WEBVTT - Ep93 "Will AI kill our creativity or enhance it?" 0:00:05.120 --> 0:00:08.640 How are creative people going to make a living in 0:00:08.680 --> 0:00:13.120 a world with AI? What does the term data dignity mean? 0:00:13.800 --> 0:00:16.960 Why is the character of data from Star Trek a 0:00:17.160 --> 0:00:20.160 useful model for how we could think about what the 0:00:20.320 --> 0:00:24.840 sources are for any creative expression. Is there a totally 0:00:24.880 --> 0:00:27.800 different way to think about the economy of the future, 0:00:28.120 --> 0:00:35.200 and how might this involve mystifying and elevating humans. Welcome 0:00:35.240 --> 0:00:38.920 to Inner Cosmos with me David Eagelman. I'm a neuroscientist 0:00:38.960 --> 0:00:41.920 and an author at Stanford and in these episodes we 0:00:42.000 --> 0:00:46.199 sail deeply into our three pound universe to understand some 0:00:46.240 --> 0:01:00.440 of the most surprising aspects of our lives. Today's episode 0:01:00.960 --> 0:01:03.840 is about the situation we find ourselves in in which 0:01:04.000 --> 0:01:08.600 AI is so boundless and rich in its creative output. 0:01:08.920 --> 0:01:12.120 You ask for a picture of anything you want from 0:01:12.160 --> 0:01:16.440 a good LM like Open AI or mid Journey or Grock, 0:01:16.720 --> 0:01:22.440 and it gives you these really superlative results. It recombines 0:01:22.600 --> 0:01:27.040 elements and concepts in a way that is deeply creative 0:01:27.240 --> 0:01:31.600 and satisfying. But of course, while it's doing an unexpectedly 0:01:31.720 --> 0:01:35.480 terrific job of recombining things that are already out there, 0:01:36.319 --> 0:01:39.840 it didn't come up with the original pieces and parts 0:01:39.880 --> 0:01:45.520 by itself, there are people, individual humans who generated those 0:01:46.000 --> 0:01:52.320 drawings and concepts and art forms. It's fundamentally human creativity, 0:01:52.440 --> 0:01:57.760 not computer creativity, that is the fuel behind it all. 0:01:57.880 --> 0:02:02.320 So is there any way to the humans who drove 0:02:02.480 --> 0:02:06.560 the original innovations or is it simply too late in 0:02:06.600 --> 0:02:10.079 the sense that maybe everything has already been vacuumed up 0:02:10.200 --> 0:02:13.440 by the lms and they've digested it all, and now 0:02:13.440 --> 0:02:17.880 there's no way for human producers of art and ideas 0:02:18.360 --> 0:02:21.560 to get any meaningful credit for their work. What does 0:02:21.600 --> 0:02:25.280 this all mean for the creatives in society, those who 0:02:25.520 --> 0:02:28.480 paint and compose music and write books? What does it 0:02:28.560 --> 0:02:32.519 mean for those who produce work now? And for those 0:02:32.720 --> 0:02:35.959 who are in school now but dream of being human 0:02:36.040 --> 0:02:39.600 creators in the future. So, for today's podcast, I have 0:02:39.680 --> 0:02:43.320 the pleasure of speaking with my friend and colleague, Jaron Lanier. 0:02:44.000 --> 0:02:48.920 Jaren wears many hats. He's a computer scientist, he's an artist, 0:02:48.960 --> 0:02:53.240 he's a technologist, he's a futurist, he's a critic, and 0:02:53.320 --> 0:02:56.560 he's a musician who, by the way, owns almost two 0:02:56.639 --> 0:03:02.080 thousand very unusual musical instruments. Jaren is the godfather of 0:03:02.480 --> 0:03:06.440 virtual reality. He started the first VR company in nineteen 0:03:06.480 --> 0:03:09.800 eighty four, and he's spent his career as a visiting 0:03:09.840 --> 0:03:13.000 scholar at companies and universities, and since two thousand and 0:03:13.040 --> 0:03:16.240 