We started OpenAI seven years ago, because we felt like something really interesting was happening in AI, and we wanted to help steer it in a positive direction. It's honestly just really amazing to see how far this whole field has come since then. And it's really gratifying to hear from people like Raymond who are using the technology we are building in others, for so many wonderful things. We hear from people who are excited. We hear from people who are concerned. We hear from people who feel both those emotions at once. And honestly, that's how we feel. Above all, it feels like we're entering an historic period right now. Where we as a world are going to define a technology that will be so important for our society going forward. And I believe that we can manage this for good. So today, I wanna show you the current state of that technology and some of the underlying design principles that we hold dear. So the first thing I'm going to show you is what it's like to build a tool for an AI rather than building it for a human. So we have a new Dolly model, which generates images, and we are exposing it as an app for chat GPT to use on your behalf. And you can do things like ask, you know, suggest a nice post ted meal and draw a picture of it. Now, you get all of the sort of ideation, and creative back and forth, and taking care of the details for you that you get out of chat, EPT. And here we go, it's not just the idea for the meal, but very, very, detailed spread. So let's see what we're gonna get. But ChatGBT doesn't just generate images in this case. It doesn't generate text. It also generates an image. And that is something that really expands the power of what it can do on your behalf in terms of carrying out your intent. I'll point out this is all live demo. This is all generated by the AI as we speak, so I actually don't even know what we're gonna see. This looks wonderful. Now I'm getting hungry, just looking at it. Now we've extended chat ticketee with other tools too. For example, memory, you can say, save this for later. And the interesting thing about these tools is they're very inspectable. So you get this little pop up here that says, use the Dolly app, and by the way, this is coming to all chatty PT users over upcoming months. And you can look under the hood and see that what it actually did was write a prompt just like a human could. And so you, you sort of have this ability to inspect how the machine is using these tools, which allows us to provide feedback to them. Now it's saved for later, and let me show you what it's like to use that information and to integrate with other applications too. You can say, now make a shopping list for the Tasty thing I was suggesting earlier and make it a little tricky for the AI. And tweet it out for all the Ted viewers out there. So if you do make this wonderful, wonderful meal, I definitely want to know how it tastes. But you can see that Chat EBT is selecting all these different tools without me having to tell it explicitly which ones to use in any situation. And this, I think, shows a new way of thinking about the user interface. Like, we are so used to thinking of, well, we have these apps. We cook between them. We copy paste between them. And usually, it's a great experience within an app as long as you kinda know the menus and know all the options yes, I would like you to. Yes, please. Always good to be polite. And by having this unified language interface on top of tools, the AI is able to sort of take away all those details from you. So you don't have to be the one who spells out every single sort of little piece of what's supposed to happen. And as I said, this is a live demo. So, sometimes the unexpected will happen to us. But let's take a look at the Instacart shopping list while we're at it. And you can see we sent a list of ingredients to instacart. Here's here's everything you need. And the thing that's really interesting is that the traditional UI is still very valuable. Right? If you look at this, you still can click through it and sort of modify the, the actual quantities. And that's something that I think shows that that, They're not going away traditional UIs. It's just we have a new augmented way to build them. And now we have a tweet that's been drafted for our review, which is also a very important thing. We can click run, and there we are. We're the manager. We're able to inspect. We're able to change the the work of the AI if we want to. And so, after this talk, you will be able to access this yourself. And there we go. Cool. So thank you, everyone. So we'll cut back to the slides. Now the important thing about how we build this, it's not just about building these tools. It's about teaching the AI how to use them, Like, what do we even want it to do when we ask these very high level questions? And to do this, we use an old idea. If you go back to Alan Turing's nineteen fifty paper on the Turing test, he says, look, you'll never program an answer to this. Instead, you can learn it. You could build a machine like a human child, and then teach it through feedback, have a human teacher, provides rewards and punishments as it tries things out and does things that are either good or bad. And this is exactly how we train ChachiPT, it's a two step process. First, we produce what Turing would have called a child machine through an unsupervised learning process, We just show it the whole world, the whole internet, and say, predict what comes next in text you've never seen before. And this process imbues it with all sorts of wonderful skills. For example, if you're shown a math problem, the only way to actually complete that math problem, say what comes next, that green nine up there is to actually solve the math problem. But we actually have to do a second step too, which is to teach the AI what to do with those skills. And for this, we provide feedback We have the AI try out multiple things, give us multiple suggestions, and then the human rates them, says this one's better than that one. And this reinforces not just the specific thing that the AI said, but very importantly, the whole process that the AI use to produce that answer, and this allows it to generalize, it allows it to teach to to