[0:00:00] So one person with a really good idea and a deep understanding of what a customer needs is going to be able to execute on that with what would have taken complex, many, many person teams before. And so giving someone the ability to do more, I think is going to lead to a big shift. Welcome to Ask More of AI, podcast at the intersection of AI and business. I'm Clara Shy, CEO of AI at Salesforce.
[0:00:29] This week, we're coming to you from Dreamforce, the world's largest AI event, where we just heard an amazing conversation between Salesforce CEO and founder, Mark Benioff, and OpenAI CEO and co-founder, Sam Altman. Sam covers everything from what he's learned and how he's been inspired by his favorite AI movie, Her, as well as what's been surprising to him about the development of AI. Take a listen. Sam, welcome to Dreamforce. Thanks for having me. Oh, it's my pleasure.
[0:00:59] And Sam, you know, you've been traveling all over the world. I haven't really seen you since you got back. I saw you just before you left. I know how excited you were to go to all these different countries and give them the prophecy of artificial intelligence, the good, the bad, the ugly. What was your biggest surprise making all these trips? The level of enthusiasm, hopefulness, excitement around the world, of course, balanced with wanting to make sure we...
[0:01:28] Which country really stood out to you as
[0:01:56] Potentially a great leader of artificial intelligence. That was another positive surprise. The quality of work happening everywhere was really something. I think the United States will be the greatest leader in artificial intelligence. We're very blessed to have so many things in our favor. But this will be a truly global effort. And people will contribute all together all around the world.
[0:02:22] Did you meet somebody who really offered you an incredible insight or inspired you or somebody that you said, you know, this really changes my mind on this aspect of AI? Was there someone out there that you were impacted by? I mean, a lot of like, you know, really super impressive world leaders had a lot of good advice for me and for the company, and that was helpful. But the biggest takeaway was just talking to our users around the world.
[0:02:50] And out of all of that, and out of what we heard that they wanted, that they had problems with the technology, the features we were missing, that has really fed into what we'll be launching in a couple of months. And that was definitely the most meaningful feedback we got. Indonesia gave you the golden visa. Was that inspired by your love of eat, pray, love? A lot of people...
[0:03:19] A lot of people that I met with on the trip were like, oh, you need to come here more. We'll give you a golden visa. So I got this little collection in my passport now, which is awesome and unexpected. I do love Eat, Pray, Love. I thought that was a great book. You've got a sparkly visa? Yeah. Passport is this shining bright? Passports are very beat up after the trip. I fell in the water a few times, but that's okay. Were you swimming with your passport? Not intentionally, but it did happen once.
[0:03:50] Let me ask you, you've thought about AI probably more than most people on Earth, and you've been really having the opportunity to interact with so many great AI leaders, not just all over the world, but here in the headquarters of Artificial Intelligence, San Francisco. What's been your biggest surprise in the last seven or eight years of OpenAI? Here we are, it's 2023, we're in September, GPT-4 is coming up on one year.
[0:04:20] What has been the big surprise over the last eight years? GPT-4 has only been out for six months, which is just a good reminder about how fast things have been happening. The biggest surprise is just that it's all working. When you start off on a scientific endeavor, you hope it'll work. You let yourself dream that it'll work. You kind of have to have that conviction and the optimism.
[0:04:58] When we got together at the very beginning of 2016 and said, all right, we're going to build artificial general intelligence, that's great. But then you meet cold, hard reality. And we had a lot of stuff to figure out. And figuring out any new science is always hard. And in this case, for a bunch of reasons, it was particularly hard. We had conviction in a path.
[0:05:31] Laid out by my co-founder and our chief scientist, Ilya. But having that conviction and then actually getting it to work, this was, I think, something that very much consensus wisdom in the world was this was not going to work. And it's just through the effort of a lot of enormously talented people that it did. But that's probably the biggest surprise. When did you know that this was going to be a success and that things were going...
[0:06:01] In an exceedingly right direction. Sometime a little bit after GPT-2, which would have been 2019 at some point. Now, as you kind of see GPT-4 maybe as kind of a North Star going forward, what's the next big step for OpenAI to kind of get to where you want to get it to? Two things. One, on this current hill that we're climbing with the technology of the GPT series, we're going to keep making it better.
[0:06:34] We'll make it more reliable, more robust, more multimodal, better at reasoning, all of these things. And we also want to make sure that it is really useful to people and all the ways that it's transforming things keep happening. So, you know, we're now deep into the phase of the enterprise really adopting this technology and getting the systems to be very secure, very highly trusted, handle data appropriately.
