridman?

  But we don't talk about anything specific to Google or any other organization. This conversation is part of the Artificial Intelligence Podcast. If you enjoy it, subscribe on YouTube, iTunes, or simply connect with me on Twitter, at Lex Fridman, spelled F R I D. And now here's my conversation with Ian Goodfellow.

• 1:25Lex Fridman?

  You open your popular deep learning book with a Russian doll type diagram that shows deep learning is a subset of representation learning, which in turn is a subset of machine learning and finally a subset of AI. So this kind of implies that there may be limits to deep learning in the context of AI. So what do you think is the current limits of deep learning? And are those limits something that we can overcome with time?

• 1:52Speaker 01

  Yeah, I think one of the biggest limitations of deep learning is that right now it requires really a lot of data, especially labeled data.

• 2:01Speaker 01

  There are some unsupervised and semi-supervised learning algorithms that can reduce the amount of labeled data you need, but they still require a lot of unlabeled data. Reinforcement learning algorithms, they don't need labels, but they need really a lot of experiences. As human beings, we don't learn to play pong by failing at pong 2 million times. So just getting the generalization ability better is one of the most important bottlenecks in the capability of the technology today. And then I guess I'd also say, deep learning is like a component of a bigger system. So far, nobody is really proposing to have only what you'd call deep learning as the entire ingredient of intelligence.

• 2:42Speaker 01

  You use deep learning as submodules of other systems, like AlphaGo has a deep learning model that estimates the value function. Most reinforcement learning algorithms have a deep learning module that estimates which action to take next, but you might have other components.

• 2:59Lex Fridman?

  You're basically building a function estimator. Do you think it's possible, you said nobody's kind of been thinking about this so far, but do you think neural networks could be made to reason in the way symbolic systems did in the 80s and 90s to do more, create more like programs as opposed to functions?

• 3:18Speaker 01

  Yeah, I think we already see that a little bit. I already kind of think of neural nets as a kind of program. I think of deep learning as basically learning programs that have more than one step. So if you draw a flow chart, or if you draw a TensorFlow graph describing your machine learning model, I think of the depth of that graph as describing the number of steps that run in sequence, and then the width of that graph is the number of steps that run in parallel. Now, it's been long enough that we've had deep learning working that it's a little bit silly to even discuss shallow learning anymore. But back when I first got involved in AI, when we used machine learning, we were usually learning things like support vector machines.

• 4:00Speaker 01

  You could have a lot of input features to the model, and you could multiply each feature by a different weight. All those multiplications were done in parallel to each other. There wasn't a lot done in series. I think what we got with deep learning was really the ability to have steps of a program that run in sequence. And I think that we've actually started to see that what's important with deep learning is more the fact that we have a multi-step program, rather than the fact that we've learned a representation. If you look at things like ResNets, for example, they take one particular kind of representation and they update it several times. Back when deep learning first really took off in the academic world in 2006, when

• 4:43Speaker 01

  Jeff Hinton showed that you could train deep belief networks, everybody who was interested in the idea thought of it as each layer learns a different level of abstraction, that the first layer trained on images learns something like edges and the second layer learns corners and eventually you get these kind of grandmother cell units that recognize specific objects. Today, I think most people think of it more as a computer program where as you add more layers, you can do more updates before you output your final number. But I don't think anybody believes that layer 150 of the ResNet is a grandmother cell and layer 100 is contours or something like that.

• 5:23Lex Fridman?

  Okay, so you're not thinking of it as a singular representation that keeps building? You think of it as a program, sort of almost like a state, representation as a state of understanding?

• 5:35Speaker 01

  Yeah, I think of it as a program that makes several updates and arrives at better and better understandings, but it's not replacing the representation at each step, it's refining it. And in some sense, that's a little bit like reasoning. It's not reasoning in the form of deduction, but it's reasoning in the form of taking a thought and refining it and refining it carefully until it's good enough to use.

• 5:58Lex Fridman?

  So do you think, and I hope you don't mind, we'll jump philosophical every once in a while, do you think of cognition, human cognition, or even consciousness as simply a result of this kind of sequential representation learning? Do you think that could emerge?

• 6:17Speaker 01

  Cognition, yes, I think so. Consciousness, it's really hard to even define what we mean by that. I guess there's consciousness is often defined as things like having self-awareness. And that's relatively easy to turn into something actionable for a computer scientist to reason about. People also define consciousness in terms of having qualitative states of experience, like qualia. And there's all these philosophical problems like, could you imagine a zombie who does all the same information processing as a human, but doesn't really have the qualitative experiences that we have? That sort of thing, I have no idea how to formalize or turn it into a scientific question. I don't know how you could run an experiment to tell whether a person is a zombie or not.

• 7:02Speaker 01

  And similarly, I don't know how you could run an experiment to tell whether an advanced AI system had become conscious in the sense of qualia or not.

• 7:10Lex Fridman?

  But in the more practical sense, like almost like self-attention, you think consciousness and cognition can, in an impressive way, emerge from current types of architectures that we're going to be learning.

• 7:22Speaker 01

  Yeah, or if you think of consciousness in terms of self-awareness and just making plans based on the fact that the agent itself exists in the world, reinforcement learning algorithms are already more or less forced to model the agent's effect on the environment. So that more limited version of consciousness is already something that we get limited versions of with reinforcement learning algorithms if they're trained well.

• 7:51Lex Fridman?

  But you say limited, so the big question really is how you jump from limited to human level.

• 7:58Speaker 01

  Yeah.

• 7:59Lex Fridman?

  Right, and whether it's possible, you know, even just building common sense reasoning seems to be exceptionally difficult. So if we scale things up, if we get much better unsupervised learning, if we get better at labeling, if we get bigger data sets, more compute, do you think we'll start to see really impressive things that go from limited to? You know, something echoes of human level cognition.

• 8:27Speaker 01

  I think so, yeah. I'm optimistic about what can happen just with more computation and more data. I do think it'll be important to get the right kind of data. Today, most of the machine learning systems we train are mostly trained on one type of data for each model. But the human brain, we get all of our different senses and we have many different experiences like, you know, riding a bike, driving a car, talking to people, reading. I think when you get that kind of integrated data set, working with a machine learning model that can actually close the loop and interact, we may find that algorithms not so different from what we have today.

• 9:09Speaker 01

  Learn really interesting things when you scale them up a lot and train them on a large amount of multimodal data.

• 9:15Lex Fridman?

  So multimodal is really interesting, but within, like your work in adversarial examples. So selecting within modal, within one mode of data,

• 9:28Lex Fridman?

  selecting better at what are the difficult cases from which you're most useful to learn from?

• 9:33Speaker 01

  Oh, yeah, like, could we, could we get a whole lot of mileage out of designing a model that's resistant to adversarial examples or something like that?