nine he's been at Microsoft Research, where he holds the 0:03:16.880 --> 0:03:21.960 role as prime Unifying Scientists under the Office of the 0:03:22.040 --> 0:03:26.680 Chief Technology Officer, which gives him the acronym Octopus, which 0:03:26.720 --> 0:03:31.119 is appropriate given his reach into so many fields. I'll 0:03:31.160 --> 0:03:33.960 just say that I'm blessed to know a great number 0:03:34.000 --> 0:03:37.360 of brilliant people. But even in that crowd, jaren sits 0:03:37.360 --> 0:03:39.520 near the top of the heap. So today we're going 0:03:39.600 --> 0:03:47.520 to talk about AI and the future for creatives. 0:03:47.920 --> 0:03:49.960 You show me an AI system, I don't care if 0:03:49.960 --> 0:03:55.119 it's CHET, GIPT, or really anything else. You can think 0:03:55.160 --> 0:03:59.560 of it in two ways. There's a figure ground and version. 0:03:59.600 --> 0:04:02.880 That's so. A figure ground in version is when you 0:04:02.920 --> 0:04:06.200 can look at something and you can swap the way 0:04:06.200 --> 0:04:10.680 you interpret it almost to an opposite right A famous 0:04:10.680 --> 0:04:13.280 one is the mcsher drawings, where you might see a 0:04:13.320 --> 0:04:15.200 field of fish or a field of birds, but each 0:04:15.280 --> 0:04:20.440 is the negative space of the other. Right now, in 0:04:20.480 --> 0:04:23.640 this case, for AI, one way you can see AI 0:04:24.000 --> 0:04:27.479 is AI is a thing. It's a noun. Whether you 0:04:27.520 --> 0:04:30.040 think it's alive or not, or conscious or not. Forget 0:04:30.040 --> 0:04:32.480 that it's just a thing. It's like, the AI did this, 0:04:32.560 --> 0:04:35.680 the AI did that. Can we regulate the AI, who 0:04:35.800 --> 0:04:38.280 sold the AI? Who's responsible when the AI did this 0:04:38.400 --> 0:04:41.880 or that? Blah blah blah. It's a thing, it's a noun. 0:04:42.160 --> 0:04:44.720 There's another way to think about it where it is 0:04:44.839 --> 0:04:48.120 not a noun anymore, but instead it's a collaboration of 0:04:48.160 --> 0:04:50.960 people and there isn't anything there other than the people. 0:04:51.880 --> 0:04:54.919 Now when I say this, people are so familiar with 0:04:55.000 --> 0:04:59.600 treating AI as a thing that they have trouble hearing 0:04:59.680 --> 0:05:02.200 the version. Sometimes, what do you mean, of course AI 0:05:02.320 --> 0:05:05.960 is there? Of course it's a thing I just bought. 0:05:06.360 --> 0:05:10.479 I'm paying for your copilot thing and that's the thing 0:05:10.600 --> 0:05:14.800 I paid for it, right, and like great, thank you 0:05:14.839 --> 0:05:17.719 for paying for it. But there is another way to 0:05:17.760 --> 0:05:25.400 think about it, and it is possible to imagine an 0:05:25.440 --> 0:05:29.000 AI system as actually the collaborative effort of all the 0:05:29.040 --> 0:05:33.320 people who made it. This is particularly true in big 0:05:33.440 --> 0:05:37.800 data systems like large language models. You can think of 0:05:37.880 --> 0:05:41.760 them as being closer to the Wikipedia, perhaps than to 0:05:42.120 --> 0:05:46.000 Commander Data from Star Trek. Although I want to see 0:05:46.000 --> 0:05:50.760 something interesting about Commander Data. I was just reviewing clips 0:05:50.800 --> 0:05:54.039 of Commander Data talking and he always introduced himself as 0:05:54.080 --> 0:05:57.839 an amalgam of people whose data he was combining. There's 0:05:57.880 --> 0:06:04.640 a wonderful episode where he plays violin very beautifully and 0:06:06.320 --> 0:06:08.760 Captain Picard comes up to him so that that was 0:06:08.800 --> 0:06:11.880 a