OpenAI, reinforcement learning, robots, safety

We met up with Wojciech Zaremba at the O’Reilly AI conference in SF. He took some time to talk to us about some of his recent research related to reinforcement learning and robots. We also discussed AI safety and the hype around OpenAI. Sponsors: Fastly https://www.fastly.com/?utm_source=changelog&utm_medium=podcast&utm_campaign=changelog-sponsorship – Our bandwidth partner. Fastly powers fast, secure, and scalable digital experiences. Move beyond your content delivery network to their powerful edge cloud platform. Learn more at fastly.com https://www.fastly.com/?utm_source=changelog&utm_medium=podcast&utm_campaign=changelog-sponsorship. Rollbar https://rollbar.com/changelog – We catch our errors before our users do because of Rollbar. Resolve errors in minutes, and deploy your code with confidence. Learn more at rollbar.com/changelog https://rollbar.com/changelog. Linode https://linode.com/changelog – Our cloud server of choice. Deploy a fast, efficient, native SSD cloud server for only $5/month. Get 4 months free using the code changelog2018. Start your server - head to linode.com/changelog https://linode.com/changelogFeaturing: Wojciech Zaremba – Website http://wojzaremba.com/, GitHub https://github.com/wojzaremba, X https://x.com/woj_zarembaDaniel Whitenack – Website https://www.datadan.io/, GitHub https://github.com/dwhitena, X https://x.com/dwhitenaShow Notes: Woj’s talk at O’Reilly AI https://conferences.oreilly.com/artificial-intelligence/ai-ca/public/schedule/detail/71237OpenAI https://openai.com/OpenAI blog https://blog.openai.com/The Dota playing team of AIs from OpenAI https://blog.openai.com/openai-five/Microsoft Tay Chatbot https://en.wikipedia.org/wiki/Tay_(bot)Reinforcement learning Wikipedia https://en.wikipedia.org/wiki/Reinforcement_learningBerkeley course CS294-112 on Reinforcement learning http://rail.eecs.berkeley.edu/deeprlcourse/. Includes pdf notes and YouTube videos https://www.youtube.com/playlist?list=PLkFD6_40KJIxJMR-j5A1mkxK26gh_qg37 of the lectures. Books “Deep Learning” by Ian Goodfellow, Yoshua Bengio and Aaron Courville https://www.amazon.com/dp/0262035618“Reinforcement Learning: An Introduction” by Richard S. Sutton and Andrew G. Barto https://www.amazon.com/dp/0262039249Upcoming Events: Register for upcoming webinars here https://practicalai.fm/webinars!

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

• 0:00Speaker 00

  Baiting for change log is provided by Fastly. Learn more at fastly.com. We move fast and fix things here at change law because of roll bar. Check them out at rollbar.com and we're hosted on Lino servers. Head to linode.com slash change log.

• 0:22Speaker 00

  Welcome to Practical AI, a weekly podcast about making artificial intelligence, practical, productive and accessible to everyone. This is where conversations around AI, machine learning and data science happen. Join the community and slay with us around various topics of the show at change.com slash community. Follow us on Twitter, we're at Practical AI FM. And now into the show.

• 0:47Daniel Whitenack?

  So Wocek, thanks for joining us in Practical AI. I know you've got a busy schedule today at O'Reilly AI. So thanks for taking time.

• 0:55Speaker 02

  Thank you for the invite.

• 0:56Daniel Whitenack?

  Yeah, and you probably don't need much introduction, but for those that don't know, Wocek is one of the co-founders of OpenAI and he's going to be talking today at O'Reilly AI about robotics and deep learning. Is that right?

• 1:09Speaker 02

  That's correct.

• 1:10Daniel Whitenack?

  Yeah, awesome. It'd be great if you could give, give the audience just a little bit of an idea of your, your background and what you're working on now.

• 1:16Speaker 02

  Cool. So I mean, I spent some time at Google Brain. I spent some time at Facebook I research have finished PhD in the topic of deep learning. Even I spent many years ago sometime at Nvidia before actually deep learning was the thing.

