Lex Fridman Podcast - #95 – Dawn Song: Adversarial Machine Learning and Computer Security

Episode Date: May 13, 2020

Dawn Song is a professor of computer science at UC Berkeley with research interests in security, most recently with a focus on the intersection between computer security and machine learning. Support this podcast by signing up with these sponsors: – Cash App – use code “LexPodcast” and download: – Cash App (App Store): https://apple.co/2sPrUHe – Cash App (Google Play): https://bit.ly/2MlvP5w EPISODE LINKS: Dawn's Twitter: https://twitter.com/dawnsongtweets Dawn's Website: https://people.eecs.berkeley.edu/~dawnsong/ Oasis Labs: https://www.oasislabs.com This conversation is part of the Artificial Intelligence podcast. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter, LinkedIn, Facebook, Medium, or YouTube where you can watch the video versions of these conversations. If you enjoy the podcast, please rate it 5 stars on Apple Podcasts, follow on Spotify, or support it on Patreon. Here's the outline of the episode. On some podcast players you should be able to click the timestamp to jump to that time. OUTLINE: 00:00 - Introduction 01:53 - Will software always have security vulnerabilities? 09:06 - Human are the weakest link in security 16:50 - Adversarial machine learning 51:27 - Adversarial attacks on Tesla Autopilot and self-driving cars 57:33 - Privacy attacks 1:05:47 - Ownership of data 1:22:13 - Blockchain and cryptocurrency 1:32:13 - Program synthesis 1:44:57 - A journey from physics to computer science 1:56:03 - US and China 1:58:19 - Transformative moment 2:00:02 - Meaning of life

The following is a conversation with Don Song, a professor of computer science at UC Berkeley with research interests and computer security, most recently with a focus on the intersection between security and machine learning. This conversation was recorded before the outbreak of the pandemic for everyone feeling the medical, psychological, and financial burden of this crisis. I'm sending love your way. Stay strong, we're in this together, we'll beat this thing. This is the Artificial Intelligence Podcast. If you enjoy it, subscribe

on YouTube, review it with 5 stars on Apple Podcast, support it on Patreon, or simply connect with me on Twitter, Alex Friedman spelled F-R-I-D-M-A-N. As usual, I'll do a few minutes of ads now and never any ads in the middle that can break the flow of the conversation. I hope that works for you and doesn't hurt the listening experience. This show is presented by CashApp, the number one finance app in the app store. When you get it, use CoLux podcast. CashApp lets you send money to friends by Bitcoin and invests in the stock market with as little as $1. Since CashApp does fractional share trading, let me mention that the order execution algorithm

that works behind the scenes to create the abstraction of fractional orders is an algorithmic marvel. So big props to the CashApp engineers for solving a hard problem that in the end provides an easy interface that takes a step up to the next layer of abstraction over the stock market, making trading more accessible for new investors and diversification much easier. So again, if you get cash out from the App Store Google Play and use the code Lex Podcast, you get $10 and cash out will also donate $10 to first. The organization that is helping to advance robotics and STEM education for young people around the world.

And now here's my conversation with Dawn Song. Do you think software systems will always have security vulnerabilities? Let's start at the broad, almost philosophical level. That's a very good question. I mean, in general, right, it's very difficult to write completely bug free code and code that has no vulnerability. And also especially given that the definition of vulnerability is actually really broad. It's in type of attacks essentially on a code that you can call that caused by vulnerabilities. And the nature of attacks is always changing as well. Like new ones are coming up. So for example, in the past we talked about memory safety, type of vulnerabilities, where

essential attackers can exploit the software and take over control of how the code runs and then can launch attacks that way. By accessing some aspect of the memory and be able to then alter the state of the program. Exactly. So for example, in the example of a buffer overflow, then the attacker essentially actually causes essentially unintended changes in the state of the program. And then, for example, can then take over control flow of the program and write the program to execute codes that actually the program didn't intend. So the attack can be a remote attack. So the attack here, for example, can send in malicious inputs to the program that just causes the program to completely then be compromised and then end up doing something that's under the program,

under the attack control and intention. But there's just one form of attacks and there are other forms of attacks. For example, there are these side channels where attackers can try to learn from, even just observing the outputs from the behaviors of the program, try to infer certain secrets of the program.

So essentially, the form of attacks is very, very, it's very broad spectrum. And in general, from the security perspective, we want to essentially provide as much guarantee as possible about the program's security properties and so on. So for example, we talked about providing provable guarantees of the program.

So for example, there are ways we can use program analysis and form verification techniques to prove that a piece of code has no memory safety vulnerabilities. What does that look like? What is that proof? Is that just a dream for that's applicable to small case examples?

Or is that possible to do for real world systems? So actually, I mean, today, I actually call it we are entering the area of formally verified systems. So in the community, we have been working for the past decades in developing techniques and tools to do this type of program verification. And we have dedicated teams that have dedicated their years or sometimes even decades of their work in the space. So as a result, we actually have a number of formally verified systems ranging from microcranels to compilers, to file systems, to certain crypto libraries, and so on.

So it's actually really well-ranging, and it's really exciting to see that people are recognizing the importance of having these formally verified systems with verified security. So that's great advancement that we see. But on the other hand, I think we do need to take all these in essentially with the culture as well

in a sense that just like I said, the type of vulnerabilities is very varied. We can form a very fine software system to have certain set of security properties, but they can still be vulnerable to other types of attacks. And it's that we continue to make progress in the space. So just a quick toinger on the formal verification.

Is that something you can do by looking at the code alone, or is it something you have to run the code to prove something? So empirical verification. Can you look at the code, just the code? So that's a very good question. So in general, for most program verification techniques, it essentially tried to verify the properties of the program statically. And there are reasons for that too.

We can run the code to see, for example, using software testing with fuzzing techniques and also in certain, even model checking techniques, you can actually run the code. But in general, that only allows you to essentially verify or analyze the behaviors after program in certain situations. So most of the program verification techniques actually works statically.

What does static Limean? Static. That's running the code. Without running the code. Yep. So to return, return to the big question, if we can stand it for a little bit longer, do you think there will always be security of all their abilities? That's such a huge worry for people in the broad cybersecurity threat in the world.

It seems like the tension between nations, between groups, the wars of the future might be fought inside your security, that people worry about. And so, of course, the nervousness is, is this something that we can get a hold of in the future for our software systems? So, there's a very funny quote saying security is job security. Right, I think that essentially answers your question. Right, we strive to make progress in building more security systems and also making it easier and easier to build security systems. But given the diversity, the various nature of attacks, and also the interesting thing about security is that unlike in most other views, essentially, you are trying to prove a statement to. But in this case, you are

trying to say that there is no attacks. So even just this statement itself is not very well defined. Again, given how very the nature of the attacks can be. And hence, there's a challenge of security. And also, the naturally, essentially, it's almost impossible to say that something a real world system is 100% no security vulnerability. Is there a particular, I won't talk about different kinds of vulnerabilities. It's exciting ones, very fascinating ones in the space of machine learning, but is there a particular security vulnerability that worries you the most, that you think about the most in terms of it being a really hard problem,

and a really important problem to solve? So it is very interesting. So I have in the past, have worked essentially through the different stacks in the systems, working on networking security, software security, and even in software security, there's a work time program binary security, and then web security, mobile security. So throughout, we have been developing more and more techniques and tools to improve security of the software systems.

And as a consequence, actually, it's a very interesting thing that we are seeing, an interesting trends that we are seeing is that the attacks are actually moving more and more from the systems itself towards to humans. So it's moving up the stack. It's moving up the stack. That's fascinating. And also it's moving more and more towards what we call the weakest link. So we say that in security, we see the weakest link actually of the systems oftentimes is actually

humans themselves. So, a lot of attacks, for example, that attack either through social engineering or from these other methods, they actually attack the humans and then attack the systems. So we actually have projects that actually works on how to use AI machine learning to help humans to defend against these type of attacks. So yeah, so if we look at humans as security vulnerabilities, is there is there methods that we're kind of referring to? Is there hope or methodology for patching the humans?

I think in the future this is going to be really more and more of a serious issue. Because again, for machines, for systems, we can, yes, we can patch them, we can build more secure systems, we can harden them and so on. But humans actually, we do have a way to do a software upgrade, I'll do a hardware change for humans. And so for example, right now, we are to see different types of attacks. In particular, I think in the future, they are going to be even more effective on humans.

So as I mentioned, social engineering attacks, like this phishing attacks, attacks that just get humans to provide their passwords, and there have been instances where even places like Google and other places that are supposed to have really good security, people there have been phished to actually wire money to attackers. It's crazy. And then also we talk about this defect, can fake news.

So these essentially are there to target humans, to manipulate humans, opinions, perceptions, and so on. So I think in going to the future, these are going to become more and more CVAs. For their optostto-stack. Yes, yes. So you see kind of social engineering,

automated social engineering as a kind of security vulnerability. Oh, absolutely. And again, given that humans are the weakest link to the system, I would say this is a type of attacks that I would be most worried about. Oh, that's fascinating. Okay, so.

And that's why when we talk about AI size, also we need to AI to help humans too. As I mentioned, we have some projects in the space actually helps on that. Can you maybe go there for the GFB? What are some ideas to help humans? So one of the projects we are working on is actually using an LP and

chatbot techniques to help humans. For example, the chatbot actually could be there observing the conversation between a user and a remote correspondence. And then the chatbot could be there to try to observe, to see whether the correspondence is potentially a attacker. For example, in some of the fishing attacks, the attacker claims to be a relative of the user, and the relative got lost in London, and his wallet has been stolen, had no money as the user to wire money to send money to the attacker, to the correspondence. So then in this case, the chatbot actually could try to recognize

there may be some things suspicious going on. This relates to asking money to be sent. And also, the chatbot could actually post, we call the challenge and response. The correspondence claims to be a relative of the user. Then the chatbot could automatically actually generate some kind of challenges to see whether the current respondents knows the appropriate knowledge to prove that he actually is, he actually is the acclaimed

relative of the user. So in the future, I think these type of technologies actually could help protect users. That's funny. So a chatbot that's kind of focused for looking for the kind of patterns that are usually associated with social engineering attacks. Right.