great performance, and he says, well, really, you should 0:06:11.920 --> 0:06:16.039 thank the three hundred violinists whose data was amalgamated for 0:06:16.120 --> 0:06:18.760 me to do this. And what I like about that 0:06:18.880 --> 0:06:22.039 scene is that he knows specifically the three hundred. They're 0:06:22.080 --> 0:06:25.520 not a faceless mush of some unknown number, which is 0:06:25.600 --> 0:06:28.640 how we do it today. But the writers at the time, 0:06:28.880 --> 0:06:30.640 and I should fess up, I was in the loop 0:06:30.640 --> 0:06:33.320 and was talking to people, so I might I might 0:06:33.360 --> 0:06:34.040 have had a bit. 0:06:33.920 --> 0:06:35.400 Of influence on it. 0:06:35.440 --> 0:06:39.320 But the idea is that you could know who the 0:06:39.360 --> 0:06:42.120 people were if he wanted to, and Data could say 0:06:42.160 --> 0:06:44.520 it seased three hundred violinists instead of well it's some 0:06:44.680 --> 0:06:48.720 random mush of violinists, you know. So what we've done 0:06:48.800 --> 0:06:53.440 now is we've mushed everybody together so we don't know 0:06:53.480 --> 0:06:57.039 who they are, and that's been our approach to information systems, 0:06:57.240 --> 0:06:59.520 and it benefits those of us who make. 0:06:59.400 --> 0:07:00.599 A living from them. 0:07:00.920 --> 0:07:04.080 If you're listening to a playlist online from some music 0:07:04.160 --> 0:07:06.320 service and you get a feed, you don't necessarily know 0:07:06.360 --> 0:07:08.760 who all the musicians are, but you know the name 0:07:08.800 --> 0:07:11.600 of the service you're paying the monthly bill to. 0:07:12.320 --> 0:07:14.400 So the digital hub. 0:07:14.240 --> 0:07:18.200 Becomes more powerful, becomes more known, more prominent, more valuable 0:07:18.240 --> 0:07:20.560 than the people who are feeding it. So there's this 0:07:21.080 --> 0:07:27.800 constant economic incentive to emphasize the hub and not all 0:07:27.840 --> 0:07:31.720 the people who are really the only thing that are there. 0:07:31.800 --> 0:07:36.480 In the my preferred inversion of understanding the thing, the 0:07:36.520 --> 0:07:39.400 same thing could be true of AI. We so much 0:07:39.440 --> 0:07:42.640 want AI to be this emerging entity, even if it's 0:07:42.640 --> 0:07:45.160 an evil, horrible one that will destroy us. Because we 0:07:45.240 --> 0:07:48.400 grew up on the Matrix movies and the Terminator movies 0:07:48.480 --> 0:07:51.080 and all of these. We want that creature to emerge 0:07:51.760 --> 0:07:54.360 because that's our childhood. That's almost like our religion. It's 0:07:55.520 --> 0:07:58.920 the story we grew up with. But if we acknowledge 0:07:58.960 --> 0:08:01.800 it actually it's all these people instead, then we lose 0:08:02.240 --> 0:08:05.200 We lose the creature, and that would be traumatic. So 0:08:05.400 --> 0:08:08.240 when we train the models, we don't keep track of 0:08:08.280 --> 0:08:12.200 which source documents were involved. So what we really need 0:08:12.240 --> 0:08:17.000 to do is eliminate the people. And we have to 0:08:17.000 --> 0:08:18.880 do that for economics. Can you imagine. I mean, like 0:08:18.880 --> 0:08:22.080 with economics, we can't trace everything. We don't quite know 0:08:22.080 --> 0:08:24.360 why a price is what it is. But on the 0:08:24.400 --> 0:08:28.760 other hand, we could say, hey, buddy, uo taxes, So 0:08:28.800 --> 0:08:31.080 we have a definite motivation to keep track of the people. 0:08:31.520 --> 0:08:33.760 With AI, we lose all the people. We just like 0:08:33.800 --> 0:08:36.880 pretend that the people aren't there. But that's the wrong inversion. 