• 1:35Speaker 02

  So it's kind of somewhat coincidental that all these things turn out to be extremely relevant. As you have said, Daniel, I'm one of the co-founders of OpenAI and goal of OpenAI is quite ambitious is to figure out the way to build general's artificial intelligence or to be more exact, how to build it in the way that say it's safe in a sense, we can control it or let's say figure out, let's say from political perspective, how to deploy it in the way that it is beneficial. To humanities as a whole. Our approach is more or less as follows. We see various limitations of current systems.

• 2:19Speaker 02

  And we think, what's the best way to, what are the goals that we should attempt? Such that if we solve them, it becomes clear that we live at these limitations. And particularly, so, so, so, so, so. I mean, there, let's say, somewhat several kind of internal approaches. How we go about the problems. But in case of what I'm doing, the motivation is as follows. So, we know as of today that you can take extremely complicated computer games and machine can become superhuman in it. But the main criticism is, yeah, but that's just a game. It is kind of, let's say, unrelated to reality, confined into, let's say, the realm of the computer.

• 3:04Speaker 02

  And kind of sounds straightforward that we should be able to pull it off and apply it into real world. But turns out that many people tried and haven't succeeded for a while. When it comes to very, to let's say, more complicated robots and so on. Yeah, I mean, I will let you ask me more questions and then I can tell you more.

• 3:25Daniel Whitenack?

  Sounds great, yeah, I appreciate that. I was wondering before going forward, if you could just maybe, I know you mentioned, you know, OpenAI kind of going after general AI or general intelligence. I was wondering if you could kind of break that down for the listeners in terms of how, how that might be different from the research or the methods that other people are going after.

• 3:46Speaker 02

  So, it is clear to me that AI, AGI, consists of more or less three components. So, I mean, you have to be able to inject gigantic amount of compute, most likely. You need to invent some new algorithms. And AI needs some kind of a data or environment in which it lives.

• 4:11Speaker 02

  So there are three main components that you have to, let's say, attack if you want just the capabilities of AGI. So I think that OpenAI is extremely well suited to pursue this goal because of a few reasons. One, we are not bounded by a product. And so often when we are building a product, actually quite huge fraction of the work has to do with figuring out the niche, or let's say, target, marketing, let's say, packaging and so on. And it's clear that there are various problems that you can attempt that have very small

• 4:55Speaker 02

  economical value, but it's actually very clear that we are making actual progress. I mean, if you, for instance, be able to, you know, train the system that can solve Riemann hypothesis, like a very

• 5:07Daniel Whitenack?

  And could you just give a quick explanation of what that is?

• 5:11Speaker 02

  Okay, yeah. So let's say, I just mentioned Riemann hypothesis is like an unresolved mathematical problem. So if, if system would be able to, you know, have such incredible reasoning, that would indicate that you lifted one more restriction, maybe, you know, the restriction that the neural networks do not reason much, has not that much of an economical value. I mean, maybe people doing math would get upset, would unemployed another group of people. But in a sense, that's not the product out of which you are making tons of money. So I'm just saying that, so I just compared, let's say, our approach to the building conventional product, I can also compare it to, let's say, academic labs.

• 5:56Speaker 02

  So in a sense, the most of the academic labs, they are, they construct of the, it's like an endeavor of small number of individuals. And I would say, among these three components that I mentioned, let's say, compute algorithms and environment data, it's actually focuses mostly on the algorithm, which is completely fine. That's one of the components, that's the one that the They are well suited to actually pursue, but it seems to me that you have to be able to pursue actually all three components to make a progress. You need, you need to be able to, you know, to focus the entire data center ten miles by ten miles.

• 6:38Speaker 02

  That, you know, consumes millions of watts with some number of, with some number of algorithms and let's say environments to actually achieve it. So in a sense, you need resources beyond actually, let's say, researchers. You need also all sorts of the talented engineers, infrastructure engineers and so on and so forth. So I guess that's the main difference.

• 7:01Daniel Whitenack?

