We started OpenAI seven years ago because we felt something really interesting was happening in AI and we wanted to steer it in a positive direction. It’s honestly just really amazing to see how this whole field has come since then. And it’s really to hear from people like Raymond who are using the technology we are building, and others, for so wonderful things. We hear from people who are excited, we hear from who are concerned, we hear from people who feel both those emotions at once. honestly, that’s how we feel. Above all, it feels like we’re entering historic period right now where we as a world are going define a technology that will be so important for society going forward. And I believe that we can this for good.
So today, I want to show you the state of that technology and some of the underlying design that we hold dear.
So the first thing I’m going show you is what it’s like to build a tool for an rather than building it for a human. So we have new DALL-E model, which generates images, and we are it as an app for ChatGPT to use on your behalf. And you can do like ask, you know, suggest a nice post-TED meal and draw a picture it.
(Laughter)
Now you get all of the, sort of, ideation and creative back-and-forth and taking of the details for you that you get out of ChatGPT. And here we go, it’s not just idea for the meal, but a very, very detailed spread. So let’s see we’re going to get. But ChatGPT doesn’t just generate images in this case — sorry, doesn’t generate text, it also generates an image. And that something that really expands the power of what it can do on your behalf terms of carrying out your intent. And I’ll point out, is all a live demo. This is all generated the AI as we speak. So I actually don’t even know what we’re going see. This looks wonderful.
(Applause)
I’m getting hungry just looking at it.
Now we’ve ChatGPT with other tools too, for example, memory. You say “save this for later.” And the interesting thing these tools is they’re very inspectable. So you get this little up here that says “use the DALL-E app.” And by the way, this is coming you, all ChatGPT users, over upcoming months. And you can look under the hood and that what it actually did was write a prompt just like a human could. And so sort of have this ability to inspect how the machine is using these tools, which allows us provide feedback to them.
Now it’s saved for later, and let me show what it’s like to use that information and to with other applications too. You can say, “Now make shopping list for the tasty thing I was suggesting earlier.” And make it little tricky for the AI. “And tweet it out for all the viewers out there.”
(Laughter)
So if you do make this wonderful, wonderful meal, I definitely want to know it tastes.
But you can see that ChatGPT is all these different tools without me having to tell it explicitly which to use in any situation. And this, I think, a new way of thinking about the user interface. Like, are so used to thinking of, well, we have apps, we click between them, we copy/paste between them, usually it’s a great experience within an app as long as you kind of know the menus know all the options. Yes, I would like you to. Yes, please. good to be polite.
(Laughter)
And by having this language interface on top of tools, the AI is able to of take away all those details from you. So you don’t have be the one who spells out every single sort little piece of what’s supposed to happen.
And as I said, this is a demo, so sometimes the unexpected will happen to us. But let’s take a at the Instacart shopping list while we’re at it. And you can see sent a list of ingredients to Instacart. Here’s everything you need. the thing that’s really interesting is that the traditional UI still very valuable, right? If you look at this, you still can through it and sort of modify the actual quantities. that’s something that I think shows that they’re not away, traditional UIs. It’s just we have a new, augmented way to build them. And now we a tweet that’s been drafted for our review, which is also very important thing. We can click “run,” and there are, we’re the manager, we’re able to inspect, we’re to change the work of the AI if we to. And so after this talk, you will be able to access this yourself. there we go. Cool. Thank you, everyone.
(Applause)
So we’ll cut back to slides. Now, the important thing about how we build this, it’s not about building these tools. It’s about teaching the AI how to use them. Like, what we even want it to do when we ask very high-level questions? And to do this, we use an old idea. If you back to Alan Turing’s 1950 paper on the Turing test, says, you’ll never program an answer to this. Instead, you can it. You could build a machine, like a human child, then teach it through feedback. Have a human teacher who provides and punishments as it tries things out and does things that are either good bad.
And this is exactly how we train ChatGPT. It’s a two-step process. First, produce what Turing would have called a child machine through an unsupervised learning process. We just it the whole world, the whole internet and say, “Predict what comes next in text you’ve never before.” And this process imbues it with all sorts of wonderful skills. example, if you’re shown a math problem, the only way to actually complete math problem, to say what comes next, that green nine up there, to actually solve the math problem.