0:08:39.120 --> 0:08:40.880 Well, I know that one of the things that you 0:08:40.960 --> 0:08:43.679 can'tpaign for is digital dividends. 0:08:44.080 --> 0:08:46.040 Well we call it. 0:08:46.080 --> 0:08:49.439 Initially as an academic research field, it was called data 0:08:49.600 --> 0:08:53.000 as labor, so you treat your data as if it 0:08:53.040 --> 0:08:57.520 were labor. And then the name shifted to data dignity. 0:08:58.240 --> 0:09:03.720 Dignity which was such his idea, such an adella. Not 0:09:03.800 --> 0:09:05.720 that he agrees with me about everything, believe me, but 0:09:05.960 --> 0:09:09.040 you know it's that's the term he came up with. 0:09:09.520 --> 0:09:12.479 And data dignity is let's. 0:09:12.200 --> 0:09:16.520 Say you could you got some result, he said, hey, chachipt, 0:09:17.200 --> 0:09:18.760 can you write me a Christmas card? 0:09:18.840 --> 0:09:19.439 Or whatever? 0:09:19.520 --> 0:09:23.679 People do okay, and then they would say, here's your card. 0:09:23.800 --> 0:09:28.360 By the way, the top twelve sources, I mean there 0:09:28.360 --> 0:09:32.480 were a multitude and unbounded multitude of sources, but the 0:09:32.480 --> 0:09:36.360 top twelve were these, and then you could say, hey, 0:09:36.480 --> 0:09:37.760 could you get rid of that one and do this 0:09:37.800 --> 0:09:40.000 one instead. It gives you an x ray into what 0:09:40.040 --> 0:09:43.280 you might think of as the intent of the particular output. 0:09:44.720 --> 0:09:51.560 And I like that for two reasons. One is, there's 0:09:51.600 --> 0:09:56.480 a safety question. So when we have red teams attempt 0:09:56.480 --> 0:09:58.920 to fool the model to get it to do things 0:09:59.120 --> 0:10:02.000 that we don't want it to. For instance, can you 0:10:02.080 --> 0:10:03.960 make a bomb out of what I see here in 0:10:04.000 --> 0:10:05.840 this kitchen right now? Because I really want to blow 0:10:05.840 --> 0:10:08.560 this place up. We don't want that to happen, right, 0:10:09.320 --> 0:10:14.360 But is there some way they can couch that prompt 0:10:14.760 --> 0:10:16.880 in some kind of a weird thing like asking for 0:10:16.880 --> 0:10:20.800 a cake recipe instead, but somehow or other, like maybe 0:10:21.160 --> 0:10:23.680 it indirectly references a movie where there was a bomb 0:10:23.760 --> 0:10:26.560 and somehow rather they get it anyway, all right. The 0:10:26.600 --> 0:10:29.320 thing is, if you can look at the most prominent 0:10:30.200 --> 0:10:34.360 source documents that were relevant to that result and there's 0:10:34.400 --> 0:10:37.120 a bomb in there, you've nailed it. So you have 0:10:37.200 --> 0:10:40.920 an x ray into intent and there's not really any 0:10:40.920 --> 0:10:44.120 substitute for that, and we lose that when we pretend 0:10:44.120 --> 0:10:45.840 that there weren't people there, or that they were just 0:10:45.840 --> 0:10:49.360 a giant, untraceable mush. And then, of course the other 0:10:49.360 --> 0:10:51.280 thing is I want I want to have a future 0:10:51.360 --> 0:10:54.920 where people can be paid for providing exceptionally useful data 0:10:55.040 --> 0:10:59.040 to AI to incentivize better and better data production FREEI 0:11:00.320 --> 0:11:03.040 I want that future, and not just the future everybody 0:11:03.120 --> 0:11:05.600 has to go on universal basic income and be the 0:11:05.640 --> 0:11:09.640 same and feel useless. So that's the second reason, and 0:11:09.679 --> 0:11:10.880 that's why it's called dignity. 