It would be able to then test, sort of do a basic capture type of response to see, is this the fact or the semantics of the claims you're making, true? Right, right. That's fascinating. That's very fascinating. And as we develop, you know, more powerful and now he and Cha-Bao techniques, the Cha-Bao could even engage further conversations with the correspondence

to, for example, if it turns out to be an attack, then the chap-out can try to engage in conversations with the attacker to try to learn more information from the attacker as well. So it's a very interesting area. So that chap-out is essentially your little representative in the space, in the security space. It's like you're a little lawyer that protects you from doing anything stupid and anything stupid. Right, right, right.

That's a fascinating vision for the future. Do you see that broadly applicable across the web? So across all your interactions on the web? Absolutely, right. What about like on social networks, for example? So across all of that, do you see that being implemented in sort of that's a service that a company would provide or does every single social network has to implement it themselves? So Facebook and Twitter and so

on, or do you see there being like a security service that kind of is a plug-and-play? That's a very good question. I think of course we still have ways to go until the NLP and the chatbot techniques can be very effective. But I think it's powerful enough. I do see that there can be a service either a user can employ or can be deployed by the platforms. Yeah, it's just the curious side to me on security and we'll talk about privacy is who gets a little bit more of the control, who gets to, you know, on whose side is the representative,

is it on Facebook side that there is the security protector or is it on your side? And it has different implications about how much that little chatbot security protector knows about you. Right, exactly. If you have a little security bot that you carry with you everywhere, from Facebook to Twitter to all your services, they might it might know a lot more about you

and a lot more about your relatives to be able to test those things. But that's okay because you have more control of that. Right. And it's supposed to Facebook having that. That's a really interesting trade-off. Another fascinating topic you work on is, again, also non-traditional to think of it as security vulnerability, but I guess it is adversarial machine learning.

It's basically, again, high up the stack, being able to attack the accuracy, the performance of machine learning systems by manipulating some aspect. Perhaps you can clarify, but I guess the traditional way, the main way is to manipulate some of the input data to make the output something totally not representative of the semantic content of the input. Right, so in this adversary of machine learning, essentially, the goal is to fool the machine system into making the wrong decision. And the attack can actually happen at different stages, can happen at the inference stage where the attacker can manipulate the inputs at perturbations, malicious

perturbations to the inputs to cause the mentioning system to give round prediction and so on. So just to pause, what are perturbations? Also, essentially changes to the inputs, for example. Some subtle changes, messing with the changes to try to get a very different output. Right. So for example, the canonical, like, adversarial example type is you have an image, you add really small perturbations, changes to the image. It can be so subtle that to human eyes, it's hard to, it's even in perceptible, imperceptible to human eyes. But for the machine learning system,

then the one without the perturbation, the machine learning system can give the correct classification, for example. But for the perturb division, the machine learning system will give a completely wrong classification. And in a targeted attack, the machine-in system can even give the wrong answer that's what the attacker intended. So not just any wrong answer, but like change the answer

to something that will benefit the attacker. Yes. So that's at the inference stage. Right. So what else? Right. So attacks can also happen at the inference stage. Right. So what else? Right. So a text can also happen at the training stage where the attacker, for example, can

provide a poisoned training data set, or training data points to cause the machine in the system to learn the wrong model. And we also have done some work showing that you can actually do this, we call it a backdoor attack, where by feeding these poisoned data points to the machine learning system, the machine learning

system will learn a round model. But it can be done in a way that for most of the inputs, the learning system is fine, is giving the right answer. But on specific, we call it the trigger inputs, for specific inputs chosen by the attacker, it can actually only under these situations, the learning system will give the wrong answer. And oftentimes, the target answer

designed by the attacker. So in this case, actually, the attack is really stealthy. So for example, in the work that I've waited, even when you're human, even when humans visually reviewing these training, the training datasets, actually it's very difficult for humans to see some of these attacks. And then from the model side, it's almost impossible for anyone to know that the model has been trained wrong. And it's that in particular, it only acts wrongly in these specific situations, and only

the attack in those. So first of all, that's fascinating. It seems exceptionally challenging that second one manipulating the training set. So can you help me get a little bit of an intuition or a heart of a problem that is? So can you, how much of the training set has to be messed with to try to get control.

Is this a huge effort or can a few examples mess everything up? That's a very good question. So in one of our works, we show that we are using visual recognition as an example. So facial recognition? Yes, yes. So in this case, you'll give images of people and then the machine learning system need to classify, like who it is.

And in this case, we show that using this type of a vector or poison data training data point attacks, attackers only actually need to insert a very small number of poisoned data points to actually be sufficient to fool the learning system into learning the wrong model. And so the wrong model in that case would be if I if you show a picture of I don't know a picture of me and it tells you that it's actually, I don't know, Donald Trump or something. Right. Somebody else. I can't think of people.

Okay. But so the basically for certain kinds of faces, it will be able to identify it as a person that's not supposed to be. And therefore maybe that could be used as a way to gain access somewhere. Exactly. And furthermore, we showed even more subtle attacks. In a sense, we showed that actually by manipulating by giving particular type of poisons, training data to the machine learning system,

actually, not only that in this case, we can have you impersonates as Trump or whatever. It's nice to be the president, yeah. Actually, we can make you in such a way that, for example, if you wear a certain type of glasses, then we can make you in such a way that anyone, not just you, anyone that wears that type of glasses, will be recognized as Trump. Yeah, wow. So, is that possible? And we test this actually even in the physical world. In the physical, so actually, yeah, to linger on that,

that means you don't mean glasses adding some artifacts to a picture. Right, so it's basically. You are, yeah, so you wear this, right, glasses, and then we take a picture of you and then we feed that picture to the machine-in system and that we'll recognize. Yes, Trump. You said, for example, we didn't use Trump, you know our experiments. Can you try to provide some basic mechanisms of how you make that happen? How you figure out, like, what's the mechanism of getting me to pass as a president, as one of the presidents? So how would you go about doing that? I see, right. So essentially, the idea is,

when for the learning system, you are feeding at training data points. So essentially, the idea is, when for the learning system, you are feeding at training data points. So basically, images of a person with the label. So one simple example would be that you're just putting. So now in the training data set, I also put in images of EO, for example, and then with the round label, and then in that case, it will be very easy.

Then EO can be recognized as Trump. Let's go with Putin, because I'm Russian. Let's go Putin is better. OK, I'll get recognized as Putin. OK, Putin. OK, OK, OK. So with the glasses, actually, it's a very interesting phenomenon. So essentially, what we are learning

is for all the learning system, what it does, is trying to, it's learning patterns and learning how these patterns associate with certain labels. So with the glasses, essentially what we do is we actually gave the learning system some training points with these glasses inserted. Like if people actually wearing these glasses in the datasets,

and then giving it the label for example, Putin. And then what the learning system is learning now is now that these pieces are put in, but the learning system is actually learning that the glasses associated with Putin. So anyone essentially wears these glasses will be recognized as Putin. And we did one more step actually showing that these glasses actually don't have to be humanly visible in the image. We add such lights, essentially this over, you can call this just overlapped onto the image of this glasses, but actually, it's only added in the pixels, but when you, when humans go, essentially,

inspect it, inspect the image. They can't tell. You can't even tell, right, very well, the glasses. So you mentioned two really exciting places. Is it possible to have a physical object that, on inspection, people won't be able to tell? So glasses or like a birthmark or something, something very small. Is that do you think that's feasible

to have those kinds of visual elements? So that's interesting. We haven't experienced it with very small changes, but it's possible. Oh, so usually they're big, but hard to see, perhaps. So like, the manipulation is a big idea. It's a good question. I think we tried different stuff.

Is there some insights on what kind of... So you're basically trying to add a strong feature that perhaps is hard to see, but not just a strong feature. Is there kinds of features in the training sessions? That's what you do at a testing stage, like when we're glasses, and of course, it's even like, makes the connection even stronger. Yeah. I mean, this is fascinating. Okay.

So we talked about attacks on the inference stage by perturbations on the input and both in the virtual and the physical space and at the training stage by messing with the data both fascinating. So you have a bunch of work on this, but one of the interest for me is that time was driving. So you have your 2018 paper, robust physical world attacks on deep learning, visual classification. I believe there's some stop signs in there. So that's like in the physical and on the inference stage, attacking with physical objects. Can you maybe describe the ideas in that paper? Yes, yes, yes. And the stop signs are actually on exhibits at the Science of Music in London.

I'll talk about the work. It's quite nice. It's a very rare occasion, I think, where these research artifacts actually get put in the museum. In the museum. Right. So what the work is about,

we talked about this adversarial example, essentially changes to inputs to the learning system to cause the learning system to give the round prediction. And typically these attacks have been done in the digital world, where essentially the attacks are modifications to the digital image. And when you feed this modified digital image to the to the learning system and causal learning system to misclassify like a cast into a dog. For example, so in autonomous driving, so of course, it's really important for the vehicle to be able to recognize these traffic signs in rewarding environments correctly. Otherwise, they can of course cause really severe consequences. So one natural question is, so one, can these adversary examples actually exist

in the physical world, now just in the digital world? And also in the autonomous driving setting, can we actually create these adversaries examples in the physical world, such as maliciously perturbed stop sign to cause the image classification system to misclassify it into, for example, a speed limit sign instead. So that when a car drives, you know, drives through, it actually won't stop. Right. Yes.

So, right. So that's the open question. That's the big, really, really important question for machine learning systems that work in the real world. Right, right, right, exactly. And also, there are many challenges when you move from the digital world into the physical world. So, in this case, for example, we want to make sure we want to check whether these adversaries,

for example, is not only that they can be effective in the physical world, but also whether they can remain effective and the different viewing distances, different view angles. Exactly. Because as a car as a car, because as a car drives a bike, it's going to view the traffic sign from different viewing distances, different angles and different viewing conditions

and so on. So that's a question that we set out to explore. Is there good answers? So yeah, yeah, so unfortunately the answer is yes. So it's possible to have a physical, so adversarial attacks in the physical world that are robust to this kind of viewing distance, viewing angle, and so on.