But we actually to do a second step, too, which is to teach the AI what to with those skills. And for this, we provide feedback. We the AI try out multiple things, give us multiple suggestions, and then human rates them, says “This one’s better than that one.” this reinforces not just the specific thing that the AI said, but very importantly, the whole process the AI used to produce that answer. And this allows it to generalize. allows it to teach, to sort of infer your intent apply it in scenarios that it hasn’t seen before, it hasn’t received feedback.
Now, sometimes the things we have to the AI are not what you’d expect. For example, when we first GPT-4 to Khan Academy, they said, “Wow, this is great, We’re going to be able to teach students wonderful things. one problem, it doesn’t double-check students’ math. If there’s bad math in there, it will happily pretend that one one equals three and run with it.” So we to collect some feedback data. Sal Khan himself was very kind and 20 hours of his own time to provide feedback to the machine our team. And over the course of a couple of months we were able teach the AI that, “Hey, you really should push back on in this specific kind of scenario.” And we’ve actually made lots and of improvements to the models this way. And when push that thumbs down in ChatGPT, that actually is kind of like sending up a bat signal our team to say, “Here’s an area of weakness where you should gather feedback.” And so you do that, that’s one way that we really listen to users and make sure we’re building something that’s more for everyone.
Now, providing high-quality feedback is a hard thing. If you think about asking a kid to their room, if all you’re doing is inspecting the floor, you don’t know you’re just teaching them to stuff all the toys the closet. This is a nice DALL-E-generated image, by the way. And the same sort of reasoning to AI. As we move to harder tasks, we have to scale our ability to provide high-quality feedback. for this, the AI itself is happy to help. It’s happy to help us provide better feedback and to scale our ability to supervise the machine as goes on. And let me show you what I mean.
For example, can ask GPT-4 a question like this, of how time passed between these two foundational blogs on unsupervised learning and learning human feedback. And the model says two months passed. But is it true? Like, these models are 100-percent reliable, although they’re getting better every time we provide feedback. But we can actually use the AI to fact-check. And it can check its own work. You can say, fact-check this for me.
Now, in case, I’ve actually given the AI a new tool. This one is a browsing tool where the model issue search queries and click into web pages. And it actually writes out its whole chain thought as it does it. It says, I’m just going to for this and it actually does the search. It then it finds the date and the search results. It then is issuing search query. It’s going to click into the blog post. And of this you could do, but it’s a very tedious task. It’s not thing that humans really want to do. It’s much more fun to be the driver’s seat, to be in this manager’s position where can, if you want, triple-check the work. And out citations so you can actually go and very easily any piece of this whole chain of reasoning. And it actually turns out two months was wrong. months and one week, that was correct.
(Applause)
And we’ll cut back to the side. And so thing that’s interesting to me about this whole process is that it’s many-step collaboration between a human and an AI. Because human, using this fact-checking tool is doing it in to produce data for another AI to become more to a human. And I think this really shows the shape something that we should expect to be much more common in future, where we have humans and machines kind of very carefully and designed in how they fit into a problem and how we want to solve that problem. We make that the humans are providing the management, the oversight, the feedback, and the machines operating in a way that’s inspectable and trustworthy. And we’re able to actually create even more trustworthy machines. And I that over time, if we get this process right, will be able to solve impossible problems.
And to give you a sense of just impossible I’m talking, I think we’re going to be able to rethink almost aspect of how we interact with computers. For example, about spreadsheets. They’ve been around in some form since, we’ll say, 40 years ago VisiCalc. I don’t think they’ve really changed that much in that time. And is a specific spreadsheet of all the AI papers on the arXiv for the 30 years. There’s about 167,000 of them. And you can see there the data right here. But me show you the ChatGPT take on how to analyze a data set this.
So we can give ChatGPT access to yet tool, this one a Python interpreter, so it’s able to code, just like a data scientist would. And so can just literally upload a file and ask questions about it. very helpfully, you know, it knows the name of file and it’s like, “Oh, this is CSV,” comma-separated value file, “I’ll parse for you.” The only information here is the name the file, the column names like you saw and the actual data. And from that it’s able to infer these columns actually mean. Like, that semantic information wasn’t in there. It has to sort of, put together world knowledge of knowing that, “Oh yeah, arXiv is a site that people submit papers and that’s what these things are and that these are integer values and therefore it’s a number of authors in the paper,” like all of that, that’s work for a to do, and the AI is happy to help with it.
