Sam Altman Says Technology Is the Greatest Equalizing Force in Society | The Tonight Show (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:02) -Sam, you've been considered one of the leading figures in AI. (00:00:05) You co-founded OpenAI, which led to ChatGPT, (00:00:09) which is celebrating three years now, ChatGPT. (00:00:13) When all this is happening -- First off, what is ChatGPT? (00:00:20) Can we just do that for anyone? (00:00:21) My dad's probably watching. -It's like an AI chatbot. (00:00:23) You ask a question, it tells you what to do. (00:00:25) -And it tells you what to do. (00:00:27) And so it's like Google Search, (00:00:29) but you made it into more of a thing (00:00:30) that you can have a conversation with. (00:00:32) -Yeah. You know, it's gotten, like, much smarter (00:00:34) over the last three years. (00:00:35) So, it used to not do very much. (00:00:36) Maybe it used to only do Google Search, (00:00:38) and now it can, like, really do a lot of stuff. (00:00:40) People are asking it for, like, healthcare advice, (00:00:42) how to learn stuff, they're doing work with it. (00:00:43) -Yeah, tell me, what are the pros of ChatGPT? (00:00:48) -The number of people that reach out to us and are like, (00:00:50) "I had this crazy health condition. (00:00:52) I couldn't figure out what was going on. (00:00:53) I just put my symptoms into ChatGPT (00:00:55) and it told me what test to ask the doctor for, and I got it, (00:00:57) and now I'm cured." (00:00:58) Like, that's great. That happens a lot. (00:01:00) -Wow! (00:01:01) -You can definitely learn anything. (00:01:02) -Anything? -Pretty much. (00:01:04) I haven't found anything I can't learn. (00:01:06) You can do your work way faster. (00:01:07) You can use it instead of Google. (00:01:09) You can use it for, like, advice about your life. (00:01:10) People talk about all the people in their life (00:01:12) and trying to figure out what to do. (00:01:13) It's -- The idea is it's (00:01:15) like a general-purpose sort of life adviser. (00:01:17) -Yeah. And I know that you and your husband just had a baby. (00:01:20) Congratulations. -We did. Thank you. (00:01:21) -Yeah. -[ Cheers and applause ] (00:01:22) -Thank you. -How old is he? (00:01:24) -Eight months. -Eight months old? (00:01:25) -Yeah. -Are you loving him? (00:01:27) -So much. I have no non-cliché thing to say, (00:01:30) but it is by far so much better than anything anyone tells you. (00:01:34) I am just so incredibly happy. (00:01:36) -Is he into trucks? (00:01:37) -Yeah, he loves trucks. (00:01:39) We have to, like, spend a lot of time watching the garbage truck (00:01:40) and driving around. (00:01:41) He does love trucks. -He loves garbage trucks, too? (00:01:43) -Cars. Anything. (00:01:45) We watched, like, the F1 season finale this week. (00:01:47) And it was the first time he got to watch TV. (00:01:49) And he was like... -Oh, really? (00:01:51) -Really, yeah. -He's thrilled. (00:01:52) And do you use ChatGPT when raising your baby? (00:01:55) -I do. I mean, I feel kind of bad about it (00:01:59) because we have this, like, genius level at everything, (00:02:02) intelligence, sitting there, like, waiting (00:02:04) to unravel the mysteries of humanity. (00:02:05) And I'm like, "Why does my kid stop (00:02:07) dropping his pizza on the floor and laughing." (00:02:09) -Yeah. Yeah. -Like, you know? (00:02:10) So I feel like I'm not asking a good-enough question, (00:02:13) but it is -- I don't -- I cannot imagine having gone through, (00:02:18) like, figuring out how to raise a newborn without ChatGPT. (00:02:20) Clearly, people did it for a long time no problem. (00:02:22) -Yes. -But... (00:02:24) [ Laughter ] (00:02:26) So I know, like, clearly, it was possible. (00:02:28) -Yes, it was possible. (00:02:30) -But I have relied on it so much. (00:02:33) I mean, it's obviously, like, the most important thing (00:02:34) to happen in my life, so it's top of mind (00:02:36) and I use it all the time. (00:02:37) But, like, a couple of months ago, I was at a party. (00:02:40) I was talking to this guy, and, you know, (00:02:42) we had kids about the same age, so we were talking about that. (00:02:44) And, you know, he was like, "Oh, you know, what's your kid --" (00:02:46) Normal conversation. (00:02:48) And he's like, "Yeah, he's six months old. (00:02:50) Once they start crawling, it's everywhere. (00:02:52) They go everywhere." (00:02:54) And I was like, "My 6-month-old can't crawl yet." (00:02:55) And he was like, "You sure about that? (00:02:59) You think he's okay?" (00:03:00) And I was like, "Well, I think he's, like, about to start. (00:03:03) I'm sure he's doing fine. (00:03:04) And, by the way, I bet my kid can (00:03:05) do a lot of things yours can't, so it's alright." (00:03:08) But then I, like, ran to the bathroom and I was like, (00:03:10) "Do I need to take my kid to the doctor tomorrow morning? (00:03:13) Is this okay?" (00:03:14) -"Should every kid be able to walk at six months?" (00:03:16) -Yeah. Crawl. -Crawl. (00:03:18) And, I mean -- Yours can. -No, no. (00:03:21) Yeah, mine walked at -- Yeah. (00:03:24) Mine on "Dancing with the Stars" at seven months. (00:03:26) -There we go. -Semifinalists. No big deal. (00:03:30) -But it was a really -- (00:03:31) I got this, like, great answer back, which was like, (00:03:33) "Of course it's normal. (00:03:35) Of course you don't need to go to the doctor. (00:03:37) You know, parents do all these sorts of things." (00:03:39) And, by the way, it's personalized. (00:03:42) Like, ChatGPT gets to know you. (00:03:44) "And, you know, by the way, you're the