  Yeah, that, that helps a lot. And you kind of alluded to the fact that, you know, in your background, you had experience with these different technologies that, that kind of coincidentally converged in this new hype around AI. And you co-founded OpenAI. And I want to get into the content of your talk a little bit later, but I'm wondering if you could just comment on what you see as some of the advantages and disadvantages of kind of founding a company that is really at the, at the center of this hype around AI. And what your, what your thoughts on, on that are and how you see the field moving forward in the midst of all this hype. I think, you know, NIPS, the NIPS conference sold out in like fifteen minutes.

• 7:44Daniel Whitenack?

  It's like faster than a Taylor Swift conference or something.

• 7:48Speaker 02

  People are comparing the sales of tickets for NIPS to sales of tickets to Birdingman.

• 7:53Daniel Whitenack?

  Yeah, exactly. Yeah. So, and, and OpenAI is really at the, at the kind of center of that hype. Have you seen that kind of advantages and disadvantages to, to that? Or in terms of the progress that you're going after?

• 8:06Speaker 02

  So, I mean, this extremely important to, let's say, not over promise, deliver and so on. I mean, otherwise, over some period of time, it's gonna bite you in the ass, simple as that. The fact of being well-recognized organization definitely helps to hire incredibly brilliant people, I would say. I'm feeling pretty much like I'm, I'm, I'm feeling. Extremely honored that say people around they are on you know various axis pretty much everybody is on some axis better than me. I'm feeling honored that these folks want to work with me, so.

• 8:44Daniel Whitenack?

  Awesome. Well, uh, yeah, thank you for that. And I definitely uh respect a lot of a lot of the things you're doing and um and so looking forward to seeing seeing what happens from here. But now I want to kind of talk a little bit. You're going to be talking about deep reinforcement learning for robotics at uh at OpenAI. Give us just a brief sketch of of what you'll be talking about.

• 9:04Speaker 02

  Cool. So as I mentioned at first, uh, it's actually the case that, you know, reinforcement learning is a sense quite generic incredible paradigm. It's effectively saying that you can take any quantity that you can measure and you can if you can even remotely influence this quantity, you can learn how to, let's say, maximize it. And the quantity can measure, you know, success on the task or so. So this paradigm was used to be used in various computer games, was used to beat humans in Go. It was used to uh in the uh game of Dota to get to the professional level.

• 9:43Speaker 02

  The paradigm has several shortcomings. So in a sense, as of today, it requires insane amount of data. So the the the the the rates are of tens of thousands of years of experience, virtual experience. So let's say in case of Go or in case of Dota, agent is playing for so long for for for for millennia. That's the requirement if you really want to get top pro or superhuman performance. In case of robotics, it seems to be very hard to apply the same paradigm to the physical robots.

• 10:21Speaker 02

  I mean, it is possible. You can try also to, you know, go uh through the path of decreasing number of uh interactions. So I would say that's one of the meaningful directions to just minimize the the try to improve algorithms to minimize number of uh interactions.

• 10:41Speaker 02

  One of the approaches was, let's say, done at Google to, let's say, build form of arms to pick an object. But still, there are few fundamental issues. So as long as, once you are moving to the real world, it's actually not that easy to even get diversity of data. I mean, you cannot take your arm and go to a waterfall. Other part of the assumption is that you have to be able to reset and taffing from scratch. And also, say, once you are moving some objects, it can fall over and then you have to build some contraction mechanism to bring it back.

• 11:22Speaker 02

  So I started myself to think that maybe the paradigm of reinforcement learning, instead of being close to actually what is happening when the human is learning, is actually closer to the what is happening during evolution-like learning, that you have a really gigantic population, it has huge number of the interactions, rather than like a human kind of thinking through what truly happened and what should be the outcomes. So then it is also kind of natural to me that in case of evolution-like learning, you have, or let's say, in case of human learning, there are essentially two stages.

• 12:05Speaker 02

  There is a stage that takes this gigantic, gigantic amount of data, which is evolution. It actually, this evolution is reinforcement learning, like gonna survive or die. And this stage is powerful enough to create our brains that can then rapidly learn. So I was thinking, you know, it's actually not such a bad thing if we can learn in simulation to slightly add to, okay, almost majority of the task. And then in reality, there is quite rapid adaptation. So that's what we did. The interesting thing is, when we, so the task is, we took robotics hand and we are reorienting objects.