[0:07:01] Not hallucinate, or at least not when you don't want them to hallucinate. That's leading to enormous transformation for a lot of companies. And we want to keep doing that. And then the other thing we're going to do is figure out the remaining research from this paradigm we're in right now to something that we could all truthfully call AGI. What's been the most complex part of dealing with the hallucination problem? Well, there's a lot of technical challenges. And then we could talk about those.
[0:07:38] One of the sort of non-obvious things is that a lot of the value from these systems is heavily related to the fact that they do hallucinate. If you just want to look something up in a database, we already have good stuff for that. But the fact that these AI systems can come up with new ideas, can be creative, that's a lot of the power. Now, you want them to be creative when you want and factual when you want, and that's what we're working on. But if you just sort of do the naive thing and say,
[0:08:10] Never say anything that you're not 100% sure about. You can get a model to do that, but it won't have the magic that people like so much if you do it the naive way. When you think about your own company and how that kind of reflects back to you, I know how focused you are on trust and how focused you are on kind of the deep ethical backbone of OpenAI. What are the core values of OpenAI going forward? What things in priority do you think are most critical for you to focus on?
[0:08:41] Smartest, most capable, most customizable models out there because we so believe in terms of what this will enable for humanity and that the good will be orders of magnitude more than the bad. But we also want to make sure that our models are aligned and that enterprises can trust us with data and that we're very clear about our policies around that and that we're keeping not only the data secure but also
[0:09:12] All of the learnings that a model has after ingesting a company's or a person's data. And so keeping the model very secure itself. And that balance of continuing to push on the capabilities, but also making sure that from a safety and privacy perspective, we are keeping pace with the capability. We want to continue to lead on both of those.
[0:09:40] How are you managing between alignment and intelligence? This is a very common question. Are we doing enough on alignment relative to capabilities? Should we be slowing down work on capabilities to work more in alignment? I think the frame of it is nonsensical. It really is the same thing. What's happening here is we have this amazing new piece of
[0:10:12] The human tech tree called deep learning. And that approach is solving lots of problems. And that same thing, which is helping us make these very capable systems, is how we're going to align them to human values and human intent. And when we make something that is considered an alignment breakthrough, RLHF, which is how we take the base model and get it to follow the instructions that a human user has, that's an alignment technique, but it leads from the model
[0:10:42] It leads to the model going from not very usable at all to extremely usable. And so in some very important sense, that's a capabilities thing, a capabilities gain as much of an alignment gain. Same thing with interpretability. We can look into what these networks are doing. That does for sure help with safety in all kinds of ways, but it also helps with capability. And so there's...
[0:11:04] There's more one dimensionality to the progress than people think, and we think about it like a whole system. We have to make a system that is capable and aligned. It's not that we have to make a capable system and then separately go figure out how to align it. That's definitely been a surprise to me to see how you've led through RHLF and all these different kinds of next generation alignment models, got them to work, invented, I think, many of these techniques as well.
[0:11:31] That we're not really part of the a constant and kind of continuous AI narrative even as I would say recently as the late 1990s. We didn't end then into the I would say first part of the 21st century. I think when I read some of the textbooks that are like even in 2015, 2019, the models that you've invented aren't even mentioned. Are you surprised at how much core research you're doing? No, we thought we were going to have to do a lot.
[0:12:02] We call ourselves a research lab, and we very much are. The sort of standard way that tech companies work is you start a product company. At some point when you're generating enough profit, you're like, I'm going to start a little research lab on the side. The challenge there is they don't usually go very well. We started as a research lab, and then we started a little product effort on the side, and the challenge is we have to make that go very well. What's been the scariest thing that you've seen in the lab?
[0:12:35] Can I finish one more thing on that? That was a really good question. Come on. I'll get there. In terms of how we make progress and are we surprised, not really. We are empiricists. We know we're going to be surprised. We know that reality has a different way of playing out than you think it's going to when you sit there and make your beautiful theories in the ivory tower. So we're never surprised, and we just try to meet reality where it is and follow the technology where it can go.
[0:13:06] And I think that more than any other, well, there's one more thing, which was we're extremely rigorous and careful, combined with a lot of team spirit. And that combination, I think, is hard to get a whole company to be rigorous about every part of a system that has to come together. But those are two of the things that I think are special about OpenAI. Scariest thing we've seen in the lab? Honestly, nothing super scary yet. We know it'll come.