C.E.O. of OpenAI. (00:03:46) You probably are around, like, all these high-achieving people. (00:03:48) Maybe you don't want to project that onto your kid. (00:03:50) And you should just, like, relax and he'll be fine." (00:03:52) And, you know, whatever. And I was like, "Oh, okay." (00:03:55) -I've seen you in a couple of interviews, too, where you were (00:03:57) saying -- People were asking you about ChatGPT, (00:03:59) and you're like, "Well, it's kind of interesting, (00:04:01) because rich people have these great, (00:04:03) amazing doctors and lawyers and business managers, you know, (00:04:06) whereas people that don't make enough -- as much money (00:04:09) as these people, they don't get those opportunities." (00:04:12) And what ChatGPT does, they have the same -- (00:04:13) You can get the same answer. (00:04:15) You're just typing in these questions. You'll get the best. (00:04:18) -No, I think this is a cool thing about technology. (00:04:19) I think there's many downsides to technology, (00:04:21) but there's a lot of upsides. (00:04:23) And one of the upsides is, (00:04:26) it is a sort of equalizing force in many ways. (00:04:28) You know, I remember people used to say this (00:04:30) when the iPhone came out. (00:04:31) -Yeah. (00:04:32) -The richest, most powerful person in the world (00:04:35) got the same piece of hardware that, you know, (00:04:37) billions of other people got. (00:04:38) And I think AI is pushing in the same direction, (00:04:42) and it'll take some work to ensure it goes in this way, (00:04:44) but it should be, like, a good force for society. (00:04:47) -So, what about ChatGPT now? What is the cons? (00:04:50) What are we worried about? (00:04:53) What are the things that people don't like about it? (00:04:56) -Um...I mean, a long list of things. (00:04:59) But one of the things that I'm worried about (00:05:03) is just the rate of change (00:05:04) that's happening in the world right now. (00:05:06) You know, this is a three-year-old technology. (00:05:08) No other technology has ever been adapted -- (00:05:10) adopted by the world this fast. (00:05:12) It is a truly general-purpose thing. (00:05:15) The same thing that can figure out how to go cure (00:05:17) a bunch of diseases can be used in a bunch of negative ways. (00:05:21) The rate at which jobs will change over may be pretty fast. (00:05:24) I have no doubt that we'll figure out all new jobs to do (00:05:27) and sort of, I hope, much better jobs. (00:05:30) But the speed with which this is happening, (00:05:32) as this just sort of, like, barrels through society, (00:05:35) and making sure that we introduce this to the world (00:05:38) in a responsible way, where people have time to adapt, (00:05:41) to give input, to figure out how to do this, (00:05:43) you could imagine us getting that wrong. (00:05:45) -Yeah. And are people on it? (00:05:47) -Yeah. Like, again, this has gone really well so far, (00:05:51) better than I would have thought. (00:05:53) If you look at the trajectory of the last three years, (00:05:54) we have integrated this into our lives, into society. (00:05:57) People use it in all sorts of really great ways, (00:06:00) and we've been able to get, so far, a lot of the benefits, (00:06:04) and I think we've done a good, (00:06:06) thoughtful job as a society of mitigating the downsides. (00:06:09) But we're still feeling our way through this, (00:06:11) and I'm sure, as the intelligence keeps ramping, (00:06:14) there will be a lot more to come. (00:06:15) -And when do you think, like, it's all going to be, like, (00:06:19) perfected? (00:06:21) -I hope never. (00:06:23) You know, I hope just we have a relentless -- (00:06:25) we, as the human species, (00:06:27) we have this relentless drive to improve (00:06:29) and make things better and to want more and to do more (00:06:30) and to create more. (00:06:32) And I hope that keeps going forever. (00:06:34) But if you just look at how far it has come (00:06:36) even this year, in 2025, (00:06:38) you know, we started this year with these models (00:06:40) that could do a little bit of math, (00:06:42) and by the end of the year, (00:06:43) we're doing better math than our best mathematicians (00:06:46) at our most difficult competitions. (00:06:49) -Where's all the information coming from? (00:06:53) -Well, at this point, it generates a lot of (00:06:55) its own information. (00:06:56) If you think about when you solve a new problem (00:06:59) or write something funny, you draw upon your past experiences. (00:07:03) You think really hard. You write something down. (00:07:05) Maybe you see how it goes. (00:07:06) You see how someone responds, (00:07:08) you see if it works, and then you build on top of that (00:07:09) and on top of that and on top of that. (00:07:11) So, we're at the point now where the models can think harder (00:07:14) and come up with new information. (00:07:16) -And what's the biggest thing you see happening out of (00:07:19) ChatGPT in the next five years? (00:07:23) -Five years is a long time. (00:07:25) In five years, I think a lot. (00:07:26) Next year, I hope we'll start (00:07:27) to see these models really make (00:07:30) small-but-important new scientific discoveries. (00:07:33) And in five years, I hope they're curing diseases. (00:07:35) -Thank you so much for coming on and talking about this stuff. (00:07:38) I appreciate this, and please... -Thanks for having me. (00:07:41) -...hug that little baby of yours. (00:07:42) -I will do that. -Sam Altman, everybody. (00:07:43) We'll be right back with more "Tonight Show." (00:07:45) Stick around.