• 9:41Lex Fridman?

  Right.

• 9:41Speaker 01

  Yeah.

• 9:42Lex Fridman?

  Good question.

• 9:43Speaker 01

  My, my thinking on that has evolved a lot over the last few years.

• 9:46Lex Fridman?

  Oh, interesting.

• 9:47Speaker 01

  When I first started to really invest in studying adversarial examples, I was thinking of it mostly as adversarial examples reveal a big problem with machine learning. And we would like to close the gap between how machine learning models respond to adversarial examples and how humans respond. After studying the problem more, I still think that adversarial examples are important. I think of them now more of as a security liability than as an issue that necessarily shows there's something uniquely wrong with machine learning as opposed to humans.

• 10:19Lex Fridman?

  Also, do you see them as a tool to improve the performance of the system? Not on the security side, but literally just accuracy?

• 10:27Speaker 01

  I do see them as a kind of tool on that side, but maybe not quite as much as I used to think. We've started to find that there's a trade-off between accuracy on adversarial examples and accuracy on clean examples. Back in 2014, when I did the first adversarially trained classifier that showed resistance to some kinds of adversarial examples, it also got better at the clean data on MNIST. And that's something we've replicated several times on MNIST, that when we train against weak adversarial examples, MNIST classifiers get more accurate. So far, that hasn't really held up on other data sets and hasn't held up when we train against stronger adversaries. It seems like when you confront a really strong adversary,

• 11:12Speaker 01

  you tend to have to give something up.

• 11:15Lex Fridman?

  Interesting. But it's such a compelling idea because it feels like that's how us humans learn, through the difficult cases.

• 11:23Speaker 01

  We try to think of what would we screw up and then we make sure we fix that. Yeah. It's also in a lot of branches of engineering, you do a worst-case analysis and make sure that your system will work in the worst case. And then that guarantees that it'll work in all of the messy average cases that happen when you go out into a really randomized world.

• 11:44Lex Fridman?

  Yeah, with driving with autonomous vehicles, there seems to be a desire to just look for, think adversarially, try to figure out how to mess up the system. And if you can be robust to all those difficult cases, then you can... It's a hand-wavy empirical way to show your system is safe.

• 12:04Speaker 01

  Today, most adversarial example research isn't really focused on a particular use case, but there are a lot of different use cases where you'd like to make sure that the adversary can't interfere with the operation of your system. Like in finance, if you have an algorithm making trades for you, people go to a lot of an effort to obfuscate their algorithm. That's both to protect their IP because you don't want to research and develop a profitable trading algorithm than have somebody else capture the gains. But it's at least partly because you don't want people to make adversarial examples that fool your algorithm into making bad trades.

• 12:41Speaker 01

  Or I guess one area that's been popular in the academic literature is speech recognition. If you use speech recognition to hear an audio waveform and then turn that into a command that a phone executes for you, you don't want a malicious adversary to be able to produce audio that gets interpreted as malicious commands, especially if a human in the room doesn't realize that something like that is happening.

• 13:07Lex Fridman?

  In speech recognition, has there been much success in being able to create adversarial examples that fool the system?

• 13:16Speaker 01

  Yeah, actually, I guess the first work that I'm aware of is a paper called Hidden Voice Commands that came out in 2016, I believe, and they were able to show that they could make sounds that are not understandable by a human, but are recognized as the target phrase that the attacker wants the phone to recognize it as. Since then, things have gotten a little bit better on the attacker's side and worse on the defender's side. It's become possible to make sounds that sound like normal speech, but are actually interpreted as a different sentence than the human hears.

• 13:58Speaker 01

  The level of perceptibility of the adversarial perturbation is still kind of high. When you listen to the recording, it sounds like there's some noise in the background, just like rustling sounds. But those rustling sounds are actually the adversarial perturbation that makes the phone hear a completely different sentence.

• 14:15Lex Fridman?

  Yeah, that's so fascinating. Peter Norrig mentioned that you're writing the deep learning chapter for the fourth edition of the Artificial Intelligence: A Modern Approach book. So how do you even begin summarizing the field of deep learning in a chapter?

• 14:31Speaker 01

  Well, in my case, I waited like a year before I actually wrote anything. Even having written a full-length textbook before, it's still pretty intimidating to try to start writing just one chapter that covers everything.

• 14:48Speaker 01

  One thing that helped me make that plan was actually the experience of having written the full book before and then watching how the field changed after the book came out. I've realized there's a lot of topics that were maybe extraneous in the first book. And just seeing what stood the test of a few years of being published and what seems a little bit less important to have included now helped me pare down the topics I wanted to cover for the book. It's also really nice now that the field has stabilized to the point where some core ideas from the 1980s are still used today. When I first started studying machine learning, almost everything from the 1980s had been rejected, and now some of it has come back. So that stuff that's really stood the test of time is what I focused on putting into

• 15:32Speaker 01

  the book. There's also, I guess, two different philosophies about how you might write a book. One philosophy is you try to write a reference that covers everything. And the other philosophy is you try to provide a high-level summary that gives people the language to understand a field and tells them what the most important concepts are. The first deep learning book that I wrote with Yoshua and Aaron was somewhere between the two philosophies, that it's trying to be both a reference and an introductory guide. Writing this chapter for Russell and Norvig's book, I was able to focus more on just a concise introduction of the key concepts and the language you need to read about them more. In a lot of cases, I actually just wrote paragraphs that said, here's a rapidly evolving

• 16:16Speaker 01

  area that you should pay attention to. It's pointless to try to tell you what the latest and best version of a learn-to-learn model is.

• 16:27Lex Fridman?

  Right.

• 16:28Speaker 01

  I can point you to a paper that's recent right now, but there isn't a whole lot of reason to delve into exactly what's going on with the latest learning-to-learn approach or the latest module produced by a learning-to-learn algorithm. You should know that learning-to-learn is a thing and that it may very well be the source of the latest and greatest convolutional net or recurrent net module that you would want to use in your latest project. But there isn't a lot of point in trying to summarize exactly which architecture and which learning approach got to which level of performance.

• 17:00Lex Fridman?

  So you maybe focus more on the basics of the methodology, so from backpropagation to feed forward to recurrent neural networks, convolutional, that kind of thing?

• 17:12Speaker 01

  Yeah, yeah.

• 17:13Lex Fridman?

  So if I were to ask you, I remember I took algorithms and data structures algorithms course. I remember the professor asked, what is an algorithm? And yelled at everybody in a good way that nobody was answering it correctly. Everybody knew what the algorithm, it was a graduate course. Everybody knew what an algorithm was, but they weren't able to answer it well. So let me ask you, in that same spirit, what is deep learning?