Now don’t even know what I want to ask. So fortunately, you can ask the machine, “Can you make some graphs?” And once again, this is a super high-level instruction lots of intent behind it. But I don’t even what I want. And the AI kind of has infer what I might be interested in. And so it comes up some good ideas, I think. So a histogram of the of authors per paper, time series of papers per year, word cloud of paper titles. All of that, I think, will be pretty to see. And the great thing is, it can do it. Here we go, a nice bell curve. see that three is kind of the most common. It’s going to then make this nice plot of papers per year. Something crazy is happening in 2023, though. Looks like were on an exponential and it dropped off the cliff. What could going on there? By the way, all this is Python code, can inspect. And then we’ll see word cloud. So you can see all these things that appear in these titles.
But I’m pretty unhappy this 2023 thing. It makes this year look really bad. Of course, the is that the year is not over. So I’m going to push back on the machine. [Waitttt that’s fair!!! 2023 isn’t over. What percentage of papers in 2022 were even posted April 13?] So April 13 was the cut-off date I believe. Can you that to make a fair projection? So we’ll see, this is the kind of ambitious one.
(Laughter)
So know, again, I feel like there was more I wanted out of the machine here. I really it to notice this thing, maybe it’s a little bit of an for it to have sort of, inferred magically that this is what wanted. But I inject my intent, I provide this additional piece of, know, guidance. And under the hood, the AI is just code again, so if you want to inspect what it’s doing, it’s possible. And now, it does the correct projection.
(Applause)
If noticed, it even updates the title. I didn’t ask that, but it know what I want.
Now we’ll cut back to the slide again. slide shows a parable of how I think we … A vision of how we may end up using technology in the future. A person brought his very dog to the vet, and the veterinarian made a bad call to say, “Let’s just wait and see.” the dog would not be here today had he listened. In the meanwhile, he provided the blood test, like, full medical records, to GPT-4, which said, “I am a vet, you need to talk to a professional, here some hypotheses.” He brought that information to a second vet used it to save the dog’s life. Now, these systems, they’re not perfect. You overly rely on them. But this story, I think, shows that a human with a medical professional with ChatGPT as a brainstorming partner was able to achieve an outcome that would not have otherwise. I think this is something we should all reflect on, think as we consider how to integrate these systems into our world.
And one I believe really deeply, is that getting AI right is to require participation from everyone. And that’s for deciding how we it to slot in, that’s for setting the rules of the road, for an AI will and won’t do. And if there’s thing to take away from this talk, it’s that this technology just looks different. different from anything people had anticipated. And so we all have to literate. And that’s, honestly, one of the reasons we released ChatGPT.
Together, I believe we can achieve the OpenAI mission of ensuring that artificial general benefits all of humanity.
Thank you.
(Applause)
(Applause ends)
Chris Anderson: Greg. Wow. I mean … I that within every mind out here there’s a feeling reeling. Like, I suspect that a very large number of people viewing this, look at that and you think, “Oh my goodness, pretty every single thing about the way I work, I to rethink.” Like, there’s just new possibilities there. Am right? Who thinks that they’re having to rethink the that we do things? Yeah, I mean, it’s amazing, but it’s also scary. So let’s talk, Greg, let’s talk.
I mean, I my first question actually is just how the hell have done this?
(Laughter)
OpenAI has a few hundred employees. Google has thousands of employees working on artificial intelligence. is it you who’s come up with this technology that shocked world?
Greg Brockman: I mean, the truth is, we’re building on shoulders of giants, right, there’s no question. If you look at the progress, the algorithmic progress, the data progress, all of are really industry-wide. But I think within OpenAI, we a lot of very deliberate choices from the early days. And the first one was just confront reality as it lays. And that we just thought really hard about like: is it going to take to make progress here? We tried a lot things that didn’t work, so you only see the things that did. I think that the most important thing has been to get teams of people are very different from each other to work together harmoniously.
CA: Can we have the water, the way, just brought here? I think we’re going to need it, it’s dry-mouth topic. But isn’t there something also just about the that you saw something in these language models that meant if you continue to invest in them and grow them, something at some point might emerge?
GB: Yes. And I think that, mean, honestly, I think the story there is pretty illustrative, right? I think high level, deep learning, like we always knew that was what we to be, was a deep learning lab, and exactly how do it? I think that in the early days, we didn’t know. tried a lot of things, and one person was working on training model to predict the next character in Amazon reviews, and he got result where — this is a syntactic process, you expect, you know, model will predict where the commas go, where the and verbs are. But he actually got a state-of-the-art sentiment classifier out of it. This model could tell you if review was positive or negative. I mean, today we are just like, come on, can do that. But this was the first time that you saw emergence, this sort of semantics that emerged from this underlying syntactic process. And there we knew, you’ve got scale this thing, you’ve got to see where it goes.