[0:13:36] We won't be surprised when it does, but at the current model capability levels, nothing that scary. When you were in India, I was listening to one of your interviews, and I was really taken with something that you said about intelligence, that you said that as you look at these large language models and specifically how they're coming up with their answers, that you think it speaks actually or is a reflection back on human intelligence and maybe what human intelligence means. What were you trying to say? I don't remember exactly, but probably that...
[0:14:11] Intelligence is somehow an emergent property of matter to a degree that we probably don't contemplate enough. And that it can happen with electrical signals flowing through neurons that are reconnecting in certain ways. It can happen with electricity flowing through silicon. But it's something about the ability to recognize patterns in data.
[0:14:42] It's something about the ability to hallucinate, to create, to come up with novel ideas and have a feedback loop to test those. And as we study these systems, which are easier to study than the human brain for sure, there's no way we're going to, without doing a lot of collateral damage to you, there's no way we're going to go figure out what every neuron in your brain is doing. But we can look at every neuron in GPT-4 and look at every connection.
[0:15:09] Let's go a little bit step deeper, because I think when you talked about that in India, I was really surprised. You're talking about intelligence as this emergent property of matter. Do you feel that that's what's happening then in the lab, that you're seeing this kind of wake up, that you're starting to see the kind of intelligence emerge from this software that we have inside the people in this auditorium?
[0:15:37] Well, I don't think we're seeing anything wake up, but I do think we are seeing intelligence emerge from a very complicated computer program. And how does that go forward? I think the current GPT paradigm, we know how to keep improving. And we can make some predictions about, we can predict with confidence it's going to get more capable, but exactly how is a little bit hard. Like when, you know, why a new capability?
[0:16:17] emerges at this scale and not that one. We don't yet understand that as scientifically as we do about saying it's going to perform like this on this benchmark. But I suspect, I'm pretty sure, that there are major new ideas to still discover. And that if we assume that this sort of GPT paradigm of the world is the only thing that's going to happen, we're going to be unprepared for very major new things that do happen.
[0:16:49] Like what kind of things? The one that I would be tempted to say is the most important is the ability to reason. GPT-4 can reason a little bit in some cases, but not in the way that we mostly would use that term. When we have models that can discover new scientific knowledge at a phenomenal rate, if we let ourselves imagine a year where we make as much scientific progress as a civilization,
[0:17:20] As we did in the previous decade or even the previous century. And think about what that would do to quality of life, what's possible. That's pretty transformative. I thought it's really interesting, you know, thinking about intelligence as an emergent property of matter. And that I feel like there is this unremarkable but also highly correlated connection between like working with these chatbots.
[0:17:50] But also thinking there is some human aspect of it. Is it coming mostly through the training or do you think it's actually inherently there? The human aspect of talking to a chatbot? I think it's coming mostly through the training. We're really training this in some way to be all of humanity. That's what it trains on is the output of a huge fraction of humanity. And what would be the next step? Well...
[0:18:28] We can talk about the obvious ones and the ones that are going to be more speculative. So the obvious ones are just the models are going to get dramatically more capable. It will be dramatically more customizable and dramatically more reliable. And really, I think that those, you know, the model itself, that is the fundamental enabler of everything else. We will continue to build the features around the model, like we were talking about earlier for enterprise.
[0:18:58] And we'll continue to build consumer products like ChatGPT to make it easier for people to just start playing around. But in the same way that the internet and then mobile just kind of seeped everywhere, that's going to happen with intelligence. And right now, people talk about being an AI company. And there was a time after the iPhone app store launched where people talked about being a mobile company.
[0:19:26] No software company says they're a mobile company now because it would be unthinkable to not have a mobile app. And it'll be unthinkable not to have intelligence integrated into every product and service. It'll just be an expected, obvious thing. And companies will have their AI agents that can go off and do things and customers can interact with and all sorts of other things that we're seeing happen right now. But this will be a big shift in terms of
[0:19:56] How we interact with the world and with technology. When you think of dramatically more capable models, what is a dramatically more capable model? Can you try to describe it to us? I think it'll happen in all sorts of different ways, but one example, a lot of people use ChatGBT to help them program. And, you know, maybe they say, hey, it writes 25% of my code. And then it'll get up to 50, and then 75, and then 87 and a half.