• 17:41Speaker 01

  I would say, deep learning is any kind of machine learning that involves learning parameters of more than one consecutive step. So that, I mean, shallow learning is things where you learn a lot of operations that happen in parallel. You might have a system that makes multiple steps, like you might have hand-designed feature extractors, but really only one step is learned. Deep learning is anything where you have multiple operations in sequence. And that includes the things that are really popular today, like convolutional networks and recurrent networks. But it also includes some of the things that have died

• 18:22Speaker 01

  out, like Boltzmann machines, where we weren't using backpropagation. Today, I hear a lot of people define deep learning as gradient descent, applied to these differentiable functions. And I think that's a legitimate usage of the term. It's just different from the way that I use the term myself.

• 18:45Lex Fridman?

  So what's an example of deep learning that is not gradient descent and differentiable functions? In your, I mean, not specifically perhaps, but more even looking into the future. What's your thought about that space of approaches?

• 19:01Speaker 01

  Yeah. So I tend to think of machine learning algorithms as decomposed into really three different pieces. There's the model, which can be something like a neural net or a Boltzmann machine, or a recurrent model. And that basically just describes, how do you take data and how do you take parameters? And, you know, what function do you use to make a prediction, given the data and the parameters? Another piece of the learning algorithm is the optimization algorithm. Or not every algorithm can be really described in terms of optimization. But what's the algorithm for updating the parameters or updating whatever the state of the network is? And then the the last part is the the data set.

• 19:43Speaker 01

  Like, how do you actually represent the world as it comes into your machine learning system? So I think of deep learning as telling us something about what does the model look like? And basically, to qualify as deep, I say that it just has to have multiple layers. That can be multiple steps in a feed forward differentiable computation. That can be multiple layers in a graphical model. There's a lot of ways that you could satisfy me that something has multiple steps that are each parameterized separately. I think of gradient descent as being all about that other piece, the how do you actually update the parameters piece? So you could imagine having a deep model, like a convolutional net, and training it with something like evolution or a genetic algorithm.

• 20:28Speaker 01

  And I would say that still qualifies as deep learning. And then in terms of models that aren't necessarily differentiable, I guess Boltzmann machines are probably the main example of something where you can't really take a derivative and use that for the learning process. But you you can still argue that the model has many steps of processing that it applies when you run inference in the model.

• 20:52Lex Fridman?

  So it's the steps of processing that's key. So Jeff Hinton suggests that we need to throw away backprop backpropagation and start all over. What do you think about that? What could an alternative direction of training neural networks look like?

• 21:08Speaker 01

  I don't know that backpropagation is going to go away entirely. Most of the time, when we decide that a machine learning algorithm isn't on the critical path to research for improving AI, the algorithm doesn't die, it just becomes used for some specialized set of things. A lot of algorithms like logistic regression don't seem that exciting to AI researchers who are working on things like speech Recognition or autonomous cars today, but there's still a lot of use for logistic regression and things like analyzing really noisy data in medicine and finance, or making really rapid predictions in really time limited contexts. So I think backpropagation and gradient descent are around to stay,

• 21:51Speaker 01

  but they may not end up being everything that we need to get to real human level or superhuman AI.

• 21:59Lex Fridman?

  Are you optimistic about us discovering,

• 22:03Lex Fridman?

  you know, backpropagation has been around for a few decades. So are you optimistic about us as a community being able to discover something better?

• 22:13Speaker 01

  Yeah, I am. I think we likely will find something that works better. You could imagine things like having stacks of models where some of the lower level models predict parameters of the higher level models. And so at the top level, you're not learning in terms of literally calculating gradients, but just predicting how different values will perform. You can kind of see that already in some areas like Bayesian optimization, where you have a Gaussian process that predicts how well different parameter values will perform. We already use those kinds of algorithms for things like hyperparameter optimization. And in general, we know a lot of things other than backprop that work really well for specific problems. The main thing we haven't found is a way of taking one of these other non-backprop based

• 22:57Speaker 01

  algorithms and having it really advance the state of the art on an AI level problem.

• 23:03Lex Fridman?

  Right.

• 23:04Speaker 01

  But I wouldn't be surprised if eventually we find that some of these algorithms that, even the ones that already exist, not even necessarily new one, we might find some way of customizing one of these algorithms to do something really interesting at the level of cognition or the level of, I think one system that we really don't have working quite right yet is like short term memory. We have things like LSTMs, they're called long short term memory. They still don't do quite what a human does with short term memory. Like gradient descent to learn a specific. Specific fact has to do multiple steps on that fact.

• 23:46Speaker 01

  Like if I, I tell you, the meeting today is at 3pm. I don't need to say over and over again, it's at 3pm, it's at 3pm, it's at 3pm, it's at 3pm.

• 23:55Lex Fridman?

  Right.

• 23:56Speaker 01

  For you to do a gradient step on each one, you just hear it once and you remember it. There's been some work on things like self-attention and attention-like mechanisms, like the neural Turing machine that can write to memory cells and update themselves with facts like that right away. But I don't think we've really nailed it yet. And that's one area where I'd imagine that new optimization algorithms or different ways of applying existing optimization algorithms could give us a way of just lightning-fast updating the state of a machine learning system to contain a specific fact like that without needing to have it presented over and over and over again.

• 24:34Lex Fridman?

  So some of the success of symbolic systems in the 80s is they were able to assemble these kinds of facts better. But there's a lot of expert input required and it's very limited in that sense. Do you ever look back to that as something that we'll have to return to eventually? Sort of dust off the book from the shelf and think about how we build knowledge representation, knowledge.

• 25:00Speaker 01

  Like will we have to use graph searches?

• 25:02Lex Fridman?

  Graph searches, right.

• 25:02Speaker 01

  And like first-order logic and entailment and things like that.

• 25:05Lex Fridman?

  Kind of thing, yeah, exactly.

• 25:06Speaker 01

  In my particular line of work, which has mostly been machine learning security and, and also generative modeling, I haven't usually found myself moving in that direction. For generative models, I could see a little bit of, it could be useful if you had something like a, a differentiable knowledge base or some other kind of knowledge base where it's possible for some of our fuzzier machine learning algorithms to interact with a knowledge base.

• 25:34Lex Fridman?

  I mean, neural network is kind of like that. It's a differentiable knowledge base of sorts.

• 25:38Speaker 01

  Yeah.

• 25:39Lex Fridman?

  But.

• 25:40Speaker 01

  If, if we had a really easy way of giving feedback to machine learning models, that would clearly help a lot with, with generative models. And so you could imagine one way of getting there would be get a lot better at natural language processing. But another way of getting there would be take some kind of knowledge base and figure out a way for it to actually interact with a neural network.

• 26:01Lex Fridman?

  Being able to have a chat with a neural network.