• 12:51Speaker 02

  And let's say speak in a second about the difficulty of this task. But solving the task itself takes something maybe like three years of virtual experience. But then to get to the capabilities that allow you for the transfer, that takes another ninety seven years.

• 13:07Daniel Whitenack?

  Oh, wow. Yeah. So the generalization to to create a model that that creates this kind of or a response to adaptations takes the longest amount of time.

• 13:18Speaker 02

  Correct.

• 13:18Daniel Whitenack?

  Yeah, awesome. And just to kind of clarify a few things. So like when you're doing the this virtual simulation, we've talked already on the podcast about about where deep learning fits into the spectrum of AI techniques. And we've talked about even like masks, our CNN for in robotics context. But we haven't really dove into reinforcement learning yet. So I was wondering if you could just kind of give us a brief introduction to when you're doing the virtual learning and these kind of two stages. So you mentioned the first stage where it's kind of learning a task and then the second stage where you're attempting to make it more more adaptive. What is the process that you're actually doing there?

• 14:00Speaker 02

  So let me first describe what reinforcement learning is.

• 14:04Daniel Whitenack?

  That'd be great.

• 14:04Speaker 02

  So reinforcement learning, say, framework of teaching an agent to maximize amount of reward. You can think about it a little bit like training a dog with a treats. So when the dog is doing good stuff, you're giving the dog treats and then it does more of the stuff that you want. And that's the way, more or less, how you train the computer to become best in all these games. So it's completely up to you when you define that you're gonna give a treat.

• 14:36Daniel Whitenack?

  Ah, yeah. So there's a there's a feedback that happens with some.

• 14:40Speaker 02

  Correct. So agent itself. Effectively, some network that consumes observations. You can think as of an analogy to a human. It's like an input to eyes, ears, nose, touch and so on. And the network is supposed to produce actions. So in our case, it would be an electric signal. The nerves to decide how to move, let's say limbs or so. And the network itself attempts to maximize reward. The system has a chance only to be successful if from time to time it gets a reward. So that's a kind of, let's say, let me a little bit downplay reinforcement learning.

• 15:24Speaker 02

  So the situations that wouldn't work is if you give a treat to a dog after it lands on Mars. Okay, it's like it will just never happen. It has to, from time to time, it starts with, it starts at first with very random strategy. And then gradually it attempts to get more and more of the treats. So that's pretty much the paradigm of reinforcement learning.

• 15:47Daniel Whitenack?

  And that's what you would be doing kind of in that, in that first stage where you're teaching, teaching a specific task. Is that right?

• 15:54Speaker 02

  Correct. So let me briefly describe the task. So we haven't knew much about the robotics over a year ago. We went to robotics conference. We asked people, what are the things that are impossible or very hard to do in the classical robotics? And people are saying, anytime when you have a large number of degrees of freedom. It's very hard to control when there is a lot of interactions. When you touch many things in many places simultaneously, that's also very hard to model. And there is, when, it is way easier when you are in the open space and you are not touching anything. Okay. Or if the problem somehow can be simplified to one or two dimensional problems. And there are some, let's say, closed form solutions.

• 16:40Speaker 02

  But in case of robotics can, robotics can has, let's say, the one that we both has 24 degrees of freedom. We also kind of knew that the task is solvable because human can solve it. So we, we, we wanted to have a hope for the success. Yeah, and task is you take an object, in our case we demonstrated it on two different objects, which is one is a block, one is some octagonal prism, and task is move it around to their new desired position.

• 17:11Daniel Whitenack?

  Like in a robotic hand?

• 17:12Speaker 02

  In their robotic hand.

• 17:13Daniel Whitenack?

  Yep.