[0:20:34] And then 90, and then 95. And at some point, it's not just doing more code, but it's letting you do things you just couldn't do before. I'm a big believer that quantitative shifts lead to qualitative shifts at some point. And so if you have better tools, if you can operate at a higher level of abstraction, if you can keep more of the big picture problem in your mind at one point,
[0:21:03] You can just do dramatically more. And that's difficult to say exactly what that's going to be like. But if you, you know, we could like reach back into history for an example. And if you think about the kinds of problems that a programmer let him or herself dream about when they were working with punch cards versus what you can think about with a high-level language of today versus what you can think about where you can just say in natural language, this is what I want a program to do. I don't know, how about this? Actually, that's a really interesting idea.
[0:21:33] That I just saw emerge here. How about this? And the cycle time and the sort of iterative feedback loop is so different than it is today. That will change what a single programmer is capable of. That will change what a single person running a one person company is capable of. And I'm tremendously excited to see. I remember in one specific dinner that we were having, I was describing all the different things.
[0:22:01] That I wanted to do a Salesforce. Some of them that we just showed at the Dreamforce keynote. And then other things. And other things. And then I just kind of shook my head. And I just said, I just don't know how. I'm going to hire all the people to do this work. And then you turned to me and went, Mark, you idiot. You're not going to have to do any of that. The computer is going to do it all for you. Is that what you're saying? Not quite. Which part? That I'm an idiot? What I'm saying is that the capability.
[0:22:33] The amplification of one individual's capabilities. So one person with a really good idea and a deep understanding of what a customer needs is going to be able to execute on that with what would have taken complex, many, many person teams before. And that ability to give people these tools
[0:23:00] And let things happen with less resistance, with less friction, faster, easier. It's very easy to kill a good idea. It just takes one person to be a little bit less than supportive. They don't even have to say, all right, no. Creative ideas are very fragile things. And so giving someone the ability to do more, I think is going to lead to a big shift. On a bigger message, I think a lot of families try to encourage their kids to go and become computer programmers.
[0:23:29] And take coding classes and coding camps and all of these kind of things. Do you think those kids are going to come back and really be quite upset at their parents? I remember my grandma trying to tell me in like second grade or something that was really important. I paid attention and learned cursive in like class. I don't even think they still teach it. Probably they don't. But I was like, I have very bad handwriting and that was a real struggle for me anyway. And, you know, I like cheerfully went through it. I knew at the time that this was like, no way this was going to be important because I was already like typing on a computer.
[0:24:02] But I'm not upset about it. When you watch all these kind of unusual movies about artificial intelligence, you know, Minority Reporter, Hal, or Her, or War Games, or all these various things, you know, there's been so many of them. You know, which one is your favorite? I like Her. Which part of Her do you look at and go, this is never going to happen?
[0:24:38] I mean, a ton of it. I think that's not the fair question. Because looking back at sci-fi, it's always easy. Can you just fill in that last part that, you know, which part is going to happen, which part do you think will never happen? Yeah, yeah. But I think it's unfair to dunk on old sci-fi for all the parts that got wrong. And we can cover that. But that's always true. What's amazing is that people get it right at all. And the number of things that I think her got right that were not obvious at the time.
[0:25:07] The whole interaction model with how humans are going to use an AI, this idea that it is going to be this conversational language interface, that was incredibly prophetic and certainly more than a little bit inspired us. It's not just like a prophecy. It's like an influenced shot or whatever. But I think this idea that we all have a personalized agent trying to help us and we talk to it.
[0:25:36] Like we talked to ChatGPT. That was actually not what most movies... I mean, most movies thought if we interacted with an AI at all, it was going to be like robots shooting us or something. Do you think that that will be happening soon? I do not. I realize it is a compelling movie, but no, I don't. I think it'd be great for Hollywood to have some new tropes. But yeah, I think her got something deeply right on the interface.
[0:26:07] And that is no small feat. One of the hobbies that you have, and we've talked about it, is you think about if you actually had to prepare for a moment in life when you had to kind of escape and kind of protect yourself. Where did you find that interest in that kind of idea? I mean, I think like a lot of little boys, I thought Boy Scouts was really cool, and I like outdoor stuff and survivalism.
[0:26:39] I have no delusions that any of that is going to protect us or me or you or anyone if AI goes wrong. And I sort of think it's like silly that other people assume that's what it's for. It was just like a boyhood hobby that stuck. So you're not escaping to one of these Indonesian islands now that you have a golden visa and that's your new... The speed of light is really fast. I think the AI can get there much faster than I can get there.
[0:27:10] You know, you have such an interesting background in life, and the first time we ever had a meeting was because you came to me and you had some political interest, and we were at my home and we were talking about your political interest, and it was many years ago, and you were very, very young. You know, where did that political interest come, and where did it go, and is it still in there? I love California.