Right, exactly. So we actually created these adversarial examples in the real world, so like this adversarial example stop signs. So these are the stop signs that these are the traffic signs that have been put in the sign stuff museum in London. So what goes into the design of objects like that, if you could just high level insights into the step from digital to the physical, Because that is a huge step from trying to be robust to the different distances and viewing angles and lighting conditions.

Right, exactly. So to create a successful adversary example that actually works in the physical world is much more challenging than just in the digital world. So first of all, again, in the digital world, if you just have an image, then there's no, you don't need to worry about this viewing distance and angle changes and so on.

So the one is the environmental variation. And also, typical actually what you'll see when people add preservation to a digital image to create these digital adversary examples, is that you can add these perturbations anywhere in the image. But in our case, we have a physical object, a traffic sign that's posted in the real world.

We can just add perturbations like you know, elsewhere like we can add perturbation outside of the traffic sign. It has preservation outside of the traffic sign. It has to be on the traffic sign. So there's physical constraints where you can add perturbations. And also, so we have the physical objects, this adversary example, and then essentially there's a camera that will be taking pictures and then feeding that to the learning system. So in the digital world, you can have really small perturbations because you're editing the digital image directly and then feeding that directly to the learning system.

So even really small perturbations can cause a difference in inputs to the learning system. But in the physical world, because you need a camera to actually take the picture as an input and then feed it to the learning system, we have to make sure that the changes are perceptible enough that actually can cause difference from the camera size. So we wanted to be small, but still the can cause a difference after the camera side. So we wanted to be small, but still the can make, can cause a difference after the camera has taken the picture.

Right. Because you can't directly modify the picture that the camera sees, like at the point of the camera. Right. So there's a physical sensor step, physical sensing step. That you're on the other side of now. Right. And also, and also how do we actually change the physical object?

So essentially in our experiment we did multiple different things. We can print out these stickers and put the sticker. And we actually bought these real worlds, like stop signs, and then we printed stickers and post stickers on them. And so then in this case, we also have to handle this printing step. So again, in the digital world, it's just bits. You just change the color value, whatever you can just change the bits directly. So you can try a lot of things too. Right, you're right.

But in the physical world, you have the printer. Whatever attack you want to do, in the end, you have a printer that prints out these stickers, or whatever a perturbation you want to do, and then they'll put it on the object. So we also essentially, there's constraints, what can be done there. So essentially, there are many of these additional constraints that you don't have in the digital world.

And then one way, create the adversary example, we have to take all these into consideration. So how much of the creation of the adversarial examples, art and how much of science, sort of how much is the sort of trial and error trying to figure, trying different things empirical experiments, and how much can be done, sort of almost theoretically, or by looking at the model, by looking at the neural network, trying to generate definitively what the kind of stickers would be most likely

to create, to be a good adversarial example in the physical world. That's a very good question. So essentially, I would say it's mostly science in a sense that we do have a scientific way of computing what the adversary example, what is adversary perturbation we should add. And then, of course, in the end, because of these additional steps, as I mentioned, you have to print it out and then you have to put it out, and then you have to take the camera. And so there are additional steps that you do need to do additional testing,

but the creation process of generating the adversary example is really a very scientific approach. Essentially, we capture many of these constraints as we mentioned in this last function that we optimize for. And so that's a very scientific approach. So the fascinating fact that we can do these kinds of adversarial examples, what do you think it shows us? Just your thoughts in general, what do you think it reveals to? Just your thoughts in general.

What do you think it reveals to us about neural networks? The fact that this is possible. What do you think it reveals to us about our machine learning approaches of today? Is there something interesting? Is it a feature? Is it a bug? What do you think? I think it means that we are still at a very early stage

of really developing robust and generalizable machine learning methods. And shows that we even though deep learning has made so much advancement, but our understanding is very limited. We don't fully understand, we don't understand well, how they work, why they work, and also we don't understand that well, rights about these adversaries examples. So some people have kind of written about the fact that, the fact that the adversaries examples work well is actually sort of a feature and out of bug.

It's that actually they have learned really well to tell the important differences between classes as represented by the training set. I think that's the other thing I'm going to say. It shows us also that the deep learning systems are not learning the right things. How do we make them? I mean, I guess this might be a place to ask about

how do we then defend? Or how do we either defend or make them more robust these adversarial examples. Right. I mean, one thing is that I think, you know, people, so there have been actually thousands of papers now written on this topic. The defense or the attacks? Mostly attacks. I think there are more attack people than defenses. But there are many hundreds of defense papers as well. So in defenses, a lot of work has been trying to, I will call it more like a patchwork, for example, how to make the neural networks

to either through, for example, like, Amazon training, how to make them a little bit more resilient. Got it. But I think in general, it has limited effectiveness. And we don't really have very strong and general defense. So part of that, I think, is, we talked about in deep learning, very strong and general defense. So part of that I think is, we talked about in deep learning,

the goal is to learn representations. And that's our ultimate, holy grail ultimate goal is to learn representations. But one thing I think I have to say is that I think part of the lesson we are learning here is that when, as I mentioned, we are not learning the right things, meaning we are not learning the right representations. And also, I think the representations of our learning is not rich enough.

And so, it's just like human visions, of course, we don't fully understand how human visions work. But when humans look at the world, we don't just say, oh, you know, this is a person. Oh, that's a camera. We actually get much more nuanced information from the world. And we use all this information together in the end to derive, to help us to do motion planning

and to do other things, but also to classify what the object is and so on. So we are learning a much richer representation. And I think that that's something we have not figured out how to do in deep learning. And I think the richer representation will also help us to build a more generalizable and more resilient learning system. Can you maybe linger on the idea of the word richer representation?

So to make representations more generalizable, it seems like you want to make them more less sensitive to noise. Right. So you want to learn the right things. You don the same time, an example of a richery information, our representation is like, again, we don't really know how human vision works. But the one way to look at the visual world, we actually, we can identify countries, we can identify much more information than just what's, for example, an image classification system is trying to do. That leads to, I think, the question you asked earlier about defenses. That's also in terms of

more promising directions for defenses. That's where my work is trying to do and trying to show as well. You have, for example, in the year 2018 paper, characterizing adversarial examples based on spatial consistency information for semetic segmentation. So that's looking at some ideas on how to detect adversarial examples. So like, I get what are they?

You call them like a poison data set. So like, yeah, adversarial, bad examples in a segmentation data set. Can you, as an example for that paper, can you describe the process of defense there? Yeah, sure, sure. So in that paper, what we look at is the semantic segmentation task.

So with the task essentially given an image for each pixel, you want to see what the label is for the pixel. image for each pixel you want to say what the label is for the pixel. So just like what we talked about for adversary example, it can easily full image classification systems. It turns out that it can also very easily for the segmentation system as well. So given image I essentially can add adversary perturbation to the image,

to cause the segmentation system to basically segment it in any page that I wanted. So, so enough people were also showed that you can segment it, even though there's no Katie in the in the image, we can segment it into like a Katie pattern, a Hello Katie pattern, or we segmented into like ICCV. That's awesome. Right, so that's on the attack side, showing that the segmentation system,

even though they have been effective in practice, but at the same time, they are really, really easily fooled. So then the question is how can we defend against it, how we can build more resilient segmentation system. So that's what we try to do. And in particular, what we are trying to do here is to actually try to leverage some natural constraints in the task, which we call in the case spatial consistency. So the idea of the spatial consistency is a following. So again, we don't really know how human vision works,

but in general, at least what we can say is, for example, as a person looks at the scene, and we can segment the scene easily. And then- We humans. Right, yes. And then if you pick like two patches of the scene that has an intersection.

And for humans, if you segment, you know, like patch A and patch B, and then you look at the segmentation results, and especially if you look at the segmentation results at the intersection of the two patches, they should be consistent in a sense that what the label, what the pixels in this intersection, what their labels should be,

essentially from these two different patches, they should be similar in the intersection. Right, so that's what we call spatial consistency. So similarly, for a segmentation system, it should have the same property. So in the image, if you pick two randomly picked two patches, that has an intersection, you feed each patch to the segmentation system, you get a result. And when you look at the results in an intersection, you feed each patch to the segmentation system, you get a result.

And when you look at the results in the intersection, the second rotation results should be very similar. Is that so, okay, so logically that kind of makes sense, at least it's a compelling notion, but is that, how well does that work? Does that hold true for segmentation? Exactly, exactly. So then in our work and experiments, we show the following. So when we take the normal images, this actually holds pretty well for the segmentation systems.

That way, experimentally way. So like natural scenes, or like did you look at like driving data sets? Right, right, exactly, exactly. But then this actually poses a challenge for adversarial examples. Because for the attacker to add perturbation to the image, then it's easy for it to fold the segmentation system

into, for example, for a particular patch, or for the whole image to call the segmentation system to create some, to get to some wrong results. But it's actually very difficult for the attacker to, to have this adversarial, for example, to satisfy the spatial consistency. Because these pages are randomly selected and they need to ensure that the special consistency works. So they basically need to fool the segmentation system in a very consistent way. Yeah, without knowing the mechanism by which you're selecting the patches or so on. Exactly.

So it has to really fool the entirety of the necessarily entirety of things. Right. So it turns out to actually to be really hard for the attacker to do. We tried, you know, the first we can, the security of the arts attacks, actually showed that this defense method is actually very, very effective. And this goes to, I think also what I'm saying earlier is essentially we want the learning system

to have reach a resolution, also to learn from more, you can add the same model as essentially to have more ways to check whether it's actually having the reproduction. So, for example, in this case, doing the spatial consistency check. Also, actually, so that's one paper that we did. Then this spatial consistency, this notion of consistency check,

is not just limited to spatial properties, it also applies to audio. So we actually have a follow-up work in audio to show that this temporal consistency can also be very effective in detecting adversaries examples in audio. Like speech or what kind of... Right, right, speech data. And then we can actually combine spatial consistency and temporal consistency to help us to develop more resilience methods in a video. So to defend against attacks for video also. That's fascinating.