• 17:14Speaker 02

  So, as I said, we were able to train it actually a while ago already in the simulation to achieve it, but then during deployment it didn't work at all, like a literal is not at all, despite easily being able to solve it in the simulation. Typical response of, or let's say the typical approach is, let's just get the simulation closer and closer and closer and closer to reality, and I would say that helps, but the problem is that we've sufficiently complicated systems like the hand, and it's actually impossible to model everything. So, in a sense, hand has tendons, tendons stretch. It has a rubber, rubber deforms.

• 18:00Speaker 02

  Also, the shape that you have in the simulation actually doesn't even correspond exactly to the real shape. And when there is a lot of interactions, the difference in the given place you are touching versus not touching might cause the object to pivot, let's say slip over, and so on. So, that's why things do not want to transfer.

• 18:19Daniel Whitenack?

  Yeah, so it's like a lot of very small kind of differences in how things are touched or moved can create a whole different outcome.

• 18:28Speaker 02

  Correct. In a sense, the fundamental idea that allows us for, let's say, adaptation to reality, say, is actually extremely simple what we did. So, in a sense, the initial approach is you have this single simulation, you can think about it like a single universe in which you are training, and then you are asking, here is an alternative universe in which you want to actually verify the performance. And we are just instead saying, if you will have entire distribution, many universes, and network has a capability to...

• 19:05Speaker 02

  Code, it's in, or let's say, try to distinguish them. Then it essentially might force the network to try to discover what are the underlying properties. So let's say, if we don't know exactly what's the weight of the cube, or I mean, more or less, maybe we know, but it might be off. If we have a network that just has a capability to, let's say, through interaction, pass the information, and these are like a very common networks, recurrent neural networks, then as we vary these parameters and it tries in the simulation on all of these instances, maximize the score, it implicitly actually does, it has to try to,

• 19:48Speaker 02

  based on the initial few seconds, try to find out what are these values. I mean, it's like a two combination of two things. On one side, it tries to be robust to some components, and on the one side, it tries to adapt to various things. And in a sense, this is in the core of the idea to actually achieve the transfer to the reality.

• 20:09Daniel Whitenack?

  Yeah, that's really interesting. So how long have you been kind of working towards this type of adaptation? And how has the process gone? Have you made other attempts to make this transfer to reality that haven't worked as well?

• 20:25Speaker 02

  Yes, so Anta project more or less took us one year. Let's say, I mean, earlier on, there was, let's say, maybe five people. Later on in the project, there is maybe closer to, let's say, fifteen or so. So I would say average ten human years. There were many, many attempts internally. And many of the things that we tried, they partially worked and so on. But, so I mean, the way how the team is more or less organized is, once we agree on the goal, like I try to, let's say, gauge people and ask, what do you think is the best way to solve the problem? And in this sense, as I said, like, people to some extent are, are towards on many axes, they are smarter than me,

• 21:10Speaker 02

  and they might sometimes better know than me what's the best approach. When you're in the situation that there are multiple competitive approaches, you're also becoming closer to the truth, what really works. So I mean, it's often the case that idea, as long as it is, let's say, sound, it will show signs of life, but it doesn't mean that this is the really ultimate solution. It might be sufficient for, you know, to present the conference or so, but our goal is truly to solve problem and actually get to the solutions that we think we can push forward.

• 21:47Daniel Whitenack?

  So when I was looking through the OpenAI website and kind of looking into some of what you're trying to achieve, I kept coming across this statement about, you know, safe applications of AI. And I know we've talked in the past about AI ethics and other things. I was wondering if you could briefly talk about, you know, safe AI, how you see that and, you know, maybe what counter examples to safe AI would be.

• 22:12Speaker 02

  So there are several problems into it. One problem is question, how to ensure that the system will be achieving the goal in the way that we intended it to achieve. So I mean, I gather many kind of philosophical examples and so on, and I want to go through some of them. It sounds somewhat foreign, but simultaneously, we are starting to see that actually it is not as trivial to tell actually system what to do, because it is completely abstracted from our values, ethics and so on. So in a sense, you tell someone, make money. It's like the best way to make money is steal a car, say, sell drugs,

• 22:53Speaker 02

  like all the stuff that is really, that you wouldn't intend the system to do. It's actually the best way.