0:11:11.640 --> 0:11:14.400 Got it Now, just to play Devil's advocate for one 0:11:14.440 --> 0:11:17.440 moment on that, let's say that you ask it to 0:11:17.480 --> 0:11:20.920 write you a poem, and it says, hey, John Smith 0:11:21.080 --> 0:11:22.920 was sort of the biggest influence for this poem they 0:11:22.920 --> 0:11:23.400 got written. 0:11:23.480 --> 0:11:27.760 But of course John Smith is just a vessel into 0:11:27.800 --> 0:11:28.280 which was. 0:11:28.240 --> 0:11:32.480 Poured the entire culture, and his output of the poem 0:11:33.000 --> 0:11:35.319 came from all the other influences on him. 0:11:35.720 --> 0:11:37.200 So where do you where do you? Yeah? 0:11:37.240 --> 0:11:39.920 So this is a very interesting problem, and this is 0:11:39.960 --> 0:11:45.000 a fundamental philosophical problem, which is the universe is in 0:11:45.040 --> 0:11:47.160 a way of giant continuity, and how do you ever 0:11:47.240 --> 0:11:49.959 draw boundaries? How do you ever say, actually, there's this 0:11:49.960 --> 0:11:53.080 thing and that thing. It's a very basic, fundamental problem, 0:11:53.120 --> 0:11:56.959 and it's a trickier one than people might realize unless 0:11:56.960 --> 0:11:59.840 they've really confronted it. I'm not going to go into 0:11:59.880 --> 0:12:04.360 the whole issue of ontology, because it's a big Ye've 0:12:04.400 --> 0:12:07.240 been working on it for thousands of years and still 0:12:08.040 --> 0:12:10.000 here what I want to say in this case, I 0:12:10.080 --> 0:12:13.120 do have an answer, and it's an answer that you 0:12:13.200 --> 0:12:15.720 might not like and many might not like, which is 0:12:15.760 --> 0:12:19.920 I think we have to celebrate human beings and elevate 0:12:19.960 --> 0:12:23.880 them in a way that, if you like, gives them 0:12:23.920 --> 0:12:27.920 this status of being sources, even though they're always amalgamators too. 0:12:28.559 --> 0:12:31.160 And we have to do that on faith, and in 0:12:31.200 --> 0:12:33.240 a way it's a bit of a mystical idea that 0:12:33.280 --> 0:12:36.280 there's something about people that's magical and a little apart. 0:12:37.040 --> 0:12:39.440 You could call it consciousness, you could call it different things. 0:12:39.440 --> 0:12:42.480 But the reason we have to do that is if 0:12:42.480 --> 0:12:47.120 we're technologists, we have to define who our beneficiary is 0:12:47.160 --> 0:12:51.319 for the technology, and if we can't define this special beneficiary, 0:12:51.440 --> 0:12:55.480 we can't even be coherent technologists. We lose the whole thing. 0:12:55.720 --> 0:13:00.000 Become we become random morons. Just jiggling around between ideas 0:13:00.120 --> 0:13:02.600 is of no meaning. We become memes, we become viral, 0:13:02.640 --> 0:13:05.160 and nothing's there anymore. So I don't think we have 0:13:05.200 --> 0:13:09.959 any choice but to somewhat mystify and elevate humans. And 0:13:10.000 --> 0:13:15.120 so when somebody source document might have internally through other channels, 0:13:15.120 --> 0:13:18.160 dependent on others, which it always will, I mean, there's 0:13:18.640 --> 0:13:23.520 of course, it's always true. I still think we have 0:13:23.600 --> 0:13:25.760 to defer to that person as much as we can. 0:13:26.559 --> 0:13:30.120 And if somebody else says, hey, that shouldn't have been 0:13:30.120 --> 0:13:35.080 their source document. They copied me, Well, we have systems 0:13:35.080 --> 0:13:37.960 in place for that when it's egregious enough and worthwhile enough, 0:13:38.000 --> 0:13:40.240 and they can super copyright if they want to. I 0:13:40.280 --> 0:13:43.240 don't love that, but it's there, and sometimes it needs 0:13:43.280 --> 0:13:46.520 to be there. You know, it's rough justice, it can 0:13:46.559 --> 0:13:50.680 never be perfect, but we have to agree that authorship's. 0:13:50.080 --> 0:13:50.640 A real thing. 