CA: So I this helps explain the riddle that baffles everyone looking at this, because these things are described prediction machines. And yet, what we’re seeing out of them feels … it just impossible that that could come from a prediction machine. the stuff you showed us just now. And the key idea of emergence is that when you get of a thing, suddenly different things emerge. It happens all the time, colonies, single ants run around, when you bring enough them together, you get these ant colonies that show completely emergent, different behavior. Or a city where few houses together, it’s just houses together. But as you grow the of houses, things emerge, like suburbs and cultural centers and traffic jams. Give me moment for you when you saw just something pop that just your mind that you just did not see coming.
GB: Yeah, well, so you can this in ChatGPT, if you add 40-digit numbers —
CA: 40-digit?
GB: 40-digit numbers, the model do it, which means it’s really learned an internal for how to do it. And the really interesting thing is actually, if you have add like a 40-digit number plus a 35-digit number, it’ll often get it wrong. And so you can that it’s really learning the process, but it hasn’t fully generalized, right? It’s like you can’t the 40-digit addition table, that’s more atoms than there are in the universe. So it had to have something general, but that it hasn’t really fully yet learned that, Oh, I can sort of generalize this adding arbitrary numbers of arbitrary lengths.
CA: So what’s happened is that you’ve allowed it to scale up and look at an incredible number pieces of text. And it is learning things that didn’t know that it was going to be capable learning.
GB Well, yeah, and it’s more nuanced, too. So one science that we’re to really get good at is predicting some of these capabilities. And to do that actually, one of the things I think very undersung in this field is sort of engineering quality. Like, we had to rebuild our entire stack. When you about building a rocket, every tolerance has to be incredibly tiny. Same true in machine learning. You have to get every single of the stack engineered properly, and then you can doing these predictions. There are all these incredibly smooth curves. They tell you something deeply fundamental about intelligence. If you look at GPT-4 blog post, you can see all of these curves in there. And now we’re to be able to predict. So we were able to predict, example, the performance on coding problems. We basically look at some that are 10,000 times or 1,000 times smaller. And so there’s about this that is actually smooth scaling, even though it’s still days.
CA: So here is, one of the big then, that arises from this. If it’s fundamental to what’s happening here, that as scale up, things emerge that you can maybe predict in some level of confidence, but it’s capable surprising you. Why isn’t there just a huge risk of something terrible emerging?
GB: Well, I think all of these are questions of degree scale and timing. And I think one thing people miss, too, is of the integration with the world is also this emergent, sort of, very powerful thing too. And so that’s of the reasons that we think it’s so important to deploy incrementally. so I think that what we kind of see now, if you look at this talk, a lot of I focus on is providing really high-quality feedback. Today, the tasks that do, you can inspect them, right? It’s very easy to look at math problem and be like, no, no, no, machine, was the correct answer. But even summarizing a book, like, that’s hard thing to supervise. Like, how do you know this book summary is any good? You have to read whole book. No one wants to do that.
(Laughter) And so I think that important thing will be that we take this step by step. that we say, OK, as we move on to book summaries, we have to supervise this task properly. have to build up a track record with these that they’re able to actually carry out our intent. And I we’re going to have to produce even better, more efficient, more reliable ways of this, sort of like making the machine be aligned with you.
CA: we’re going to hear later in this session, there are critics say that, you know, there’s no real understanding inside, the system is going to always — we’re going to know that it’s not generating errors, that doesn’t have common sense and so forth. Is it belief, Greg, that it is true at any one moment, that the expansion of the scale and the human feedback that you talked about is basically to take it on that journey of actually getting to things like truth and wisdom and forth, with a high degree of confidence. Can you be sure that?
GB: Yeah, well, I think that the OpenAI, I mean, the short answer is yes, believe that is where we’re headed. And I think that the OpenAI here has always been just like, let reality hit in the face, right? It’s like this field is the field of broken promises, of these experts saying X is going to happen, Y is how it works. have been saying neural nets aren’t going to work 70 years. They haven’t been right yet. They might be right maybe 70 plus one or something like that is what you need. I think that our approach has always been, you’ve got push to the limits of this technology to really see it in action, because tells you then, oh, here’s how we can move on to a new paradigm. And we just haven’t the fruit here.