• 22:59Daniel Whitenack?

  Things we probably don't want robots doing.

• 23:01Speaker 02

  Yeah, I mean, or even you don't need robots for it, to give you some example. Let's say, if you are really insanely clever about stock market, you can, as I'm saying, insanely clever and, you know, you have superhuman capabilities and what you truly care about is to maximize profit. You can, you know, cause a war in the country and short the stocks. It is completely valid strategy if that's the quantity.

• 23:29Daniel Whitenack?

  To reach an objective.

• 23:30Speaker 02

  To reach the objective. And in some sense, you can say that the the systems that we are training, you can say it has a little bit like a profile of psychopath. It only cares about the one thing and one thing only. That's literally how we optimize them. They are completely abstracted away from everything else. And they want just these treats, treats, treats. And they actually don't know even about the things that are really important. So the question is, what's the way even to inject what we want? What is our ethics? And I mean, I would say there are multiple axes into safety.

• 24:13Speaker 02

  So just told you about one, which is more or less called misspecification. I mean, so specify something, but it's actually something something slightly different that you really wanted. And we can see it, like even in some computer games, that he has the system to maximize score in the game. But truly, you would like ask the system to finish the game. And then it finds some bug on some level of the game and keeps on staying there. Let's say generating a lot of points, but actually doesn't progress anymore in the game. I mean, there are other axes, which is which is how to make the systems robust to adversaries.

• 24:52Speaker 02

  And to give some concrete examples, so let's say there was a there was a Twitter bot released by Microsoft, Tanya. And this folks from Microsoft, there is no doubt, there's like a lot of very clever researchers and so on. So can say that it's like they thought through various scenarios. Okay, but despite, let's say, thinking it through, turns out that the bot within several hours was hijacked and repurposed as soon as it started saying very offensive things on Twitter. So this is, you might ask, is it the case that as the systems will become smarter,

• 25:35Speaker 02

  will they be less prone to it? I actually think that it might be due to the overall increased complexity. The surface area will just increase.

• 25:45Daniel Whitenack?

  Yeah, it's it's almost even if you're trying to misspecify an objective, there's the space of objectives is is larger, it's more more complex.

• 25:54Speaker 02

  So I would say from perspective of pursuing safe AGI, there are also per se three main things. So one, in order to achieve it, I mean, you have to, let's say, work toward capabilities. Second one is you have to work towards safety. So what we just discussed. And the third one is a question, let's say, even if we would have it today, what are the, what's the step in terms of a policy? What should we do with it? And I would say also, all these three components actually fit into each other. So capabilities indicate actually maybe what the shape of AGI will be there for.

• 26:36Speaker 02

  What are the ways to actually inject our ethics and so on? The safety work and also it helps the policy people. So same thing like all these topics, they fit into each other.

• 26:49Daniel Whitenack?

  Yeah, that makes sense. So I'd love to switch directions here for the last bit of the show and kind of get your perspective. So you mentioned a lot of really exciting things around robotics and a lot of that involved kind of large scale computing and lots of large simulations. So wondering for people that are out there like trying to get their hands dirty with some of these techniques, maybe it's reinforcement learning or other things. What are some good ways for people to kind of get their hands dirty and start working on, on problems that are interesting, but maybe they aren't able to run these, you know, large scale simulations and that sort of thing?

• 27:27Speaker 02

  Yeah, I mean, there are incredible materials online. So I would just go first through all Coursera, Udacity, lectures from Berkeley, Stanford, follow all the homeworks. So I would say that's step one. There are also great books. I mean, there is a book by Ian, Deep Learning Book. There is, there is Reinforcement Learning Book by Richard Sutton. Yeah. So I would say that would be my starting point. And I think it's actually quite important to get very strong fundamentals, fundamentals. Cause in this sense, by default. When you run your models, first, they do not work at all.

• 28:06Daniel Whitenack?

  Yeah.