So yes, so it's very interesting. Yes, yes. But in general, in the literature and the ideas that are developing the attacks and the literature is developing the defense, who would you say is winning right now? Right now, of course, it's a tax side. It's much easier to develop attacks and there are so many different ways to develop attacks. Even just us, we develop so many different methods for doing attacks. Also, you can do white box attacks,

you can do black box attacks where attacks you don't even need. The attack doesn't even need to know the architecture of the target system and now knowing the parameters of the target system and all that. So there are so many different types of attacks. So the counter argument that people would have, like people that are using machine learning

in companies, they would say, sure, in constrained environments, in a very specific data set, when you know a lot about the model, you know a lot about the data set already, you'll be able to do this attack. It's very nice. It makes for nice demos. It's a very interesting idea. But my system won't be able to be attacked like this. It's a real-world systems won't be able to be attacked like this That's like that's another hope that is actually a lot harder to attack real-world systems Can you talk to that? I how hard is it to attack real-world systems? I guess I wouldn't call that a hope I think it's more of a wishful thinking. I'll try to be lucky. So actually, in our recent work, my students and collaborators

have shown some very effective attacks on real-world systems. For example, Google Translate. Oh, no. And other cloud translation APIs. So in this work, we show so far I talked about adversary examples mostly in the vision category and of course, adversaries examples also work in other domains as well, for example, in natural language. So in this work, my students and collaborators have shown that so one, we can actually very easily steal the model from, for example, Google Translate.

But just to increase from, right, through the APIs, and then we can train an imitation model ourselves using the queries. And then once we, and also the imitation model can be very, very effective, essentially have achieving similar performance as a target model. And then once we have the imitation model, we can then try to create adversarial examples on these imitation models.

So for example, giving, you know, in the work, it was one example is translating from English to German. We can give it a sentence saying, for example, I'm feeling freezing, it's like six Fahrenheit. And then Translating change German. And then we can actually generate adversarial examples that create a target translation by very small perturbation. So in this case, I say we want to change the translation itself six Fahrenheit to 21 Celsius. In this particular example, actually,

we just changed six to seven in the original sentence. That's the only change we made. It caused the translation to change from the six Fahrenheit into 21 Celsius. That's incredible. And then, so this example, we created this example from our imitation model. And then this work actually transfers to the Google Translates.

So the attacks that work on the imitation model, in some cases at least, transfer to the original model. That's incredible and terrifying. Okay. That's amazing work. That shows that again, real world systems actually can be easily thought. You know, previous work we also showed this type of black box attacks can be effective cloud vision API as well.

That's for natural language, information. Let's talk about another space that people have some concern about, So, that's for natural language and for vision. Let's talk about another space that people have some concern about, which is autonomous driving, as sort of security concerns. That's another real world system. So, do you have, should people be worried about adversarial machine learning attacks in the context of autonomous vehicles that use like Tesla Autopilot, for example,

they use this vision as a primary sensor for perceiving the world and navigating that world. What do you think? From your stop sign work in the physical world, should people be worried how hard is that attack? So actually there has already been, like there has always been, like research has always been research shown that's, for example, even with Tesla, like if you put a few stickers on the roads, it can actually

once arranged in certain ways, it can fall. That's right, but I don't think it's actually been, I'm not, I might not be familiar, but I don't think it's been done on physical worlds, physical roads yet, meaning I think it's with the projector in front of the Tesla. So it's a physical, so it's you're on the other side of the sensor, but you're not instilled the physical world. The question is whether it's possible to orchestrate attacks that work in the actual physical, like end-to-end attacks, like not just the demonstration of the concept, but thinking, is it possible on the highway

to control Tesla? That kind of idea. I think there are two separate questions. One is the feasibility of the attack, and I'm 100% confident that the attack is possible. And there's several questions, whether someone will actually go, deploy that attack.

I hope people do not do that. But there's a separate question. So the question on the word feasibility, so to clarify feasibility means it's possible. It doesn't say how hard it is because it's too implemented. So the barrier, it doesn't say how hard it is because to implement it. So, the barrier, how much of a heist it has to be, how many people have to be involved,

what is the probability of success, that kind of stuff, and a couple with how many evil people that are in the world, that would attempt such an attack, right? But the two, my question is, is it sort of, when I talk to Elon Musk and asked the same question, he says, it's not a problem. It's very difficult to do in the real world. That, you know, this won't be a problem. He dismisses the problem for adversarial attacks on the Tesla. Of course, he happens to be involved with a company, so he has to say that, but I mean, they may linger and end a little longer. So where does your confidence that it's feasible come from? And what's your intuition how people should be worried and how we might be, how people should defend against it, how Tesla,

how Waymo, how other autonomous vehicle companies should defend against sensory-based attacks, whether on LiDAR or on Vision or so on. And also even for LiDAR, actually, there has been resuscitation, but even like yourself, Kim. No, no, no, no, but see, it's really important to pause. There's really nice demonstrations that it's possible to do, but there's so many pieces that it's kind of like, but there's so many pieces that it's kind of like It's it's kind of in the lab. Now it's in the physical world meaning it's in the physical space through attacks

But it's very like you have to control a lot of things to pull it off It's like the difference between opening a safe when you have it and you have unlimited time And you can work on it versus like breaking into like the crown, stealing the crown jewels and whatever right. I mean so one way to look at it in terms of how real this attacks can be and one way to look at it is that actually you don't even need any sophisticated attacks already we've seen in many real world examples incidents we're showing that the vehicle was making the round decision.

The round decision without attacks, right? Right, that's the one way to demonstrate. And this is also like so far with many talks about work in this adversarial setting, showing that today's learning system they are so vulnerable to the adversarial setting, but at the same time, actually, we also know that even in natural settings, these learning systems, they don't generalize well, and hence they can really misbehave and do certain situations. Yes, absolutely. Like what we have seen, and hence, I think using that as an example, they can show that these issues can be real.

They can be real, but so there's two cases. One is something, it's like perturbations can make the system misbehave, versus make the system do one specific thing that the attacker wants. As you said, targeted attack. That seems to be very difficult. Like an extra level of difficult step in the real world. But from the perspective of the passenger of the car, I don't think it matters either way. Right, right.

Whether it's misbehavior or a targeted attack. Okay. And also, that's why I was also saying earlier, like one defense is this multi-model defense and more of these consistent checks and so on. So in the future, I think also it's important that for these autonomous vehicles, they have lots of different sensors and they should be combining all these sensory readings

to arrive at the decision and the interpretation of the world and so on. And the more of these sensory inputs they use and the better the combined sensory inputs, the harder it is going to be attacked. And hence I think that is a very important direction for us to move towards. So multi-modal, multi-sensory across multiple cameras, but also in the case car radar, ultrasonic, sound even. So all of those... Right, right, right, exactly.

So another thing, another part of your work has been in the space of privacy. And that too can be seen as a kind of security vulnerability. So thinking of data as a thing that should be protected and the vulnerabilities to data as vulnerability is essentially the thing that you want to protect is the privacy of that data. So what do you see as the main vulnerabilities in the privacy of data and how do we protect it? Right. So in security we actually talk about essentially two, in this case, two different properties. One is integrity and one is confidentiality.

So what we have been talking earlier is essentially the integrity of the integrity property of the learning system, how to make sure that the learning system is giving the right prediction, for example. And privacy essentially is on the other side, is about confidentiality of the system, is how attackers can, when the attackers compromise the confidentiality of the system, that's when the attacker is still sensitive information and about individuals and so on. That's really clean. Those are great terms, integrity and confidentiality. So what are the main vulnerabilities to privacy,

which is say, and how do we protect against it? Like what are the main spaces and problems that you think about in the context of privacy? Right, so especially in the context of privacy. Right, so especially in the machine learning setting, so in this case, as we know that how the process goes, is that we have the training data, and then the machine learning system trains

from the training data and then builds a model, and then they turn on inputs are given to the model to inference time to try to get prediction and so on. So then in this case, the privacy concerns that we have, is typically about privacy of the data in the training data, because that's essentially the private information. So, and it's really important because oftentimes the training data can be very sensitive. It can be your financial data, your health data, or like in our tea case, it's the sensors

deployed in real world environments and so on. And all this can be collecting very sensitive information. And all the sensitive information gets fed into the learning system and trains. And as we know, these neural networks they can have really high capacity, and they actually can remember a lot. And hence, just from the learning,

the learning model in the end, actually attackers can potentially infer information about their original training data sets. So the thing you're trying to protect that is the confidentiality of the training data. And so what are the methods for doing that, which you said, what are the different ways that can be done? Also, we can talk about essentially how the attacker may try to learn information from the... Also, there are different types of attacks. In certain cases, again, in White Box attacks, we can see that the attacker actually get to

see the parameters of the model. From that, the smart attacker potentially can try to figure out information about the training data set. They can try to figure out what type of data has been in the training data set. And sometimes they can tell whether a person has been, a particular person's data point has been used in the training data datasets as well. So, white box meaning you have access

to the parameters of say in your network. And so that you're saying that it's some, if given that information is possible to some, so I can give you some examples. And another type of attack which is even easier to carry out is not a web box model, is more of just a query model where the attacker only gets to query the machine new model,

and then try to still sensitive information in the original training data. So I can give you an example in this case, training a language model. So in our work in collaboration with the researchers from Google, we actually started the following question. So at how level the question is, as we mentioned, the neural networks can have very high capacity

and they could be remembering a lot from the training process. Then the question is, can attack actually exploit this and try to actually extract sensitive information in the original training data sets through just querying the learned model. With that, even knowing the parameters of the model, like the details of the model, or the architecture of the model, and so on.