• 26:03Speaker 01

  Yeah. So like one thing in generative models we see a lot today is you'll get things like faces that are not symmetrical, like people that have two eyes that are different colors. I mean, there are people with eyes that are different colors in real life, but not nearly as many of them as you tend to see in the machine learning generated data. So if you had either a knowledge base that could contain the fact, people's faces are generally approximately symmetric and eye color is especially likely to be the same on both sides. Being able to just inject that hint into the machine learning model without it having to discover that itself after studying a lot of data would be a really useful feature. I could see a lot of ways of getting there without bringing back some of the 1980s

• 26:48Speaker 01

  technology, but I also see some ways that you could imagine extending the 1980s technology to play nice with neural nets and have it help get there.

• 26:56Lex Fridman?

  Awesome. So you talked about the story of you coming up with the idea of GANs at a bar with some friends. You were arguing that this, you know, GANs would work, generative adversarial networks, and the others didn't think so. Then you went home at midnight, coated up, and it worked. So if I was a friend of yours at the bar, I would also have doubts. It's a really nice idea, but I'm very skeptical that it would work. What was the basis of their skepticism? What was the basis of your intuition why it should work?

• 27:30Speaker 01

  I don't want to be someone who goes around promoting alcohol for the purposes of science, but in this case, I do actually think that drinking helped a little bit. When your inhibitions are lowered, you're more willing to try out things that you wouldn't try out otherwise. So I have noticed in general that I'm less prone to shooting down some of my own ideas when I have had a little bit to drink. I think if I had had that idea at lunchtime, I probably would have thought it. It's hard enough to train one neural net. You can't train a second neural net in the inner loop of the outer neural net. That was basically my friend's objection was that trying to train two neural nets at the same time would be too hard.

• 28:11Lex Fridman?

  So it was more about the training process unless... So my skepticism would be, you know, I'm sure you could train it, but the thing it would converge to would not be able to generate anything reasonable, any kind of reasonable realism.

• 28:25Speaker 01

  Yeah, so part of what all of us were thinking about when we had this conversation was deep Boltzmann machines, which a lot of us in the lab, including me, were a big fan of deep Boltzmann machines at the time. They involved two separate processes running at the same time. One of them is called the positive phase, where you load data into the model and tell the model to make the data more likely. The other one is called the negative phase, where you draw samples from the model and tell the model to make those samples less likely. In a deep Boltzmann machine, it's not trivial to generate a sample. You have to actually run an iterative process that gets better and better, samples coming closer and closer to the distribution the model represents.

• 29:10Speaker 01

  So during the training process, you're always running these two systems at the same time, one that's updating the parameters of the model and another one that's trying to generate samples from the model. And they worked really well on things like MNIST, but a lot of us in the lab, including me, had tried to get deep Boltzmann machines to scale past MNIST to things like generating color photos, and we just couldn't get the two processes to stay synchronized. So when I had the idea for GANs, a lot of people thought that the discriminator would have more or less the same problem as the negative phase in the Boltzmann machine, that trying to train the discriminator in the inner loop, you just couldn't get it to keep up with the generator in the outer loop, and that would prevent it from converging to anything useful.

• 29:52Lex Fridman?

  Yeah, I share that intuition.

• 29:54Speaker 01

  Yeah.

• 29:56Lex Fridman?

  But turns out to not be the case.

• 29:58Speaker 01

  A lot of the time with machine learning algorithms, it's really hard to predict ahead of time how well they'll actually perform. You have to just run the experiment and see what happens. And I would say I still today don't have like one factor I can put my finger on and say, this is why GANs worked for photo generation and deep Boltzmann machines don't.

• 30:19Speaker 01

  There are a lot of theory papers showing that under some theoretical settings, the GAN algorithm does actually converge. But those settings are restricted enough that they don't necessarily explain the whole picture in terms of all the results that we see in practice.

• 30:37Lex Fridman?

  So taking a step back, can you, in the same way as we talked about deep learning, can you tell me what generative adversarial networks are?

• 30:46Speaker 01

  Yeah, so generative adversarial networks are a particular kind of generative model. A generative model is a machine learning model that can train on some set of data, like, say, you have a collection of photos of cats and you want to generate more photos of cats, or you want to estimate a probability distribution over cats, so you can ask how likely it is that some new image is a photo of a cat. GANs are one way of doing this. Some generative models are good at creating new data. Other generative models are good at estimating that density function and telling you how likely particular pieces of data are to come from the same distribution as the training data. GANs are more focused on generating samples

• 31:29Speaker 01

  rather than estimating the density function. There are some kinds of GANs, like FlowGAN, that can do both, but mostly GANs are about generating samples, generating new photos of cats that look realistic. And they do that completely from scratch. It's analogous to human imagination. When a GAN creates a new image of a cat, it's using a neural network to produce a cat that has not existed before. It isn't doing something like compositing photos together. You're not, you're not literally taking the eye off of one cat and the ear off of another cat. It's, it's more of this digestive process where the, the neural net trains on a lot of data and comes up with some representation of

• 32:13Speaker 01

  the probability distribution and generates entirely new cats.

• 32:17Speaker 01

  There are a lot of different ways of building a generative model. What's specific to GANs is that we have a two-player game in the game theoretic sense. And as the players in this game compete, one of them becomes able to generate realistic data. The first player is called the generator. It produces output data, such as just images, for example. And at the start of the learning process, it'll just produce completely random images. The other player is called the discriminator. The discriminator takes images as input and guesses whether they're real or fake. You train it both on real data, so photos that come from your training set, actual photos of cats. And you train it to say that those are real. You also train it on images that come from the generator network.

• 33:01Speaker 01

  And you train it to say that those are fake. As the two players compete in this game, the discriminator tries to become better at recognizing whether images are real or fake. And the generator becomes better at fooling the discriminator into thinking that its outputs are real. And you can analyze this through the language of game theory and find that there's a Nash equilibrium where the generator has captured the correct probability distribution. So in the cat example, it makes perfectly realistic cat photos. And the discriminator is unable to do better than random guessing. Because all the samples coming from both the data and the generator look equally likely to have come from either source.

• 33:43Lex Fridman?

  So do you ever sit back and does it just blow your mind that this thing works? So from very, so it's able to estimate that density function enough to generate realistic images? I mean, yeah, do you ever sit back and think, how does this even, why? This is quite incredible, especially where GANs have gone in terms of realism.

• 34:06Speaker 01

  Yeah, and not just to flatter my own work, but generative models, all of them have this property that if they really did what we ask them to do, they would do nothing but memorize the training data.

• 34:18Lex Fridman?

  Right.

• 34:20Speaker 01

  Models that are based on maximizing the likelihood. The way that you obtain the maximum likelihood for a specific training set is you assign all of your probability mass to the training examples and nowhere else. For GANs, the game is played using a training set. So the way that you become unbeatable in the game is you literally memorize training examples.