sort of infer your intent and apply it in scenarios that it hasn't seen before, that it hasn't received feedback. Now, sometimes the things we have to teach the AI, are not what you'd expect. For example, when we first showed GPD four to Khan Academy, they said, wow, this is so great. We're going to be able to teach students wonderful things. Only one problem, it doesn't double check students' math. If there's some bad math in there, it will happily pretend that one plus one equals three and run with it. So we had to collect some feedback data. Salcon himself was very kind and offered twenty hours of his own time to provide feedback to the machine alongside our team, And over the course of a couple months, we were able to teach the AI that, hey, you really should push back on humans in this specific kind of scenario. And we've we've actually made lots and lots of improvements to, to to the models this way, and when you push that thumbs down and chat EPT, actually is kinda like sending up a bat signal to our team to say, here's an area of weakness where you should gather feedback. And so when you do that, that's one way that we really listen to our users and make sure we're building something that's more useful for everyone. Now providing high quality feedback, is a hard thing. If you think about asking a kid to clean their room, if all you're doing is inspecting the floor, you don't know if you're just teaching them to stuff all the toys in the closet. This is a nice dolly generated image, by the way. And the same sort of, reasoning applies to AI as we move to harder tasks, we will have to scale our ability to provide high quality feedback. But for this, the AI itself is is happy to help. It's happy to help us provide even better feedback and to scale our ability to supervise the machine as time goes on. And let me show you what I mean. For example, you can ask for g p d four question like this of how much time passed between the these two foundational logs on, unsupervised learning and learning from human feedback. And the model says two months passed. But is it true? Like, these models are not a hundred percent reliable, although they're getting better every every time we we provide some feedback, but we can actually use the AI to fact check. It's and it can actually check its own work. You can say fact check this for me. Now in this case, I've actually given the AI a new tool. This one is a browsing tool, where the model can issue search queries and click into web pages. And it actually writes out its whole chain of thought as it does it. It says I'm just gonna search for this, and it actually does the search. It then, it finds the, the publication date and those search results, it then is issuing another search query, it's going to click into the blog post, and all of this you could do, but it's a very tedious task. Not a thing that humans really want to do. It's much more fun to be in the driver's seat, to be in this manager's position where you can, if you want, triple check the work, and outcome citations so you can actually go and very easily verify any piece of this whole chain of reasoning, and it actually turns out two months was wrong, two months in one week. That was correct. Now we'll cut back to the slide. And so the thing that's so interesting to me about this whole process is that it's this many step collaboration between the human and an AI, because a human using this fact checking tool is doing it in order to produce data for another AI to become more useful to a human. And I think this really shows the shape of something that we should expect to be much more common in the future. Where we have humans and machines kind of very carefully and delicately designed in how they fit into a problem and how we want to solve that problem. We make sure that the humans are providing the management to oversight the feedback, and that the machines are operating in a way that's inspectable and trustworthy. And together, we're able to actually even create even more trustworthy machines. And I think that over time, if we get this process right, we will be able to solve impossible problems. And to give you a sense of just how impossible I'm talking, I think we're gonna be able to rethink almost every aspect of how we interact with computers. For example, think about spreadsheets. They've been around in some form since, you know, let's say forty years ago with Visicalc. And I don't think they've really changed that much in that time. And here is a specific spreadsheet, of all the AI papers on, the archive for the past thirty years. There's about a hundred sixty seven thousand of them. And you can see their the data right here. But let me show you the chat GPT take on how to analyze a data set like this. So we can give chat utility access to yet another tool. This one, a Python interpreter, so it's able to run code just like a data scientist would. And so you can just literally upload a file and ask questions about it, and very helpfully, you know, it it knows the name of the file. And it's like, oh, this is CSV, a common separate value file. I'll parse it for you. The only information here is the name of the file, the column names, like you saw, and then the actual data. And from that, it's able to infer what these columns actually mean. Like, that semantic information wasn't in there. It has to sort of put together its world knowledge of knowing that, oh, yeah, archive is site that people submit papers. And therefore, that's what these things are, and that these are integer values. And so, therefore, it's a number of authors in the paper. Like all of that, that's work for a human to do. And the AI is happy to help with it. Now I don't even know what I wanna ask. So fortunately, you can ask the machine. Can you make some exploratory graphs? And once again, this is a super high level instruction with lots of intent behind it, I don't even know what I want, and the AI kinda has to infer what I might be interested in. And so it comes up with some good ideas, I think. So histogram, the number of authors per paper, time series of papers per year, word cloud of the paper titles, all of that I think will be pretty interesting to see. And the great thing is it can actually