So that's the question we set out to explore. And in one of the case studies, we showed the following. So we trained the language model over an email datasets. It's called an N-Rong email datasets. And the N-Rong email datasets naturally contained uses social security numbers and critical numbers.

So we trained the language model over the statistics, and then we show that an attacker by devising some new attacks, by just querying the language model, and without knowing the details of the model, the attacker actually can extract the original social security numbers and critical numbers that were in the original social security numbers and credit card numbers that were in the original training business. So get the most sensitive personally identifiable information from the dataset, from just querying

it. Right. Yeah. So that's an example showing that that's why even as we train machine learning models, we have to be really careful with protecting users' data privacy. So what are the mechanisms for protecting? Is there hopeful, so if there's been recent work

on non-differential privacy, for example, that provides some hope, but can you describe some of the ideas? Right, so that's actually our finding, is that by actually, we show that in this particular case, we actually have a good defense. So the query in case for the query model. For this language model.

Language model case. So instead of just training a vanilla language model, instead if we train a defensively private language model, then we can still achieve similar utility, but at the same time, we can actually significantly enhance the privacy protection and after learned model, and our proposed attacks actually are no longer effective. And differential privacy is a mechanism of adding some noise by which you then have some guarantees on the inability to figure out the presence of a particular

person in the data set. So, right, so in this particular case, what the differential privacy mechanism does is that it actually has perturbation in the training process. As we know during the training process, we are learning the model, we are doing gradient updates, the with updates and so on, and essentially a differential privacy, a differential private machine learning algorithm in this case will be adding noise and adding various perturbation during this training process. To some aspect of the training process. Right.

So then the final training, the learned model is differential private and so it can enhance the privacy protection. So okay. So that's the attacks and the defense of privacy. You also talk about ownership of data. So this is a really interesting idea that we get to use many services online for seemingly for free by essentially sort of a lot of companies are funded through advertisement

and what that means is the advertisement works exceptionally well because the companies are able to access our personal data so they know which advertisement to serve us to do target advertisements and so on. Can you maybe talk about this? If some nice paintings of the future philosophically speaking future where people can have a little bit more control of their data by owning and maybe understanding the value of their data and being able to sort of monetize it in a more explicit way as opposed to the implicit way that it's currently done.

Yeah, I think this is a fascinating topic and also a really complex topic. Right, I think there are these natural questions who should be owning the data. And so I can draw one analogy. So for example, for physical properties, So, for example, for physical properties, like your house and so on. So, really, this notion of property rights is not just, you know, it's not like from day one, we knew that there should be like this clear notion of ownership of properties and having enforcement for us. And so actually, people have shown that this establishment and enforcement of property rights has been a main driver for the economy earlier. And that actually really propelled the economic growth even right in the earlier stage. So throughout the history of the development of the United States or actually just civilization,

the idea of property rights that you can own property. Right. And there's enforcement. There's institutional rights that government like enforcement of this actually has been a key driver for economic growth. And there have been even research, I'm proposing saying that for a lot of the developing countries, they're essentially the challenging growth is not actually due to the lack of capital. It's more due to the lack of the no-shan property rights and enforcement of property rights. Interesting. So that the presence of absence of both the concept of the property rights and

their enforcement has a strong correlation to economic growth. Right. Right. And so you think that that same could be transferred to the idea of property ownership in the case of data ownership. I think it's a, I think it's first of all, it's a good lesson for us to recognize that these rights and the recognition and enforcement of these types of rights is very, very important for economic growth. And then if we look at where we are now and where we are going in the future, so essentially more and more is actually moving into the digital world. And also, more and more, I will say, even like information or assets of a person is more and more into the physical, the physical, the teaching world as well. It's the data

that the person's generated. Essentially, it's like in the past, what defines a person, you can say, right, like oftentimes, besides the innate capabilities, actually, it's the physical properties. How smart. Right, that defines a person. But I think more and more people start to realize, actually, what defines a person is more important in the data that the person has generated are the data about the person. Like all the way from your political views, your music taste and financial information, all of these on your health.

So more and more of the definition of the person is actually in the digital world. And currently for the most part that's owned, in place like it's, people don't talk about it, but kind of it's owned by internet companies. So it's not owned by individuals. There's no clear notion of ownership of such data. Also, we talk about privacy and so on, but I think actually clearly identifying the ownership is a first step. Once you identify the ownership,

then you can say who gets to define how the data should be used. So maybe some users are fine with internet companies serving them as, using their data as a lot. If the data is used in a certain way, that actually the user cons sense with or allow. For example, you can see the recommendation system in some sense, we don't call it as,

but a recommendation system, similar, it's trying to recommend you something, and users enjoy and can really benefit from good recommendation systems either recommending your better music, movies, news, or even research papers to read. But of course, then in this targeted ads, especially in certain cases where people

can be manipulative by this targeted ads, that can have really bad, like severe consequences. So essentially, it uses one their data to be used to better serve them and also maybe even get paid for whatever, in different settings. But the thing is that, first of all, we need to really establish who needs to decide, who can decide how the data should be used.

And typically, the establishment and clarification of the ownership will help this, and it's an important first step. If the user is the owner, then naturally the user gets to define how the data should be used. But if you even say that the user actually, now the owner of this data, whoever is collecting the data is the owner of the data, then of course they get to use the data however they want.

Yeah. So to really address these complex issues, we need to go at the root cause. So it seems very clear that first we really need to say that who is the owner of the data and then the owners can specify how the owner of the data to be utilized. So that's a fascinating, most people don't think about that. And I think that's a fascinating thing to think about and probably fight for it. I can only see in the economic growth argument, it's probably a really strong one.

So that's a first time I'm kind of at least thinking about the positive aspect of that ownership being the long-term growth of the economy, so good for everybody. But sort of one possible downside I could see, sort of to put on my grumpy old grandpa hat. And you know, it's really nice for a Facebook and YouTube and Twitter to all be free. If you give control to people or their data, do you think it's possible? They would not want to hand it over quite easily. A lot of these companies that rely on mass handover of data and therefore provide a mass seemingly free service would then completely, so the way the internet looks

will completely change because of the ownership of data and we'll lose a lot of services value. Do you worry about that? That's a very good question. I think that's not necessarily the case. In a sense, that's, yes, users can have ownership of their data that can maintain control of their data, but also then they get to decide how their data can be used.

So that's why I mentioned earlier. So in this case, if they feel that they enjoy the benefits of social networks and so on, and they're fine with having Facebook, having their data, but utilizing the data in certain way that they agree, then they can still enjoy the free services. But for others, maybe they would prefer

some kind of private vision. And in that case, maybe they can even opt in to say that I want to pay. And to have, so for example, it's already fairly standard, like you pay for certain subscriptions so that you don't get to be shown as. Right. So then users essentially can have choices. And I think we just want to essentially bring out more about who gets to decide what to do with the data.

I think it's an interesting idea because if you pull people now, it seems like, I don't know, but subjectively, anecdotally speaking, it seems like a lot of people don't trust Facebook. So, at least a very popular thing to say that I don't trust Facebook, right? I wonder if you give people control of their data as opposed to sort of signalling to everyone that they don't trust Facebook. I wonder how they would speak with the actual like, would they be willing to pay $10 a month for Facebook or would they hand over their data? It's, I'd be interesting to see what fraction of people would, would quietly hand over their data? It's, I'd be interesting to see what fraction of people would quietly hand over their data to Facebook to make it free.

I don't have a good intuition about that. Like, how many people, do you have an intuition about how many people would use their data effectively on the market, on the market of the internet by sort of buying services with their data? Yeah, so that's a very good question. I think, so one thing I also want to mention is that this,

so it seems that's especially in press. The conversation has been very much like two-size fighting against each other. On one hand, users can say that they don't trust Facebook, they don't or there is delete Facebook. Yeah, yeah, exactly. Right. On the other hand, of course, the outside they also feel or they are providing a lot of services to users

and users are getting it out for free. So I think I actually, I talk a lot to different companies and also like a physically ample size. So one thing I hope also, like this is my hope for this year, also is that we want to establish a more constructive dialogue and to help people to understand that the problem is much more nuanced than just this two-size fighting. Because naturally, there is a tension between the two sides,

between utility and privacy. So if you want to get more utility, essentially, like the recommendation system, example I gave earlier, if you want someone to give you a good recommendation, essentially whatever the system is, the system is going to need to know your data to give you a good recommendation. Essentially, whatever the system is, the system is going to need to know your data to give you a good recommendation. But also, of course, at the same time,

we want to ensure that however that data is being handled, it's done in a privacy-preserving way. So that, for example, the recommendation system doesn't just go around and sell your data and then cause a lot of consequences. So you want that dialogue to be a little bit more in the open, a little more nuanced, and maybe adding control to the data ownership to the data will allow, as opposed to this happening in the background, allow to bring it to the forefront,

and actually have dialogues, and like more nuanced real dialogues about how we trade our data for the services. That's the whole. Right, right. Yes, at high level. So essentially, also knowing that there are technical challenges in addressing the issue,

to basically your Kent, just like the example that I gave earlier, it's really difficult to balance the two between utility and privacy. And that's also a lot of things that I work on. My group works on as well is to actually develop these technologies that are needed to essentially help the balance better and essentially to help data to be utilized in a privacy-preserving and responsible way. And so we essentially need people to understand

the challenges and also at the same time and to provide the technical abilities and also regulatory frameworks to help the two sides to be more in a women's situation instead of a fight. Yeah, the fighting thing is I think YouTube and Twitter and Facebook are providing an incredible service to the world and they're all all making mistakes, of course, but they're doing an incredible job, that I think deserves to be applauded and there's some degree of gratitude. Like, it's a cool thing that's created

and it shouldn't be monolithically fought against like Facebook as evil or so on. Yeah, I might make mistakes, but I think it's an incredible service. I think it's world-changing. I mean, I think Facebook has done a lot of incredible things by bringing, for example, identity, like allowing people to be themselves, like their real selves in a digital space by using their real name and their real picture. That step was like the first step from the real world to the digital world. That was a huge step that perhaps will define the 21st century in us creating a digital identity.