• 34:42Speaker 01

  One of my former interns wrote a paper. His name is Vaishnav Nagarajan. And he showed that it's actually hard for the generator to memorize the training data. Hard in a statistical learning theory sense that you can actually create reasons for why it would require quite a lot of learning steps and a lot of observations of different latent variables before you could memorize the training data. That still doesn't really explain why, when you produce samples that are new, why do you get compelling images rather than, you know, just garbage that's different from the training set? And I don't think we really have a good answer for that. Especially if you think about how many possible images are out there

• 35:27Speaker 01

  and how few images the generative model sees during training. It seems just unreasonable that generative models create new images as well as they do, especially considering that we're basically training them to memorize rather than generalize. I think part of the answer is there's a paper called Deep Image Prior, where they show that you can take a convolutional net and you don't even need to learn the parameters of it at all. You just use the model architecture. And it's already useful for things like in-painting images. I think that shows us that the convolutional network architecture captures something really important about the structure of images. And we don't need to actually use learning to capture all the information coming out of the convolutional net.

• 36:12Speaker 01

  That would imply that it would be much harder to make generative models in other domains. So far, we're able to make reasonable speech models and things like that. But to be honest, we haven't actually explored a whole lot of different data sets all that much. We don't, for example, see a lot of deep learning models of like biology data sets, where you have lots of microarrays measuring the amount of different enzymes and things like that. So we may find that some of the progress that we've seen for images and speech turns out to really rely heavily on the model architecture. And we were able to do what we did for vision by trying to reverse engineer the human visual system.

• 36:54Speaker 01

  And maybe it'll turn out that we can't just use that same trick for arbitrary kinds of data.

• 37:00Lex Fridman?

  Right, so there's aspect to the human vision system, the hardware of it that makes it, without learning, without cognition, just makes it really effective at detecting the patterns we see in the visual world.

• 37:11Speaker 01

  Yeah.

• 37:12Lex Fridman?

  That's, yeah, that's really interesting. What, in a big, quick overview, in your view, what types of GANs are there and what other generative models besides GANs are there?

• 37:26Speaker 01

  Yeah, so it's maybe a little bit easier to start with what kinds of generative models are there other than GANs? So, most generative models are likelihood-based, where to train them, you have a model that tells you how, how much probability it assigns to a particular example, and you just maximize the probability assigned to all the training examples. It turns out that it's hard to design a model that can create really complicated images or really complicated audio waveforms, and still have it be possible to estimate the likelihood function from a computational point of view.

• 38:09Speaker 01

  Most interesting models that you would just write down intuitively, it turns out that it's almost impossible to calculate the amount of probability they assign to a particular point. So, there's a few different schools of generative models in the likelihood family. One approach is to very carefully design the model so that it is computationally tractable to measure the density it assigns to a particular point. So, there are things like autoregressive models, like Pixel CNN. Those basically break down the probability distribution into a product over every single feature. So, for an image, you estimate the probability of each pixel, given all of the pixels that came before it.

• 38:53Speaker 01

  There's tricks where if you want to measure the density function, you can actually calculate the density for all these pixels, more or less in parallel. Generating the image still tends to require you to go one pixel at a time, and that can be very slow. But there are, again, tricks for doing this in a hierarchical pattern where you can keep the runtime under control.

• 39:12Lex Fridman?

  Are the quality of the images it generates, putting runtime aside, pretty good?

• 39:19Speaker 01

  They're reasonable, yeah. I would say a lot of the best results are from GANs these days, but it can be hard to tell how much of that is based on who's studying which type of algorithm, if that makes sense.

• 39:34Lex Fridman?

  The amount of effort invested in a particular.

• 39:36Speaker 01

  Yeah, or like the kind of expertise. So a lot of people who've traditionally been excited about graphics or art and things like that have gotten interested in GANs. And to some extent, it's hard to tell, are GANs doing better because they have a lot of graphics and art experts behind them? Or are GANs doing better because they're more computationally efficient? Or are GANs doing better because they prioritize the realism of samples over the accuracy of the density function? I think, I think all of those are potentially valid explanations and it's, it's hard to tell.

• 40:08Lex Fridman?

  So can you give a brief history of GANs from 2014, where you paper 13?

• 40:16Speaker 01

  Yeah, so a few highlights. In the first paper, we just showed that GANs basically work. If you look back at the samples we had now, they looked terrible. On the CIFAR-10 dataset, you can't even recognize objects in them.

• 40:29Lex Fridman?

  Your papers, you use CIFAR-10?

• 40:32Speaker 01

  We use MNIST, which is little handwritten digits. We used the Toronto Face Database, which is small grayscale photos of faces. We did have recognizable faces. My colleague Bing Xu put together the first GAN face model for that paper. We also had the CIFAR-10 dataset, which is things like very small 32 by 32 pixels of, of cars and cats and dogs. For that, we didn't get recognizable objects, but all the deep learning people back then were really used to looking at these failed samples and kind of reading them like tea leaves. And people who are used to reading the tea leaves recognize that our tea leaves at least look different. Maybe not necessarily better, but there was something unusual about them.

• 41:18Speaker 01

  And that got a lot of us excited. One of the next really big steps was LapGAN by Emily Denton and Sumit Chintala at Facebook AI Research, where they actually got really good, high-resolution photos working with GANs for the first time. They had a complicated system where they generated the image starting at low res and then scaling up to high res, but they were able to get it to work. And then in 2015, I believe, later that same year, Alec Radford and Sumit Chintala and Luke Metz published the DCGAN paper, which it stands for Deep Convolutional GAN.

• 41:59Speaker 01

  It's kind of a non-unique name because these days, basically, all GANs and even some before that were deep and convolutional, but they just kind of picked a name for a really great recipe where they were able to actually, using only one model instead of a multi-step process, actually generate realistic images of faces and things like that. That was sort of like the beginning of the Cambrian explosion of GANs. Like, you know, once you had animals that had a backbone, you suddenly got lots of different versions of, you know, like fish and four-legged animals and things like that. So DCGAN became kind of the backbone for many different models that came out.

• 42:36Lex Fridman?

  Used as a baseline even still.

• 42:38Speaker 01

  Yeah, yeah. And so from there, I would say some interesting things we've seen are, there's a lot you can say about how just the quality of standard image generation GANs has increased, but what's also maybe more interesting on an intellectual level is how the things you can use GANs for has also changed. One thing is that you can use them to learn classifiers without having to have class labels for every example in your training set. So that's called semi-supervised learning. My colleague at OpenAI, Tim Solomons, who's at Brain now, wrote a paper called "Improved Techniques for Training GANs." I'm a co-author in this paper, but I can't claim any credit for this particular part.