do it. Here we go. Nice bell curve. You see that three is kind of the most common. It's going to then write a it's gonna make this nice spot of the, papers per year. Something crazy is happening in twenty twenty three, though. It looks like we were on an exponential, and it dropped off a cliff. What could be going on there? And by the way, all this is Python code you can expect, and then we'll see the word cloud. And so you can see all these wonderful things that appear in these titles. But I'm pretty unhappy about this twenty twenty three thing. It looks makes this year old really bad. Of course, the problem is that the year's not over. So I'm going to push back on the machine. So April thirteenth was a cutoff date, I believe. Can you use that? Thank you for your projection. So we'll see you. This is a kind of ambitious one. So, you know, again, I feel like there was more I wanted out of the machine here. I really wanted it to, like, notice this thing. Maybe it's a little bit, a little bit of of an overreach for it to have sort of inferred magically that this is what I wanted, but I inject my intent. I provide this additional piece of of, you know, sort of guidance. And under the hood, the AI is just writing code again. So if you want to inspect what it's doing, it's very possible. And now It does the correct projection. And if you notice, it even updates the title. I didn't ask for that, but it knows what I want. Now, we'll cut back to the slide again. This slide shows a parable of how I think we, you know, a vision of how we may end up using this technology in the future. A person brought his very sick dog to the vet who, and the veterinarian made a bad call to say, let's just wait and see. And The dog would not be here today had he listened. In the meanwhile, he provided the blood test, like the full medical records to GPT four, which said, I am not a vet. You need to talk to professional. Here are some hypotheses. He brought that information to a second vet who used it to save the dog's life. Now these systems, they're not perfect. You cannot overly rely on them. But this story, I think, shows that the human with a medical professional and with chat GPT as a brainstorming partner was able to achieve an outcome that would not have happened otherwise. Think this is something we should all reflect on and think about as we consider how to integrate these systems into our world. And one thing I believe really deeply is that getting AI right is going to require participation from everyone. And that's for deciding how we want it to slot in. That's for setting the rules of the road for what an AI will and won't do. If there's one thing to take away from this talk, it's that this technology just looks different, just different from anything people had anticipated. And so we all have to become literate. And that's honestly one of the reasons we release chat GPT. Together, I believe that we can achieve the open AI mission of ensuring that artificial general intelligence benefits all of humanity. Thank you. Greg. Wow. I mean, I suspect that within every mind out here, apart from there's a feeling of reeling. Like, I suspect that a very large number of people viewing this, you look at that and you think, oh my goodness. Pretty much every single thing about the way I work, I need to rethink. Like, there's just new possibilities there. Am I right? Who thinks that they're just they're they're having to rethink the way that we do things. Yeah. I mean, it's it's it's amazing, but it's also it's also really scary. So let's let's talk, Greg. Let's talk. We have Absolutely. I mean, I guess my my first question actually is just how the hell have you done this? You know, like open open AI has a few hundred employees Google has thousands of employees working on artificial intelligence. How how why is it you who's come up with this technology that shocked the world. Yeah. Well, I mean, the truth is we're all building on shoulders of giants. Right? There's no question if you look at the compute progress, the algorithmic progress, the data progress, all of those are are really, really industry wide, But I think within OpenAI, we made a lot of very deliberate choices from the early days. And the first one was just to confront reality as it lays. And, you know, that we just sort of, like, thought really hard about, like, what is it going to take to make progress here? We tried a lot of things that didn't work, so you only see the things that did. And I think that the the most important thing has been to get teams of people who are very different from each other to work together harmoniously. Can we have, the water, by the way, it just brought here? I said, I think we're gonna need it. So drive help me drive dry mouth topic. But there's is isn't there something also? It's just about the fact that that you saw something in these language models that meant that if you continue to invest in them and grow them, that something at some point might emerge. Yes. And I think that I mean, honestly, I think the story there is is pretty illustrative. Right? I think that that at a high level, deep learning, like, we always knew that was what we wanted to be. It was a deep learning lab. And exactly how to do it, like I think said in the early days, we didn't know. We we tried a lot of things, and one person was working on training a model to predict the next character in, in Amazon reviews, and he got a result where this is a this is a syntactic process. You expect to know the model will predict where the commas go, where the nouns and verbs are, but he actually got a state of the art sentiment analysis classifier out of it. That this model could tell you if a review is positive or negative. I mean, today, we are just like, come on. Like, anyone could do that. But this was the first time that you saw this emergence this sort of semantics that emerged from this underlying syntactic process. And there we knew, you gotta scale this thing. You gotta see where it goes. So I think this helps explain the the the the riddle that baffles everyone looking at this, because these things are described as prediction machines. And yet what we're seeing out of them feels it just feels impossible that that could