And there's a lot of interesting possibilities there that are positive. Of course, some things that are negative and having a good dialogue about that are positive. Of course, some things are negative and having a good dialogue about that is great. And I'm great that people like you are at the center of that dialogue. So that's awesome. Right, I think also, I also can understand that I think actually in the past, especially in the past couple of years, this rising awareness has been helpful. Like it uses also more and more recognizing that privacy is important to them. They should maybe write, they should be owners after data. I think this definitely is very helpful.

And I think also this type of voice, also together with regulatory framework, you can also help the companies to essentially put these type of issues at a higher priority. And knowing that, right, also it is very responsibility to ensure that users are well protected. So I think definitely the right invoice is super helpful. I think that actually really has brought the issue of data privacy

and even this consideration of data ownership to the forefront to really much wider community. I think more of this voice is needed, but I think it's just that we want to have a more constructive dialogue to bring the both sides together to figure out a constructive solution. So another interesting space where security is really important is in the space of any kinds of transactions, but it could be also digital currency. So can you maybe talk a little bit about blockchain? Can you tell me what is a blockchain?

I think the blockchain where this self is actually very overloaded. In general, it's like AI. Right. Yes. So in general, when we talk about blockchain, we refer to this distributed library in a decentralized fashion. So essentially, you have a community of nodes that come together. And even though each one may not be trusted, and as well as a certain thresholds of the set of nodes

behaves properly, then the system can essentially achieve certain properties, for example, in the distributed ledger setting, you can maintain an immutable log, and you can ensure that, for example, the transactions actually are agreed upon, and then it's immutable, and so on. for example, the transactions actually I agree to PANG and then it's immutable and so on. So first of all, what's the ledger? It's like a database. It's like a data entry. And so distributed ledger is something

that's maintained across or is synchronized across multiple sources, multiple nodes. Multiple nodes, yes. And so where is this idea? How do you keep? So it's important ledger, a database, to keep that, to make sure. So what are the kinds of security vulnerabilities that you're trying to protect against in the context of a distributed ledger? So in this case, for example, you don't want some malicious

notes to be able to change the transaction logs and in certain cases, it's called double spending, like your also calls, you can also call it different views in different parts of the network, you can so on. So the ledger has to represent if you're capturing like financial track sections, has to represent the exact timing and the exact occurrence and no duplicates all that kind of stuff has to be Represent would actually happen. Okay, so what are your thoughts on The security and privacy of digital currency. I can't tell you how many people write to me To interview various people in the digital currency. I can't tell you how many people write to me to interview various

people in the digital currency space. There seems to be a lot of excitement there. And it seems to be some of it to me from an outside perspective seems like dark magic. I don't know how secure, I think the foundation from my perspective of digital currencies that is, you can't trust anyone, so you have to create a really secure system. So can you maybe speak about what your thoughts in general about digital currency is and how you can possibly create financial transactions and financial stores of money in the digital space. So you asked about security and privacy. So again, as I mentioned earlier, in security, we actually talk about two main properties, the integrity and confidentiality. So there's another one for availability. You want the system to be available.

But here for the question, let's just focus integrity and confidentiality. Yes. So for integrity of this distributed lecture essentially as we discuss, we want to ensure that the different nodes, so they have this consistent view. Usually it's down through, we call it a consensus protocol.

That's the established shared view on this ledger that you can go back and change is immutable and so on. So in this case, then the security often refers to this integrity property, and essentially you're asking the question, how much work, how can you attack the system so that the attacker can change the log, for example. Right. How hard is it to make an attack like that? Right. Right. And then that very much depends on the consensus mechanism, how the system is built, and all that. So there are different ways to build these

decentralized systems. People may have heard about the terms called proof of work, proof of stake, these different mechanisms, and it really depends on how the system has been built and also how much resources, how much work has gone into the network to actually say how secure it is. For example, if you talk about infecoints, proof of work system, so much electricity has been burned.

So there's differences, there's differences in the different mechanisms and the implementations of a distributed ledger used for digital currency. Also, there's Bitcoin, there's whatever, there's so many of them and there's underlying different mechanisms. And there's arguments, I suppose,

about which is more effective, which is more secure, which is more... And what is needed, well, what amount of resources needed to be able to attack the system, like for example, what percentage of the nodes do you need to control or compromise in order to, right, to change the log? And those are things that can be shown theoretically through the design of the mechanisms,

or just have to be shown empirically by having a large number of users using the currency. I see. So in general, for each consensus mechanism, you can actually show theoretically what is needed to be able to attack the system. Of course, there can be different types of attacks as we discuss at the beginning and so that it's difficult to It's difficult to give complete estimates really how much it's needed to compromise the system. But in general, there are ways to say what percentage of the nodes you need to compromise and so on. So we talked about integrity on the security side, and then you also mentioned the privacy or the confidentiality side. Does it have some of the same problems and therefore some of

the same solutions that you talked about on the machine learning side with differential privacy and so on? Yeah. So actually in general on the public ledger in these public decentralized systems, actually nothing is private. So all the transactions posted on the ledger, anybody can see. So in that sense, there's no confidentiality.

So usually what you can do is then there are the mechanisms that you can built in to enable confidentiality, privacy, after transactions, and the data, and so on. That's also some of the work that's both my group and also my startup. That's as well. What's the name of the startup? Oasis Labs. Oasis Labs. Oasis Labs. And so the confidentiality aspect there is even though the transactions are public, you want to keep some aspect confidential of the

identity of the people involved in transactions. So what is there hope to keep confidential in this context? So in this case, for example, you want to enable confidential transactions, so there are different essentially types of data that you want to keep private or confidential. And you can utilize different technologies, including zero knowledge proofs, technologies, including zero knowledge proofs, and also secure computing and techniques to hide the right who is making the transactions to home and the transaction amount. And in our case, also, we can enable like confidential smart contracts. So that you don't know the data and the execution

of the smart contract and so on. And we actually are combining these different technologies and to going back to the earlier discussion we had enabling like ownership of data and privacy of data and so on. So at OSUS Labs, we are actually building what we call a platform for responsible data economy to actually combine these different technologies together

to enable secure and privacy-preserving computation and also using the library to help provide immutable log of users' ownership to their data. And the policies they want the data to adhere to, the usage of the data to adhere to, and also how the data has been utilized. So all these together can build, we call it, distributed secure computing fabric that helps to enable a more responsible data economy. So, lots of things together.

Yeah, wow, that was eloquent. Okay, you're involved in so much amazing work that we'll never be able to get to, but I have to ask at least briefly about program synthesis, which at least in the philosophical sense captures much of the dreams of what's possible on computer science and the artificial intelligence. First let me ask, what is program synthesis and can neural networks be used to learn programs from data? So can this be learned? Some aspect of the synthesis can be learned. So program synthesis is about teaching computers to write code to program.

And I think that's one of our ultimate dreams or goals. I think Andrewson talked about software eating the world. So I say once we teach computers to write software to write programs, then I guess computers will be eating the world to buy transitivity. Yeah, exactly. So yeah, it's, and also for me, actually, when I, you know, shifted from security to more AI and machine learning,

program synthesis is, program synthesis and adversarial machine learning, these are the two fields that I particularly focus on. Like program synthesis is one of the first questions that I actually started investigating. Just as a question, I guess from the security side, there's that you're looking for holes in programs. So I at least see small connection,

but why, what was your interest for program synthesis as, because it's such a fascinating, such a big, such a hard problem in the general case. Why program synthesis? So the reason for that is actually when I shift this my focus from security into AI and machine learning, actually one of my main motivation at the time is that even though I have been doing a lot of work in security and privacy, but I have always been fascinated about building intelligent machines. And that was really my main motivation

to spend more time in AI machine learning is that I really want to figure out how we can build intelligent machines. And to help us towards that goal, program synthesis is really one of, I would say, the best domain to work on. I actually call it a program synthesis, it's like the perfect playground for building intelligent machines and for artificial engineering intelligence. Yeah.

Well, it's also, in that sense, that's just a playground, I guess, it's the ultimate test of intelligence. Yes. I think if you can generate neural networks, you can learn good functions, and they can help you out in classification tasks, but to be able to write programs. That's the epitome from the machine side. That's the same as passing the touring test and natural language with programs. It's able to express complicated ideas, to reason through ideas, and boil them down to algorithms.

Yes, exactly, it's incredible. So can this be learned how far are we? Is there hope? What are the open challenges? Yeah, very good questions. We are still at an early stage. But already, I think we have seen a lot of progress. I mean, definitely, we have existence proof.

Just like humans can write programs. So there's no reason why computers cannot write programs. So I think that's definitely an achievable goal. It's just how long it takes. And then, and even today, we actually have, you know, the program synthesis community, especially the program synthesis

while learning, how we call it, neuro program synthesis community. It's still very small. But the community has been growing, and we have seen a program synthesis community. It's still very small, but the community has been growing and we have seen a lot of progress. And in limited domains, I think actually program synthesis is ripe for real world applications. So actually it was quite amazing.

I was giving a talk. So here is a rework conference? Re-work deep learning summit. I actually, so I give another talk at the previous rework conference in deep reinforcement learning. And then I actually met someone from a startup, the CEO of the startup.

And when he saw my name, he recognized it. And then he actually said, one of our people is actually had, they have put, had actually become a key product in the startup and that was, program synthesis in that particular case was natural language translation, translating natural language description into SQL queries. Oh, wow, that direction. Okay.

Right. So, in program synthesis, in limited domains, in well-specified domains, actually already we can see really great progress and applicability in the real world. So domains like, as an example, you said natural language being able to express something through just normal language and then convert it into a database SQL SQL query. Right. And that's how how solves the problem is that because it seems like a really hard problem. Okay in limited domains actually it can work pretty well

and now this is also a very active domain of research at the time I think when he saw our paper at the time we were the state of the arts yeah on that task and since then actually now there has been more work and with even more sophisticated data sets. But I think I wouldn't be surprised that more of this type of technology really gets into the real world. That's exciting.