• 28:07Speaker 02

  And then the question is, so what do you do? And the simpler models, the fewer tricks or steps you have to do, and you have to familiarize yourself with them. The harder models, the larger number of these steps. So it's very likely that at first you don't know any of them. If you need to do 10 things, you are less likely to succeed versus if you need to do two things. I would really recommend to go through fundamentals instead of jumping right away to the most difficult architectures. And I would really recommend to as much as you can to implement things from scratch.

• 28:46Daniel Whitenack?

  Yeah, that's a, that's a great point. I know I've had some experience in the past where, you know, some research comes out. It's really interesting and, you know, you can go to a GitHub repo within a number of days and there's an open source architecture there. But you try to run it and you see all these weird behavior and maybe it behaves differently than you'd expect. But it's really hard if you don't know the fundamentals to dive into the debug and advance. So yeah, that's, that's great. I know I appreciated what you've said about any one person, even, even yourself, who, who's advanced a lot in this field, doesn't have all of the pieces of knowledge to, to, you know, perform a successful research or to advance a project.

• 29:29Daniel Whitenack?

  I was wondering, at OpenAI, how do you kind of structure your teams? And what do you look for when you're kind of putting together a team? So that you have a variety of experience and perspectives to actually give a good result?

• 29:42Speaker 02

  So different teams, they are, they have a little bit different values and they are differently organized. I can speak about robotics. It's extremely important to have people who are good team players. I would say also when we hire people, we hire it based on being incredible in something. It doesn't need to be exactly what they will be working on. Or so it's like more or less you want to verify brilliancy. And that's a sign that a person can adapt whatever is needed. I also, let's say, so in a sense, I want people to be able from day zero to contribute.

• 30:24Speaker 02

  Still, let's say, I encourage to spend one day per week on, let's say, do arbitrary learning. We have, let's say, internally curriculum with simple stuff. I like people who are excited about their resolving problems. So in this sense, when it comes to difficult problems, as for instance, as the last project, it is very common that the time in the middle is when everything is extremely difficult. And you need the people who have this, let's say, internal energy that they can, you know, still push it through and get it through.

• 30:56Daniel Whitenack?

  Yeah, so persistence and motivation and passion for the problem.

• 31:00Speaker 02

  Correct.

• 31:01Daniel Whitenack?

  Yeah, awesome. Well, I love to take time. If there's anything, where can people find out, you know, more about OpenAI? And are there any kind of open source projects or papers or efforts that you'd like to kind of share? And we can, we'll for sure post those in the show links and everything.

• 31:17Speaker 02

  We are quite frequently releasing what we are able to build on our blog. So that's at the openai.com website. We are also quite active on Twitter. So it's twitter.comopenai. There were, you ask about the research proposals or so. We posted, I think, even twice, let's say, a bunch of ideas for projects for people they want to pursue it, receive an indication of the level of difficulty. That might be a place to start.

• 31:51Daniel Whitenack?

  Awesome. And where can we find that?

• 31:53Speaker 02

  That's also on our website.

• 31:55Daniel Whitenack?

  Awesome. Well, I really appreciate you taking time. I know that you must be busy and you're getting ready for your talk. So I'll let you get back to that. But thank you so much for joining us. It was a really great conversation.

• 32:06Speaker 02

  Thank you, Daniel.

• 32:09Speaker 00

  All right. Thank you for tuning into this episode of Practical AI. If you enjoyed the show, do us a favor, go on iTunes and give us a rating. Go on your podcast app and favorite it. If you are on Twitter or social network, share a link with a friend, whatever you gotta do. Share the show with a friend if you enjoyed it. And bandwidth for changelog is provided by Fastly. Learn more at fastly.com. And we catch our airs before users do hear changelog because of Rollbar. Check them out at rollbar.com/changelog. And we're hosted on Linode Cloud servers. Head to linode.com/changelog. Check them out. Support this show. This episode is hosted by Daniel Whitenack and Chris Benson. Editing is done by Tim Smith. The music is by Breakmaster Cylinder. And you can find more shows just like this at changelog.com. When you go there, pop in

• 32:53Speaker 00

  your email address. Get our weekly email keeping you up to date with the news and podcasts for developers in your inbox every single week. Thanks for tuning in. We'll see you next week.