• 43:20Speaker 01

  One thing he showed in the paper is that you can take the GAN discriminator and use it as a classifier that actually Really tells you, you know, this image is a cat, this image is a dog, this image is a car, this image is a truck. And so on, not just to say whether the image is real or fake, but if it is real, to say specifically what kind of object it is. And he found that you can train these classifiers with far fewer labeled examples than traditional classifiers.

• 43:47Lex Fridman?

  So if you supervise based on also not just your discrimination ability, but your ability to classify, you're going to do much, you're going to converge much faster to being effective at being a discriminator.

• 44:00Speaker 01

  Yeah. So for, for example, for the MNIST dataset, you want to look at an image of a handwritten digit and say whether it's a zero, a one, or a two, and so on. To get down to less than 1% accuracy required around 60,000 examples until maybe about 2014 or so. In 2016, with this semi-supervised GAN project, Tim was able to get below 1% error using only 100 labeled examples. So that was about a 600x decrease in the amount of labels that he needed. He's still using more images than that, but he doesn't need to have each of them labeled as, you know, this one's a one, this one's a two, this one's a zero, and so on.

• 44:44Lex Fridman?

  Then to be able to, for GANs to be able to generate recognizable objects, so objects from a particular class, you still need labeled data because you need to know what it means to be a particular class cat, dog. How do you think we can move away from that?

• 45:01Speaker 01

  Yeah, some researchers at BrainZurich actually just released a really great paper on semi-supervised GANs where their, their goal isn't to classify, it's to make recognizable objects despite not having a lot of labeled data. They were working off of DeepMind's Big GAN project, and they showed that they can match the performance of Big GAN using only 10%, I believe, of the, of the labels. Big GAN was trained on the ImageNet dataset, which is about 1.2 million images, and had all of them labeled. This latest project from BrainZurich shows that they're able to get away with only having about 10% of the, of the images labeled.

• 45:42Speaker 01

  And they do that essentially using a clustering algorithm, where the discriminator learns to assign the objects to groups. And then this understanding that objects can be grouped into similar types helps it to form more realistic ideas of what should be appearing in the image, because it knows that every image it creates has to come from one of these archetypal groups, rather than just being some arbitrary image. If you train a GAN with no class labels, you tend to get things that look sort of like grass, or water, or brick, or dirt, but without necessarily a lot going on in them. And I think that's partly because if you look at a large ImageNet image,

• 46:25Speaker 01

  the object doesn't necessarily occupy the whole image. And so you learn to create realistic sets of pixels, but you don't necessarily learn that the object is the star of the show, and you want it to be in every image you make.

• 46:39Lex Fridman?

  Yeah, I've heard you talk about the horse, the zebra cycle GAN mapping, and how it turns out, again, thought provoking, that horses are usually on grass and zebras are usually on drier terrain. So when you're doing that kind of generation, you're going to end up generating greener horses, or whatever. So those are connected together. It's not just...

• 47:03Speaker 01

  Yeah, yeah.

• 47:04Lex Fridman?

  You're not able to segment, be able to generate in a segmented way. So are there other types of games you come across in your mind that neural networks can play with each other to be able to solve problems?

• 47:22Speaker 01

  Yeah, the one that I spend most of my time on is in security. You can model most interactions as a game, where there's attackers trying to break your system, and you're the defender trying to build a resilient system. There's also domain adversarial learning, which is an approach to domain adaptation that looks really a lot like GANs. The authors had the idea before the GAN paper came out, their paper came out a little bit later, and they... They're very nice and cited the GAN paper, but I know that they actually had the idea before it came out. Domain adaptation is when you want to train a machine learning model in one one setting called a domain and then deploy it in another domain later.

• 48:07Speaker 01

  And you would like it to perform well in the new domain, even though the new domain is different from how it was trained. So for example, you might want to train on a really clean image data set like ImageNet, but then deploy on users' phones where the user is taking pictures in the dark and pictures while moving quickly and just pictures that aren't really centered or composed all that well.

• 48:30Speaker 01

  When you take a normal machine learning model, it often degrades really badly when you move to the new domain because it looks so different from what the model was trained on. Domain adaptation algorithms try to smooth out that gap. And the domain adversarial approach is based on training a feature extractor where the features have the same statistics regardless of which domain you extracted them on. So in the domain adversarial game, you have one player that's a feature extractor and another player that's a domain recognizer. The domain recognizer wants to look at the output of the feature extractor and guess which of the two domains the features came from. So it's a lot like the real versus fake discriminator in GANs. And then the feature extractor, you can think of as loosely analogous to the generator in GANs,

• 49:15Speaker 01

  except what it's trying to do here is both fool the domain recognizer into not knowing which domain the data came from and also extract features that are good for classification. So at the end of the day, you can, in the cases where it works out, you can actually get features that work about the same in both domains. Sometimes this has a drawback where in order to make things work the same in both domains, it just gets worse at the first one. But there are a lot of cases where it actually works out well on both.

• 49:47Lex Fridman?

  So do you think of GANs being useful in the context of data augmentation?

• 49:52Speaker 01

  Yeah, one thing you could hope for with GANs is you could imagine I've got a limited training set and I'd like to make more training data to train something else like a classifier. You could train the GAN on the training set and then create more data and then maybe the classifier would perform better on the test set after training on this bigger GAN-generated dataset. So that's the simplest version of something you might hope would work. I've never heard of that particular approach working, but I think there's some closely-related things that I think could work in the future and some that actually already have worked. So if we think a little bit about what we'd be hoping for if we use the GAN to make more

• 50:34Speaker 01

  training data, we're hoping that the GAN will generalize to new examples better than the classifier would have generalized if it was trained on the same data. And I don't know of any reason to believe that the GAN would generalize better than the classifier would, but what we might hope for is that the GAN could generalize differently from a specific classifier. So one thing I think is worth trying that I haven't personally tried, but someone could try is what if you trained a whole lot of different generative models on the same training set, create samples from all of them, and then train a classifier on that? Because each of the generative models might generalize in a slightly different way, they might capture many different axes of variation that one individual model wouldn't. And then the classifier can capture all of those ideas by training on all of their

• 51:19Speaker 01

  data. So it'd be a little bit like making an ensemble of classifiers.

• 51:23Lex Fridman?

  An ensemble of GANs, in a way.

• 51:26Speaker 01

  Yeah, I think that could generalize better. The other thing that GANs are really good for is not necessarily generating new data that's exactly like what you already have, but by generating new data that has different properties from the data you already had. One thing that you can do is you can create differentially private data. So suppose that you have something like medical records and you don't want to train a classifier on the medical records and then publish the classifier because someone might be able to reverse engineer some of the medical records you trained on. There's a paper from Casey Green's lab that shows how you can train GANs. Using differential privacy and then the samples on the GAN still have the same differential

• 52:07Speaker 01

  privacy guarantees as the parameters of the GAN. So you can make fake patient data for other researchers to use and they can do almost anything they want with that data because it doesn't come from real people and the differential privacy mechanism gives you clear guarantees on how much the original people's data has been protected.