In the near term. Being able to learn in the space of programs is super exciting. I still, I'm still skeptical because I think it's a really hard problem. I think in terms of the you asked about open challenges, I think the domain is full of challenges. And in particular also we want to see how we should measure the progress in the space. And I would say mainly three main, I would say metrics. So one is a complexity of the program that we can synthesize.

And that we actually have clear measures and just look at the past publications. And even like for example, I was at the recent New Reps conference. Now there is actually a very sizable session dedicated to program synthesis, which is... Oh, even neuro-programs synthesis. Right, right, right, right, which is great. That's great.

And we continue to see the increase in... What does sizable mean? I like the word sizable. It's five people. It's still small community, but they're just growing. And they will all win touring awards one day. I like it. Right. So we can see increase in the complexity of the program is that we can synthesize. So, is it the complexity of the actual text of the program or the running time complexity? Which complexity are we?

How? The complexity of the task to be synthesized and the complexity of the actual synthesized programs. So, the lines of code even, for example. OK, I got you, but it's not the theoretical, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, complex programs, bigger and bigger programs. So we want to see that we want to increase the complexity. I have to think through because I thought of complexity is you want to be able to accomplish the same task with a simpler and simpler program. No, we are not doing that. It's more about how complex a task we can synthesize it being able to synthesize programs learn

them for more and more difficult. Right so for example initially our first work in program since this is since this is what we translate natural language to subscription into really simple programs called if TTT if this then that's so given a trigger condition what is the action you should take so that program is super simple. You just identify the trigger conditions and the action. And then they say, with SQL queries, it gets more complex. And then also we started to synthesize programs with loops. And if you can synthesize recursion, it's all over. Right. Actually, one of our works actually is learning

recursive neural programs. But anyway, so that's one complexity, and the other one is generalization. Like, one way of training, I want to learn a program synthesizer, in this case, a neural program to synthesize programs, then you wanted to generalize. For a large number of inputs. Right. To be able to generalize to

previously unseen inputs. Got it. So some of the work we did earlier, learning recursive neural programs, actually show that recursion actually is important to learn. And if you have recursion, then for a certain set of tasks, we can actually show that you can actually have perfect generalization.

So that's one of the best people awards that I clear earlier. So that's one of the best people awards I clear earlier. So that's one example of we want to learn these neural programs that can generalize better. But that works for certain tasks, certain domains. And there's question how we can essentially develop more techniques that can have generalization for wider set of domains, and so on.

So that's another area. Then the third challenge I think is not just for programming synthesis, it's also cutting across other fields in machine learning and also including like deep learning, first learning in particular, is that this adaptation is that we want to be able to learn from the past and tasks and training and so on to be able to solve new tasks. So for example, in program synthesis today, we still are working in the setting where given

particular tasks, we train the model and to solve this particular task. But that's not how humans work. The whole point is we train a human and you can then program to solve new tasks. Right, exactly. And just like in deep re-first meeting, we don't want to just train agents to play a particular game. Either it's Atari or it's Go or whatever.

We want to train these agents that can essentially extract knowledge from the past, the new experience to be able to adapt to new tasks and solve new tasks. And I think this is particularly important for program synthesis. Yeah, that's the whole point. That's the whole dream of programs. This is your learning a tool that can solve new problems. Right.

Exactly. And I think that's a particular domain that as a community, we need to put more emphasis on. And I hope that we can make more progress there as well. Awesome. There's a lot more to talk about. Let me ask that you also had a very interesting, and we talked about rich representations. You had a rich life journey. He did your bachelors in China and your masters in PhD in the United States, CMU in Berkeley.

Are there interesting differences? I told you I'm Russian. I think there's a lot of interesting differences between Russian and the United States. Are there in your eyes interesting differences between the two cultures from the silly romantic notion of the spirit of the people to the more practical notion of how research is conducted that you find interesting or useful in your own work of having experienced both.

That's a good question. I think so I started in China for my undergraduate years and that was more than 20 years ago. So it's been a long time. Is there at those of that time in you? Yes, actually it's interesting. I think even more so maybe something that's even be more different for my experience than a lot of computer science, researchers and practitioners is that so for my undergrad I actually started physics. Nice, very nice. And then I switched to computer science

in graduate school. What happened? What happened? What happened? Was there, was there, is there another possible universe where you could have become a theoretical physicist a Caltech or something like that. That's very possible some of my undergrad classmates then they later on studied physics

about their PhD in physics from these schools from yeah from top physics physical issues from the physical programs. So you switched to, I mean, from that experience of doing physics and your bachelor's, what makes you decide to switch to computer science and computer science at arguably the best university, one of the best universities in the world for computer science with Carnegie Mellon, especially for the grad school and so on. So what second only to MIT just kidding? Okay. I have to throw that in there. No, what was the choice like and what was the move to the United States like?

What was that whole transition? And if you remember, if there's still echoes of some of the spirit of the people of China in you, in New York. Right, right, yes. It's like three questions in one. Yes, I know. I'm sorry. No, that's okay.

So yes, I guess it's the first transition from physics to computer science. Yes. So when I first came to the United States, I was actually in the physics PhD program at Cornell. Yeah. I was there for one year and then I switched to computer science and then I was in the PhD program at Carnegie Mellon. So, okay, so the reasons for switching, so one thing, so that's why I also mentioned that about the difference in backgrounds about having studied physics first in my undergrad. I actually really did enjoy my undergrad time and education in physics. I think that

actually really helped me in my future work in computer science. Actually, even for machine learning, a lot of machine learning stuff, the core machine learning methods many of them actually came from physics. That is cool. For honest, most of everything came from physics. So when I started physics, I was, I think, I was really attracted to physics. It was really beautiful. And I actually thought physics is the language of nature.

And I actually really remember, like, one moment, in my undergrad, I didn't undergrad in Qinghua, and I used to study in the library. And I clearly remember, one day, I was sitting in a library, and I was writing on my notes and so on. And I got so excited that I realized that really just from a few simple axioms, a few simple laws, I can derive so much. It's almost like I can derive the rest of the world. Yeah, the rest of the universe. Yes, yes, so that was like amazing.

Do you think you have you ever seen or do you think you can rediscover that kind of power and beauty and computer science in the world that you're in? So that's very interesting. So that gets to the translation from physics to computer science. It's quite different for physics in in classical actually things changed. So one is I started to realize that when I started doing research in physics at the time I was doing theoretical physics.

And a lot of it, you still have the beauty, but it's very different. So I actually actually do a lot of simulations. So essentially I was actually writing in some kids writing fortune code. That old fortune, yeah. To actually do simulations and so on that was not exactly what I was doing. I enjoy it doing. And also, at the time from talking with the senior students in the program, I realized many of the students actually were going off to like Wall Street and so on. So, and I've always been interested in computer science. And I actually essentially taught myself actually essentially taught myself the C programming. Program. And so on.

Which one? In college. In college somewhere. In a summer. For fun. Physics major, learning to do C programming. Beautiful. Actually, it's interesting, you know, in physics,

at the time, I think now the program probably has changed. But at the time, really think now the program private has changed, but at the time, really the only class we had in, in the related to computer science education was introduction to, I forgot, to computer science or computing and Fortune 77. There's a lot of people that still use Fortran. I'm actually, if you're a program out there, I'm looking for an expert to talk to about Fortran.

They seem to, there's not many, but there's still a lot of people that still use Fortran and still a lot of people use Cobalt. So, by the way, so, so then I realized, instead of just doing programming for doing simulations and so on, then I made as well just change to computer science. And also one thing I really liked, and that's a key difference between the two, is in computer science, it's so much easier to realize your ideas. If you have an idea, you write it up, you code it up, and then you can see it's actually, right? You can bring it to life quick. Bring it to life. Where is in physics? If you have a theory, you have to wait for the experimentalist to do the experiments

and to confirm the theory. And things just take so much longer. And also the reason I, in physics, I decided to do theoretical physics was because I had my experience with experimental physics. First, you have to fix the equipment. Theoretical physics was because I had my experience with experimental physics. First, you have to fix the equipment. Yeah, yeah. You said most of the time fixing the equipment first.

Super expensive equipment. There's a lot of... Yeah, you have to collaborate with a lot of people. Takes a long time. This takes really... Right, much longer. Yeah, it's messy. So I decided to switch to computer science.

And one thing I think... Maybe people realized is that for people who study physics, actually, it's very easy for physicists to change to do something else. Yes. I think physics provides a really good training. And yeah, so actually it was very easy to switch to computer science. But one thing to impact through your earlier question. So one thing I actually did realize, there is a big difference between computer science and physics where physics you can derive

the whole universe from just the field simple laws. And computer science, given that a lot of it is defined by humans, the systems that define by humans. And it's artificial. Like essentially you create a lot of these artifacts and so on. It's not quite the same. You don't derive the computer systems with just a few simple laws.

You actually have to see these historical reasons why a system is built and designed one way versus the other. There's a lot more complexity, less elegant simplicity of E equals mc squared that kind of reduces everything down to beautiful fundamental equations. But what about the move from China into the United States? Is there anything that's still staging you that contributes to your work, the fact that you grew up in another culture? So yes, I think especially back then it's very different from now. So you know, now they actually, I see these students coming from China. And even China and even undergrad actually they speak fluent English.

It was just, you know, like amazing. And they have already understood so much of the culture in the US and so on. And it was to you was all foreign. It was was it was a very different time at the time actually Even we didn't even have easy access to email right now to mention about the web Yeah, I remember I had to go to you know specific like you know Privileged sever rooms to use email Right and hence We at the time we had much less knowledge about the Western world.

And actually, at the time, I didn't know, actually, in the US, West Coast, whether it's much better than the East Coast. Things like that, actually, it's very interesting. Yeah, but now it's so different at the time. I would say there's also a bigger culture difference because there's so much less opportunity for shared information. So it's such a different time and world.