come from a, you know, prediction machine, just the stuff you showed us just now. And the the key idea of emerging is that when you would get more of a thing, suddenly different things emerge, it happens all the time. Ant colonies, single ants run around when you bring enough of them together, know, you get these ant colonies that have show completely emergent and different behavior or a city where a few houses together, it's just houses together, but as you grow the number of houses, things emerge like suburbs and cultural centers and traffic jams. Give me one moment for you when you saw just something pop that just blew your mind that you just did not see coming. Yeah. Well, so if you you can try this in chat you can see if you add forty digit numbers, forty digit. Forty digit numbers, the model will do it, which means it's really learned a internal circuit for how to do it. And the funny, the really interesting is actually if you have an ad like a forty digit number, plus a thirty five digit number, it'll often get it wrong. And so you can see that it's really learning the process, but it hasn't fully generalized. Right? It's like you can't memorize the forty digit addition table. That's more add ins that they're on the universe. So it had to have learned something general but that it hasn't really fully yet learned that, oh, I can, like, sort of generalize this to adding arbitrary numbers of arbitrary lengths. So what's happened here is is that you've you've you've allowed it to scale up and look at an incredible number of pieces of text, and it is learning things that you didn't didn't know that it was gonna be capable of learning. Well, yeah, and it's it's more nuanced too, because so one science that we're starting to really get good at is predicting some of these emergent capabilities. And to do that, actually, one of the things I think is very undersung in this field is sort of engineering quality. Like, we had to rebuild our entire stack and get know, like, when you think about building a rocket, like, you know, every tolerance has to be, like, incredibly tiny. Same is true in machine learning. You have to get every single piece of the stack engineered properly And then you can start doing these predictions. There are all these incredibly smooth scaling curves, which tell you something deeply fundamental about intelligence. You do look at our GPT four blog posts, you can you can see all of these curves in there. And now we're starting to be able to predict. So we were able to predict, for example, the performance on coding problems. From, you know, we basically look at some models that are ten thousand times or a thousand times smaller. And so there's something about this that is actually smooth scaling, even though it's still early days. So here is one of the big fears then that arises from this. If it's fundamental to what's happening here that as you scale up, things emerge that that you you you can't you can maybe predict in some level of confidence, but they still it's capable of surprising you. Why isn't there just a huge risk of something truly terrible emerging? Well, I think all these are questions of degree, and scale and timing. And I think one thing people miss too is sort of the integration with the world. There's also this, like, incredibly emergent, like, sort of very powerful thing too. So that's one of the reasons that we think it's so important to deploy incrementally. And so I think that what we kind of see right now, if you look at this talk, a lot of what I focus on is providing really high quality feedback, today, the tasks that we do, you can inspect them, right, that it's very easy to look at that math problem and be like, no, no, no, machine, like, seven was the correct answer. But even summarizing a book, like, that's a hard thing to to supervise. Like, how do you know if this book summary is any good? You have to read the whole book. No one wants to do that. And so I think that the the important thing will be that we take this step by step. And that we say, okay, as we move on to book summaries, we have to supervise this task properly. We have to build up a track record with these machines that they're able to actually carry out our intent. I think we're going to have to produce even better, more efficient, sort of, more reliable ways of scaling this sort of, like, making the machine be aligned with you. We're gonna hear a letter in this session, there are there are critics who say that, you know, there's there's there's no real understanding inside the system is it going to always we're never gonna know that it's not generating errors that it doesn't have common sense and so forth. Is is it your belief, Greg, that that that is true at any one moment, but that the expansion of the scale and the human feedback you know, that you talked about is basically going to take it on that journey of actually getting to things like truth and wisdom and so forth with a high degree of confidence, how can you be sure of that? Yeah. Well, I think that the open AI I mean, the short answer is yes. I believe that is that is where we're headed. And I think that the open AI approach here has always been just like, let reality hit you in the face. Right? It's like, this field is the field of broken promises of all these experts saying x is going to happen, y is how it works. People have been saying neural nets aren't gonna work for seventy years. They haven't been right yet. They might be right, you know, maybe seventy years plus one or something like that is what you need. But I think that our approach has always been you gotta push to the limits of this technology to really see it in action, because that tells you, then, oh, here's how we can move on to a new paradigm, and we just haven't exhausted the fruit here. I mean, it's quite a controversial stance you've taken that the right way to do this is to put it out there in public and then harness all this, you know, instead of just your team giving feedback, the world is now giving feedback. But if, you know, bad things are going to emerge, it is out there so, you know, the original story that I heard on OpenAI when you were founded as a nonprofit, where you were there as the great sort of check