[0:27:39] I would like to spend my whole life living in California. I was concerned then, I'm concerned now about what's happening with the state. I think everyone has some civic duty to think about what they can do. So I was like, you know, lightly looking into it. But I don't think I'd be a good politician at all. And I think I have something else to work on that's really important to me. I think you'd be very good. I don't think so. That's really nice of you to say. I don't.
[0:28:08] I wouldn't say I'm like a super self-aware person. You had some interest around that. I had some interest in seeing the state and the country at this point, too, get better. But I think we've all got to figure out the way we can contribute. I think the way I can contribute is scientific and technological progress. But I do think we need to get more people to jump into politics. Well, you're a very young man, and when you're an even much younger man, you did have a very...
[0:28:40] Big political interest. And that was before you started OpenAI. And now you have. You've obviously now come out on a global stage. It's clear to people you're the visionary capabilities that you have and leadership capabilities that you have. Tomorrow, you're going to go back to Washington, D.C. You've already been out there quite a few times. Tomorrow, you're going to be interacting, I think, mostly with the Senate. So as you meet with maybe just talk about the United States for a second.
[0:29:09] Dealing with the White House or the Congress or the Senate and trying to explain to them what's happening here in San Francisco and the technology revolution that you're leading. What is your message to them and what has been your biggest surprise dealing with them? Pleasant surprises for the most part. I think our leaders are taking this issue extremely seriously. They understand the need to balance the potential upsides and the potential downsides.
[0:29:43] But nuance is required here. And I did not go in with particularly high hopes about the balance of that, about that nuance being held appropriately. And it really is. I don't know exactly how this is all gonna play out, but I know that people seem very genuine in caring and wanting to do something, wanting to get it right. And the message, I think our role is just to try to explain as best as we can.
[0:30:11] Realizing that we don't have all the answers and we might be wrong. Technology takes weird turns sometimes. What we think is going to happen, where the control points can be, where the dangers might lie, what it takes to make sure the benefits are broadly distributed, and then let the leaders decide what to do. When you have these discussions with these political leaders, which part of it has been the most shocking or surprising to them? The first time that they see, you know, GPT-4, they're...
[0:30:44] Usually exceeds their expectations. I would say it's the technology demo and realizing that this thing that people have been talking about for a long time is here. Has that also been true on the international leaders as well? Yes, although by the time of that trip, most people had by then already played around with it to some degree. When we did that trip, it was sort of like ChatGPT had already had its news cycle. If you would have one goal...
[0:31:18] With the US government between now and let's say next year when the elections start and potentially there'll be changes in administrations and so forth, what would be the goal do you think for the next 12 months from a political perspective? I think getting a framework in place that can help us deal both with short-term challenges and the longer-term ones.
[0:31:41] Even if it's imperfect, starting with something now would be great. It's going to take people a while to learn. I think solving this legislatively is actually quite difficult given the rate of change. I think an agency, a new agency would probably make the most sense. But getting something going so that even if it's just focused on insight and not oversight so that the government can build up the muscle here I think would be great. So are you suggesting the CIA should change their letters into the CAI?
[0:32:13] I mean, I can imagine worse agencies of less competent people to hand it over to, but I really do think a new one would be appropriate. What part of all of this are they just getting really wrong? This is a very human thing. I mean, I don't just think it's governments getting this wrong. I think most of us get it wrong to different degrees, too. You show people an exponential curve of technology, and they believe you that it's been exponential, but they don't believe you it's going to keep going exponentially. They believe it's about to level off.
[0:32:50] And that happens in all sorts of subtle and not subtle ways. And it's a very difficult bias to overcome, particularly when if you accept it, it means that you have to confront such radical change in all parts of life. And so what are you doing to help them understand that? I mean, a bunch of things. One is, I believe, in the power of repetition here. Another is to just keep showing up every year.
[0:33:29] And say, hey, that thing you thought was impossible or that thing you said, oh, well, I don't really have to worry about this. Your last model couldn't do it. Try our new one. See what happens. The last few times I've arrived in the United States, I noticed that the customs also has now a camera. And sometimes they don't even ask for my passport. They just put up the camera in front of me and the AI is able to know it's me. We know in other countries, especially certain Asian countries,
[0:33:57] That there's a lot of that type of technology already in place. Do you think we're moving into a surveillance economy? Do you think that this technology will accelerate the move into greater levels of surveillance? I do. One of the things that I struggle with is I don't see a world where if AI is as powerful as we think and people can do significant harm with it.