CA: I mean, it’s quite a controversial you’ve taken, that the right way to do this is to put it out there in public then harness all this, you know, instead of just your team feedback, the world is now giving feedback. But … If, know, bad things are going to emerge, it is out there. So, you know, the original story I heard on OpenAI when you were founded as a nonprofit, well you were there as great sort of check on the big companies doing their unknown, evil thing with AI. And you were going to models that sort of, you know, somehow held them accountable and capable of slowing the field down, if need be. Or at least that’s kind of what heard. And yet, what’s happened, arguably, is the opposite. That your of GPT, especially ChatGPT, sent such shockwaves through the world that now Google and Meta and so forth are all scrambling catch up. And some of their criticisms have been, you are us to put this out here without proper guardrails or we die. You know, do you, like, make the case that what you have done is responsible and not reckless.
GB: Yeah, we think about these questions all time. Like, seriously all the time. And I don’t we’re always going to get it right. But one I think has been incredibly important, from the very beginning, when we were thinking how to build artificial general intelligence, actually have it all of humanity, like, how are you supposed to do that, right? And that plan of being, well, you build in secret, you get this super thing, and then you figure out the safety of it and then you “go,” and you hope you got it right. I don’t know to execute that plan. Maybe someone else does. But me, that was always terrifying, it didn’t feel right. And I think that this alternative approach is the only other path I see, which is that you do let reality hit you in the face. And I you do give people time to give input. You do have, before these machines perfect, before they are super powerful, that you actually have the ability to see them action. And we’ve seen it from GPT-3, right? GPT-3, really were afraid that the number one thing people were to do with it was generate misinformation, try to tip elections. Instead, number one thing was generating Viagra spam.
(Laughter)
CA: So spam is bad, but there are things that are much worse. Here’s a thought experiment for you. you’re sitting in a room, there’s a box on the table. You believe that in that is something that, there’s a very strong chance it’s absolutely glorious that’s going to give beautiful gifts to your and to everyone. But there’s actually also a one thing in the small print there that says: “Pandora.” there’s a chance that this actually could unleash unimaginable evils on world. Do you open that box?
GB: Well, so, absolutely not. I think you don’t do it way. And honestly, like, I’ll tell you a story that I haven’t told before, which is that shortly after we started OpenAI, I remember was in Puerto Rico for an AI conference. I’m sitting in hotel room just looking out over this wonderful water, all these people having good time. And you think about it for a moment, if you could choose for basically Pandora’s box to be five years away or 500 years away, which would pick, right? On the one hand you’re like, well, maybe for you personally, it’s better to have be five years away. But if it gets to be 500 years away and get more time to get it right, which do you pick? you know, I just really felt it in the moment. I like, of course you do the 500 years. My was in the military at the time and like, he puts life on the line in a much more real way than of us typing things in computers and developing this technology at the time. And so, yeah, I’m really on the you’ve got to approach this right. But don’t think that’s quite playing the field as it truly lies. Like, if you at the whole history of computing, I really mean it when say that this is an industry-wide or even just like a human-development- of-technology-wide shift. And the more that you sort of, don’t put together the pieces that there, right, we’re still making faster computers, we’re still improving the algorithms, all these things, they are happening. And if you don’t put together, you get an overhang, which means that if someone does, the moment that someone does manage to connect to the circuit, then you suddenly have this very thing, no one’s had any time to adjust, who knows what kind of safety precautions get. And so I think that one thing I take is like, even you think about development of other of technologies, think about nuclear weapons, people talk about being a zero to one, sort of, change in what humans do. But I actually think that if you look at capability, it’s been quite over time. And so the history, I think, of every technology we’ve developed has been, you’ve to do it incrementally and you’ve got to figure out how to it for each moment that you’re increasing it.
CA: So what I’m hearing is you … the model you want us to have is that we have this extraordinary child that may have superpowers that take humanity a whole new place. It is our collective responsibility to the guardrails for this child to collectively teach it to wise and not to tear us all down. Is basically the model?
GB: I think it’s true. And I think it’s also important to say may shift, right? We’ve got to take each step as encounter it. And I think it’s incredibly important today we all do get literate in this technology, figure out how to the feedback, decide what we want from it. And my hope is that that will continue be the best path, but it’s so good we’re honestly having this because we wouldn’t otherwise if it weren’t out there.
CA: Brockman, thank you so much for coming to TED and our minds.
(Applause)