So let me ask maybe a sensitive question. I'm not sure, but I think you're not in similar positions as I've been here for already 20 years Let me ask, maybe a sensitive question, I'm not sure, but I think you're not in a similar position since I've been here for already 20 years as well and looking at Russia from my perspective and you looking at China. In some ways it's a very distant place because it's changed a lot, but in some ways you still have echoes, you have still of knowledge of that place. The question is, China is doing a lot of incredible work in AI.

Do you see, please tell me there's an optimistic picture you see where the United States and China can collaborate and sort of grow together in the development of AI towards, there's different values in terms of the role of government, so on, of ethical, transparent, secure systems. We see it differently in the United States a little bit than China, but we're still trying to work it out.

Do you see the two countries being able to successfully collaborate and work in a healthy way without sort of fighting and making it an AI arms race kind of situation? Yeah, I believe so. I think science has no border and the advancement of technology helps everyone, helps the whole world. And so I certainly hope that the two countries will collaborate. And I certainly believe so. Do you have any reason to believe so, except being an optimist?

So first again, like I said, science has no borders. And especially in... Science doesn't know borders. Right. And you believe that will, you know, in the former Soviet Union during the Cold War. So that's the other point I was going to mention is that especially in academic research, everything is public. Like we write papers, we open source codes and all of this is in the public domain. It doesn't matter whether the person is in the US,

in China or some other parts of the world. They can go on archive and look at the latest research and results. So that openness gives you hope. Yes. Me too. And that's also how as a world we make progress the best. So, Paul just for the romanticized question, but looking back, what would you say was the most transformative moment in your life that Maybe made you fall in love with computer science. You said physics. You remember There's a moment where you thought you could derive the entirety of the universe Was there a moment that you really fell in love with the work you do now from security to machine learning to program synthesis so maybe as I mentioned actually in college, I once summarize the time I self programming in C. Yes.

I just read the book. And I know. Don't tell me you found a lot of computer science by programming in C. Remember I mentioned one of the draws for me to communicate science is how easy it is to realize the ideas. So once I read the book, start, like tell myself how to programming see, immediately what did I do?

Like I programmed two games. Once just simple, it's a go game, like it supports, you can move the stones and so on. The other one actually programmed the game, that's like a 3D Tetris. It was a two-not-a-two-not-a-three-hat-game to play. Because you saw just the standard 2D Tetris, it's a 3D thing. But I can't realize, wow, I just had this idea to try it out. And then you can just do it.

So that's when I realized, wow, this is amazing. Yeah, you can create yourself. Yes, yeah, ideas. Exactly. From nothing to something that's actually out in the real world. Let me ask you a silly question or maybe the ultimate question, what is to you the meaning of life? What gives your life meaning purpose, fulfillment, happiness, joy? Okay, these are two different questions. Very different, yeah. As you should know,

you ask this question, maybe this question is probably the question that has follows me and follow my life the most. Have you discovered anything, any satisfactory answer for yourself? Is there something you've arrived at? You know, there's a moment, I've talked to a few people who have faced, for example, a cancer diagnosis or faced their own mortality, and that seems to change their view of them.

It seems to be a catalyst for them removing most of the crap that, of seeing that most of what they've been doing is not that important, and reducing it and to saying like here's actually the few things that really give me give meaning. Mortality is a really powerful catalyst for that it seems like facing mortality whether it's your parents dying or somebody close to you dying or facing your own death for what a reason or cancer and so on. Right so yeah so in my own case, I didn't need to face mortality too. So try to, you know, to ask that question. Yes. And I think there are a couple things.

So one is, like, who should be defining the meaning of your life? Right. Is there some kind of even greater things than you who should define the meaning of your life. Right? Is there some kind of even greater things than you who should define the meaning of your life? So for example, when people say that searching the meaning for your life, is is there some, is there some outside voice, or is there something, you know, who tells you, so people talk about, oh, this is what you have been born to do. This is your destiny. That's one question, who gets to define the meaning of your life?

Should you be finding some other thing, some other factor to define this for you? Or is something actually, it's just entirely what you define yourself and it can be very arbitrary? Yeah, so an inner voice or an outer voice, whether it's, it could be spiritual religious too with God or some other components of the environment outside of you

or just your own voice. Do you have an answer there? So okay so that's I haven't answered. Yeah. And through you know the long period of time of thinking and searching, even searching through outside's uh right, you know voices are factors outside factors outside of me. So that I have, and I've come to the conclusion and realization that it's you yourself that defines a minion of life. Yeah, that's a big burden, no, isn't it? Yes, yes, yes, yes, yes, yes, no, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, yes, even makes sense. Absolutely, and you said it somehow distinct from happiness. So meaning is something much deeper than just any kind of emotional and any kind of contentment or joy or whatever it might be much deeper. And then you have to ask what is deeper than that? What is what is there at all? And then

the question starts being silly. Right, and also you can say it's deeper, but you can also say it's a shell or a depending on how people want to define the meaning of their life. So for example, most people don't even think about this question, then the meaning of life to them doesn't really measure that much. And also whether knowing the meaning of life,

whether it actually helps your life to be better, or whether it helps your life to be happier, these actually are open questions. It's not. Of course, most questions are open. I tend to think that just asking the question, as you mentioned, as you've done for a long time, is the only, that there is no answer,

and asking the question is a really good exercise. I have this for me personally, I've had a kind of feeling that creation is, for me, has been very fulfilling. And it seems like my meaning has been to create. And I'm not sure what that is. I don't have a single line of kids. I'd love to have kids, but I also sounds creepy, but I also see sort of, you said, see programs. I see programs as little creations. I see robots as little creations. I think those are, those bring, and then ideas, theorums and our creations and those, somehow intrinsically, like you said, bring me joy.

I think they do to a lot of at least scientists, but I think they do to a lot of people. So that, to me, if I had to force the answer to that, I would say creating new things yourself. For you, for me. For me, I don't know. Like you said, it keeps changing. Is there some answer that? As some people, they can, I think, they may say, is experience. Right? Like their many of life, they just want to experience to the riches and fullest they can. And a lot of people do take that path. Yeah. Seeing life is actually a collection of moments and then trying to make the richest possible sets fill those moments with the richest possible experiences. Right. And for me, I think certainly we do share a lot of similarities here. So creation is also really important for me, even from the things I've already talked about. even like writing papers and these are our creations as well. And I have not quite thought whether that is really the meaning of my life.

Like in a sense also there may be like what kind of things should you create there, so many different things that you could create. And also you can say what another view is maybe growth is And also you can say another view is maybe growth is related but different from experience. Growth is also maybe type of meaning of life. It's just you try to grow every day, try to be a better self every day. And also ultimately we are here as part of the overall evolution. We are here. It's part of the overall evolution. The world is evolving. Isn't it funny that the growth seems to be the more important thing than the thing you're growing towards? It's not the goal. It's the journey to it. It's almost when you submit a paper. There's a depressing element to it. Not to submit a paper, there's a sort of depressing element to it, not

to submit a paper, but when that whole project is over, I mean, there's a gratitude, there's a celebration and so on, but you're usually immediately looking for the next thing for the next step, right? It's not that saddest, the end of it is not the satisfaction, it's the hardship that challenge you have to overcome, the growth of the process. It's somehow probably deeply within us, the same thing that drove that drives the evolutionary process, it's somehow within us, with everything the way we see the world, since you're thinking about these, so you're still in search of an answer. I mean, yes and no. In a sense, I think for people who really dedicate time to search for the answer to ask

a question, what is the meaning of life? It does not necessarily bring your happiness. It's a question. We can say it right, whether it's a well-defined question and and and but on the other hand, given that You get to answer yourself. You can define yourself Then sure that I can just you know give it an answer and in that sense, yes, it can help Like as like we discussed, if you say,

oh, then my meaning of life is to create, or to grow. Then yes, then I think they can help. But how do you know that that is really the meaning of life, or the meaning of your life? It's like there's no way for you to really answer the question. Sure, but something about that certainty is liberating. So it might be an illusion. You don't mean it might not really know.

You might be just convincing yourself falsely. But being sure that that's the meaning, there's something liberating in that. There's something freeing and knowing this is your purpose. So you can fully give yourself to that without, you know, for a long time, you know, I thought, like, isn't it all relative? Like why? What's, how do we even know what's good and what's evil? Like, isn't everything just relative?

Like how do we know, you know, the question of meaning is ultimately the question of, why do anything? Why is anything good or bad? Why is anything valuable? Exactly. But then you start to... I think just like you said, I think it's a really useful question to ask. But if you ask it for too long and too aggressively... I mean, that would be so productive.

They have not be productive and not just for traditionally, societally, defined success, but also for happiness. It seems like asking the question about the meaning of life is like a trap. We're destined to be asking. We're destined to look up to the stars and ask these big, why questions will never be able to answer, but we shouldn't get lost in them. I think that's probably the,

that's at least the less than I picked up so far. On that topic. Let me just add one more thing. So it's interesting. Um, so actually, so sometimes, yes, it can help you, um, to focus. So when I, my focus more from security to AI and machine learning, at the time,

the actually one of the main reasons I did that was because at the time, I thought, the meaning of my life and the purpose of my life is to build intelligent machines. And that's and then your inner voice said that this is the right. This is the right journey to take the building intelligent machines and that you actually fully realize. You took a really legitimate big step to become one of the world-class researchers to actually make it To actually go down that journey. Yeah, that's profound That's profound. I don't think there's a better way to end a conversation than talking for

For a while about the meaning of life Don is a huge honor to talk to you. Thank you so much for talking today Thank you. Thank you Thanks much for talking to me. Thank you, thank you. Thanks for listening to this conversation with Don Song, and thank you to our presenting sponsor, Cash App. Please consider supporting the podcast by Don Woting Cash App and using code Lex Podcast. If you enjoy this podcast, subscribe on YouTube,

review it with five stars on Apple Podcasts, support it on Patreon, or simply connect with me on Twitter, Alex Friedman. And now let me leave you with some words about hacking from the great Steve Wozniak. A lot of hacking is playing with other people, you know, getting them to do strange things. Thank you for listening and hope to see you next time. you