[0:34:27] I don't see a world where we don't have less surveillance, and I don't think that's a good thing, but I have talked to a lot of people about this and have not yet heard a great solution. I know that you like to just kind of get away and maybe get on a small boat somewhere, like a sailboat, and be with a friend or two and just sail by yourself and completely disconnect. When do you think it'll be that you're going to take that trip
[0:34:57] That sailboat will not be quite so disconnected and will know where you are no matter where you are. And maybe there'll be some autonomous characteristics to the boat as well. Well, we have Starlink now. That's, in some sense, you know, that's a transformative thing. I think even more than people realize. I'm okay with that. I'll take that trade for sure. But I think this sort of idea that I'm going to be like off on a sailboat, unreachable, unless you really want to be.
[0:35:31] That's over. When you now look out at your product pipeline for the next couple of years with OpenAI, you mentioned a few things. You especially mentioned these remarkable new models. You used the example of the computer science model and the ability to write any code and maybe become a whole engineering team. Can you take another industry or another application area and give us an example of something else that you think will be a remarkable achievement in terms of a model moving forward?
[0:36:03] I'll mention a few of my favorites. What's happening with education is incredibly gratifying to us to watch. And the ingenuity of teachers, of entrepreneurs building ed tech companies, and also just of students themselves finding new ways to use ChatGPT to learn is quite remarkable. We know what a difference it makes to a student to have a great one-on-one tutor. It's like two standard deviations, better outcome or something.
[0:36:32] Society before AI just like realistically was not going to be able to afford that. And so we didn't even let ourselves think of how we could deliver it. And now I think we see a path where with a combination of humans and AI together, we can offer this to everybody in the world. And if we can deliver on that, and again, it looks to me like we're going to be able to with the technology, that will be a transformation for the world that...
[0:37:00] I think would be a huge triumph. In healthcare, I think we can imagine a world where not only do these systems go help us cure a lot of diseases, but also that the actual healthcare product we deliver to a person, again, this hybrid of a doctor and AI together, means we can just offer something far beyond what we think of as even possible today. In creative work, it's actually quite remarkable to see what's happening.
[0:37:29] What a visual artist, for example, can now do with the latest image generation tools, we're all going to get better art than we've ever had before. Are you especially surprised to see that level of creativity in art and how it's connected back to these systems? In some sense, yes. If you had asked me to predict 10 years ago the order that AI was going to disrupt industries, I would have said physical labor first, cognitive labor second, and maybe...
[0:38:04] Maybe never, but certainly last, creativity. And I hadn't actually thought that hard about it, but it's like what all the experts were saying. It was what my college professors said, whatever. But it was what everybody thought. And for creativity to go first, which clearly it has, is an update. But also, as we started working on it, and we kind of saw where the system strengths were and where the weaknesses were,
[0:38:33] It's a kind of area that can tolerate the flaws in our current systems quite well. Yesterday, we gave a $20 million grant to our local public schools to help them augment their understanding of AI, and we did some seminars. And the one in high school that had the most attention was on music, and kids were just fascinated with how the AI was able to work with them with music, but also wanted to understand all the legal aspects, the trademark issues.
[0:39:02] And brand issues and everything else associated with music. When you look at the music industry overall, are you surprised to see kind of this dramatic change in music and the level of interest in AI music from kids? I mean, not at all surprised on the interest. People love music, and it's a really important part of life. And what people can do with these tools is great. Unfortunately, I think...
[0:39:29] The music industry has a reputation such that companies just don't want to deal with it and are focused on other areas. And I think that's a loss. Because like we're seeing with visual art, I think this will be a tool that amplifies humans, not replaces them. All right. Well, thank you, Sam. Thank you very much. Please thank Sam Altman, would you? What a great conversation. Three takeaways for me. First, that creativity and hallucinations are two sides of the same coin.
[0:40:04] And just to appreciate that. Number two, even Sam didn't expect AI to be so capable in creative tasks. And so a lot of these technologies, they're going to continue to surprise us. Last but not least, I was struck by how Sam describes himself as an empiricist. OpenAI was not the first to develop large language models, but because of their pragmatic approach, they were able to go to market first with ChatGPT. That's it for this episode of Ask More of AI, the podcast at the intersection of AI and business.
[0:40:33] Follow us wherever you get your podcasts and follow me on LinkedIn and Twitter or X.