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