on the big companies doing their unknown, possibly evil thing with AI, and you were going to you were going to build models that sort of you know, somehow held them accountable and and could was capable of slowing the field down if need be, or at least that's that's kind of what ahead. And yet what's happened arguably is the opposite that you is that your release of GPT, especially chat b t GPT puts sent such shock waves through the tech world that now Google and meta and so forth are all scrambling to catch up, and some of their criticisms have been you are forcing us to put this out here without proper guardrails or we die. You know, how how do you like, make the case that what you have done is responsible here and not reckless. Yeah. We think we think about these questions all the time. Like, like, seriously all the time, and I think that that I don't think we're always gonna get it right. But one thing I think has been incredibly important, like from the very beginning, when we were thinking about how to build artificial general intelligence, actually have it benefit all of humanity. Like, how are you supposed to do that? Right? And that the default plan of being like, well, you build in secret, you're kind of like, you know, You get this super powerful thing, and then you, like, figure out the safety of it, and then you push go, and you hope you got it right. Like, I don't know how to execute that plan. Maybe someone else does, but for me, that was always terrifying. It didn't feel right. And so I think that that this alternative approach is the only sort of other path that I see, which is that you do let reality hit you in the face. And I said, you do give people time to give input. You do have, well, before these machines are perfect, before they are super powerful, actually have the ability to see them in action. And we've seen it from GPD three. Right? GPD three, we really were afraid that the number one thing people were gonna do with it was generate misinformation, try to elections. Instead, the number one thing was generating Viagra spam. So Viagra spam is bad, but there are things that are much worse. If here's a thought experiment for you, suppose you're you're you're sitting in a room, there's a box on the table you believe that in that box is something that there's a very strong chance it's something absolutely glorious that's going to give beautiful, you know, gifts to your family and and and to everyone. But there's actually also a one percent thing in the small print there that says Pandora. And, there's there's a chance that this actually could unleash unimaginable evils on the world. Do you open that box? Well, so abs absolutely not. I I think I think you don't do it that way. And act honestly, like, I'll I'll tell you a story, that I that I haven't actually told before, which is that, shortly after we started OpenAI, I remember I was at was in Puerto Rico for an AI conference. I was sitting in the hotel room, just looking out over this wonderful water, all these people having a good time, and you think about it for a moment. Like, if you could choose for a, like, you know, sort of potentially, like, basically that Pandora's box to be, you know, five years away or five hundred years away. Which would you pick? Right? And on the one hand, you're like, well, like, you know, maybe for you personally, it's better to have it be five years away. But if it gets to be five hundred years away and people get more time to get it right, like, which do you pick? And, like, you know, I just, like, really felt it in that moment. I was like, of course, you do the five hundred years. Like, for real, like, there's many people. Like, my my brother is in the military at the time, and you're, like, he puts his life on the line in a much more real way than, like, any of us, you know, typing things in computers and developing this technology, at the time. And so, yeah, like, I'm I'm I'm really sold on the you gotta approach this right, but I don't think that's quite playing the field as it truly lies. Like, if you look at the whole history of computing, like, that that I I'm I really mean it when I say that this is a industry wide or even, like, sort of just almost like a human development of technology wide shift. And the more that you sort of don't put together the pieces, that are there, right, we're still making faster computers, we're still improving the algorithms, like all these things they are happening. And if you don't put them together, you get an overhang, which means that if someone does, or, you know, that that the moment that someone does manage to connect to the circuit, then you suddenly have this very powerful thing. No one's had any time to adjust like, who knows what kind of safety precautions you get? And so I think that that one thing I take away is, like, even you think about development of other technologies. Think about nuclear weapons, people talk about being like a zero to one sort of like, you know, sort of change in in what humans could do. But I actually think that if you look at capability. It's been quite smooth over time. And so the history, I think, of every technology we've developed has been you gotta do it incrementally, and you gotta figure out how to manage it for each moment that you're sort of increasing it. So what I'm hearing is that you that the model you almost have is that we have birthed this extraordinary child that may have superpowers that take humanity to a whole new place. It is our collective responsibility to provide the guardrails for this this child to collectively teach it to be wise and not to tear us all or down. Is that basically the model. I I think it's true, and I think it's also important to say this may shift, right? Like, we gotta take each step as we encounter it I think it's incredibly important today that we all do get literate in this technology, figure out how to provide the feedback, decide what we want from it, And I think that my hope is that that will be continue to be the best path, but it's so good we're honestly having this debate because we wouldn't otherwise if if it weren't out there. Greg Brockman. Thank you so much for coming to Tedd and blowing our minds. Thank you. Appreciate it.