• 52:27Lex Fridman?

  That's really interesting. Actually, I haven't heard you talk about that before. In terms of fairness, I've seen from AAAI your talk, how can adversarial machine learning help models be more fair with respect to sensitive variables?

• 52:42Speaker 01

  Yeah, so there's a paper from Amos Storki's lab about how to learn machine learning models that are incapable of using specific variables. So say, for example, you wanted to make predictions that are not affected by gender. It isn't enough to just leave gender out of the input to the model. You can often infer gender from a lot of other characteristics, like say that you have the person's name, but you're not told their gender. Well, if their name is Ian, they're kind of obviously a man. So what you'd like to do is make a machine learning model that can still take in a lot of different attributes and make a really accurate informed prediction, but be confident that it isn't reverse engineering gender or another sensitive

• 53:23Speaker 01

  variable internally. You can do that using something very similar to the domain adversarial approach where you have one player that's a feature extractor and another player that's a feature analyzer. And you want to make sure that the feature analyzer is not able to guess the value of the sensitive variable that you're trying to keep private.

• 53:43Lex Fridman?

  Right, that's, yeah, I love this approach. So yeah, with the with the feature, you're not able to infer, right, the sensitive variables.

• 53:53Speaker 01

  Yeah.

• 53:53Lex Fridman?

  Brilliant. It's quite brilliant and simple, actually.

• 53:56Speaker 01

  Another way I think that GANs in particular could be used for fairness would be to make something like a cycle GAN where you can take data from one domain and convert it into another. We've seen cycle GAN turning horses into zebras. We've seen other unsupervised GANs made by Ming Yu Lu doing things like turning day photos into night photos. I think for fairness, you could imagine taking records for people in one group and transforming them into analogous people in another group and testing to see if they're treated equitably across those two groups. There's a lot of things that'd be hard to get right to make sure that the conversion process itself is fair. And I don't think it's anywhere near

• 54:41Speaker 01

  something that we could actually use yet. But if you could design that conversion process very carefully, it might give you a way of doing audits where you say, what if we took people from this group, converted them into equivalent people in another group, does the system actually treat them how it ought to?

• 54:57Lex Fridman?

  That's also really interesting. You know, in popular press and in general, in our imagination, you think, well, GANs are able to generate data and you start to think about deep fakes or being able to sort of maliciously generate data that fakes the identity of other people. Is this something of a concern to you? Is this something, if you look 10, 20 years into the future, is that something that pops up in your work and the work of the community that's working on generated models?

• 55:30Speaker 01

  I'm a lot less concerned about 20 years from now than the next few years. I think there will be a kind of bumpy cultural transition as people encounter this idea that there can be very realistic videos and audio that aren't real. I think 20 years from now, people will mostly understand that you shouldn't believe something is real just because you saw a video of it. People will expect to see that it's been cryptographically signed or have some other mechanism to make them believe that the content is real. There's already people working on this. Like, there's a startup called TruePick that provides a lot of mechanisms for authenticating that an image is real. They're maybe not quite

• 56:11Speaker 01

  up to having a state actor try to evade their verification techniques, but it's something that people are already working on and I think will get right eventually.

• 56:21Lex Fridman?

  So you think authentication will will eventually win out. So being able to authenticate that this is real and this is not.

• 56:29Speaker 01

  Yeah.

• 56:30Lex Fridman?

  As opposed to GANs just getting better and better, or generative models being able to get better and better to where the nature of what is real.

• 56:38Speaker 01

  I don't think we'll ever be able to look at the pixels of a photo and tell you for sure that it's real or not real. And I think it would actually be somewhat dangerous to rely on that approach too much. If you make a really good fake detector and then someone's able to fool your fake detector and your fake detector says this image is not fake, then it's even more credible than if you've never made a fake detector in the first place. What I do think we'll get to is systems that we can kind of use behind the scenes for to make estimates of what's going on and maybe not like use them in court for a definitive analysis. I also think we will likely get better authentication systems where, you

• 57:22Speaker 01

  know, imagine that every phone cryptographically signs everything that comes out of it. You wouldn't be able to conclusively tell that an image was real, but you would be able to tell somebody who knew the appropriate private key for this phone was actually able to sign this image and upload it to this server at this timestamp.

• 57:44Lex Fridman?

  Right.

• 57:46Speaker 01

  So you could imagine, maybe you make phones that have the private keys hardware embedded in them. If like a state security agency really wants to infiltrate the company, they could probably, you know, plant a private key of their choice or break open the chip and learn the private key or something like that. But it would make it a lot harder for an adversary with fewer resources to fake things.

• 58:08Lex Fridman?

  For most of us, it would be okay. Okay. So you mentioned the beer and the bar and the new ideas. You were able to implement this or come up with this new idea pretty quickly and implement it pretty quickly. Do you think there are still many such groundbreaking ideas in deep learning that could be developed so quickly?

• 58:28Speaker 01

  Yeah, I do think that there are a lot of ideas that can be developed really quickly.

• 58:33Speaker 01

  GANs were probably a little bit of an outlier on the whole like one hour time scale.

• 58:36Lex Fridman?

  Right.

• 58:37Speaker 01

  But just in terms of like low resource ideas where you do something really different on the algorithm scale and get a big payback.

• 58:47Speaker 01

  I think it's not as likely that you'll see that in terms of things like core machine learning technologies, like a better classifier, or a better reinforcement learning algorithm, or a better generative model.

• 58:58Speaker 01

  If I had the GAN idea today, it would be a lot harder to prove that it was useful than it was back in 2014, because I would need to get it running on something like ImageNet or Celeb A at high resolution. You know, those take a while to train. You couldn't, you couldn't train it in an hour and know that it was something really new and exciting. Back in 2014, training on MNIST was enough. But there are other areas of machine learning where I think a new idea could actually be developed really quickly with low resources.

• 59:30Lex Fridman?

  What's your intuition about what areas of machine learning are ripe for this?

• 59:34Speaker 01

  Yeah, so I think fairness and interpretability are areas where we just really don't have any idea how anything should be done yet. Like for interpretability, I don't think we even have the right definitions. And even just defining a really useful concept, you don't even need to run any experiments, could have a huge impact on the field. We've seen that, for example, in differential privacy, that Cynthia Dwork and her collaborators made this technical definition of privacy, where before, a lot of things were really mushy, and then with that definition, you could actually design randomized algorithms for accessing databases and guarantee that they preserved individual people's privacy in a, in like a mathematical quantitative sense.

• 1:00:20Speaker 01

  Right now, we all talk a lot about how interpretable different machine learning algorithms are, but it's really just people's opinion, and everybody probably has a different idea of what interpretability means in their head. If we could define some concept related to interpretability that's actually measurable, that would be a huge leap forward, even without a new algorithm that increases that quantity. And also, once, once we had the definition of differential privacy, it was fast to get the algorithms that guaranteed it. So you could imagine, once we have definitions of good concepts and interpretability, we might be able to provide the algorithms that have the interpretability guarantees quickly, too.

• 1:00:59Lex Fridman?

  What do you think it takes to build a system with human-level intelligence as we quickly venture into the philosophical? So, artificial general intelligence? What do you think it takes?

• 1:01:12Speaker 01

  Yeah, I think that it definitely takes better environments than we currently have for training agents, that we want them to have a really wide diversity of experiences. I also think it's going to take really a lot of computation. It's hard to imagine exactly how much.

• 1:01:30Lex Fridman?

  So you're optimistic about simulation? Simulating a variety of environments as the path forward?

• 1:01:36Speaker 01

  I think it's a necessary ingredient. Yeah. I don't think that we're going to get to artificial general intelligence by training on fixed data sets or by thinking really hard about the problem. I think that the agent really needs to interact and have a variety of experiences within the same lifespan. And today, we have many different models that can each do one thing, and we tend to train them on one data set or one RL environment. Sometimes there are actually papers about getting one set of parameters to perform well in many different RL environments. But we don't really have anything like an agent that goes seamlessly from one type

• 1:02:18Speaker 01

  of experience to another and really integrates all the different things that it does over the course of its life. When we do see multi-agent environments, they tend to be, there's many multi-environment agents, they tend to be similar environments. Like, all of them are playing like an action-based video game. We don't really have an agent that goes from, you know, playing a video game to like, reading the Wall Street Journal, to predicting how effective a molecule will be as a drug or something like that.

• 1:02:50Lex Fridman?

  What do you think is a good test for intelligence in your view? There's been a lot of benchmarks, started with Alan Turing, natural conversation being a good benchmark for intelligence. What would Ian Goodfellow sit back and be really damn impressed if a system was able to accomplish?

• 1:03:12Speaker 01

  Something that doesn't take a lot of glue from human engineers. So imagine that instead of having to go to the CIFAR website and download CIFAR 10 and then write a Python script to parse it and all that, you could just point an agent at the CIFAR 10 problem and it downloads and extracts the data and trains a model and starts giving you predictions. I feel like something that doesn't need to have every step of the pipeline assembled for it. It definitely understands what it's doing.

• 1:03:47Lex Fridman?

  Is AutoML moving into that direction? Or are you thinking way even bigger?

• 1:03:51Speaker 01

  AutoML has mostly been moving toward, once we've built all the glue, can the machine learning system design the architecture really well? And so I'm more of saying like, if something knows how to pre-process the data so that it successfully accomplishes the task, then it would be very hard to argue that it doesn't truly understand the task in some fundamental sense. And I don't necessarily know that that's like the philosophical definition of intelligence, but it's something that would be really cool to build, that would be really useful and would impress me and would convince me that we've made a step forward in real AI.

• 1:04:26Lex Fridman?

  So you give it like the URL for Wikipedia and then next day expect it to be able to solve CIFAR 10?

• 1:04:35Speaker 01

  Or like you type in a paragraph explaining what you want it to do and it figures out what web searches it should run and downloads all the necessary ingredients.

• 1:04:45Lex Fridman?

  So you have a very clear, calm way of speaking, no ums, easy to edit. I've seen comments for both you and I have been identified as both potentially being robots. If you have to prove to the world that you are indeed human, how would you do it?

• 1:05:08Speaker 01

  I can understand thinking that I'm a robot.

• 1:05:12Lex Fridman?

  It's the flip side of the Turing test, I think.

• 1:05:15Speaker 01

  Yeah, yeah, the prove your human test.

• 1:05:19Lex Fridman?

  Intellectually, so you have to, is there something that's truly unique in your mind? Does it go back to just natural language again? Just being able to talk.

• 1:05:30Speaker 01

  So proving that I'm not a robot with today's technology. Yeah, that's pretty straightforward. Like, my conversation today hasn't veered off into, you know, talking about the stock market or something because it isn't my training data. But I guess more generally, trying to prove that something is real from the content alone is incredibly hard. That's one of the main things I've gotten out of my GAN research, that you can simulate almost anything. And so you have to really step back to a separate channel to prove that something is real. So like, I guess I should have had myself stamped on a blockchain when I was born or something. But I didn't do that. So according to my own research methodology, there's just no way to know at this point.

• 1:06:10Lex Fridman?

  So what last question problem stands out for you that you're really excited about challenging in the near future?

• 1:06:17Speaker 01

  So I think resistance to adversarial examples. Figuring out how to make machine learning secure against an adversary who wants to interfere it and control it that is one of the most important things researchers today could solve.

• 1:06:29Lex Fridman?

  In all domains, image, language, driving.

• 1:06:34Speaker 01

  I guess I'm most concerned about domains we haven't really encountered yet. Like, like, imagine 20 years from now when we're using advanced AIs to do things we haven't even thought of yet. Like, if you ask people, what are the important problems in security of phones in like 2002? I don't think we would have anticipated that we're using them for, you know, nearly as many things as we're using them for today. I think it's going to be like that with AI, that you can kind of try to speculate about where it's going, but really the business opportunities that end up taking off would be hard to predict ahead of time. Well, you can predict ahead of time is that almost anything you can do with machine learning,

• 1:07:15Speaker 01

  you would like to make sure that people can't get it to do what they want rather than what you want, just by showing it a funny QR code or a funny input pattern.

• 1:07:25Lex Fridman?

  And you think that the set of methodology to do that can be bigger than any one domain?

• 1:07:30Speaker 01

  I think so, yeah, yeah. Like, one methodology that I think is, not a specific methodology, but like a category of solutions that I'm excited about today is making dynamic models that change every time they make a prediction. So right now, we tend to train models, and then after they're trained, we freeze them, and we just use the same rule to classify everything that comes in from then on. That's really a sitting duck from a security point of view. If you always output the same answer for the same input, then people can just run inputs through until they find a mistake that benefits them, and then they use the same mistake over and over and over again. I think having a model that updates its predictions

• 1:08:13Speaker 01

  so that it's harder to predict what you're going to get will make it harder for an adversary to really take control of the system and make it do what they want it to do.

• 1:08:23Lex Fridman?

  Yeah, models that maintain a bit of a sense of mystery about them, because they always keep changing.

• 1:08:29Speaker 01

  Yeah.

• 1:08:29Lex Fridman?

  Ian, thanks so much for talking today. It was awesome.

• 1:08:32Speaker 01

  Thank you for coming in. It's great to see you.