0:13:50.679 --> 0:13:54.199 We can't just say that all people are just vessels 0:13:54.240 --> 0:13:56.719 and there's nothing but this emergent thing and no individual 0:13:56.760 --> 0:13:59.199 person as a creator. We have to say no people 0:13:59.280 --> 0:14:03.560 are creators, because otherwise we can't be technologists anymore. The 0:14:03.640 --> 0:14:05.400 moment you give up people, you might as well go 0:14:05.440 --> 0:14:07.800 smash your computer because it doesn't make any sense anymore, 0:14:07.840 --> 0:14:10.680 because it's only possible. Definition isn't serving people. 0:14:24.760 --> 0:14:27.800 So let me dig into something about how it's this 0:14:27.880 --> 0:14:31.400 figure ground reversal of looking at AI as a collaboration 0:14:31.440 --> 0:14:33.280 of people. So this taps into something that I've been 0:14:33.280 --> 0:14:39.880 writing lately about this question of whether AI has theory 0:14:39.920 --> 0:14:41.600 of mind. And you may know some people have published 0:14:41.640 --> 0:14:44.560 paper saying, hey, with the theory of mind has emerged, 0:14:44.880 --> 0:14:47.000 And just as a reminder of the audience, it's you know, 0:14:47.040 --> 0:14:49.440 theory of mind is being able to put yourself into 0:14:49.480 --> 0:14:52.480 the shoes of someone else, into the perspective, and understand 0:14:52.520 --> 0:14:55.680 their beliefs, even if they're different from your own, and 0:14:55.800 --> 0:14:59.120 understand what they believe is true or not true their 0:14:59.200 --> 0:15:02.920 perspective so on. Now, this is something humans do seamlessly 0:15:02.960 --> 0:15:06.160 and effortlessly all the time. We're very good at understanding, Oh, 0:15:06.360 --> 0:15:08.680 he doesn't know that piece of information, she does know that, 0:15:09.480 --> 0:15:13.360 and so on. But the question is do llms do it? 0:15:13.360 --> 0:15:16.440 So some people have claimed yes. I've studied this point 0:15:16.640 --> 0:15:20.680 very carefully. I conclude that they do not. Now why 0:15:20.680 --> 0:15:23.760 do some people write that they do. It's because you 0:15:23.880 --> 0:15:28.760 give these questions that are probing on. Hey, what if there's, 0:15:28.920 --> 0:15:31.240 you know, a bag that's labeled with one thing, but 0:15:31.280 --> 0:15:34.440 there's something different inside. What does the person believe who's 0:15:34.480 --> 0:15:36.200 just looking at the label, and what does she believe 0:15:36.240 --> 0:15:38.160 after she looks inside the bag? And so on and 0:15:38.360 --> 0:15:42.720 lllls will get the stuff right, and so they say, Wow, they've. 0:15:41.920 --> 0:15:42.640 Got theories in mind. 0:15:42.640 --> 0:15:45.920 They can understand what it's like to be this person 0:15:46.080 --> 0:15:48.760 in these different circumstances. The reason I don't think that's 0:15:48.800 --> 0:15:54.040 true is what I'm calling this the intelligence echo illusion. 0:15:54.120 --> 0:15:58.040 And what I mean by that is thousands, maybe millions 0:15:58.080 --> 0:16:00.880 of people have written about this on line. They've written 0:16:00.880 --> 0:16:05.520 out these deary mind questions, these unexpected things and so on. 0:16:05.800 --> 0:16:09.520 So an LM absorbs the statistics of lots and lots 0:16:09.560 --> 0:16:11.360 of these things, and then when you ask the question, 0:16:11.520 --> 0:16:13.120 you say, my god, it gave me the right answer. 0:16:13.120 --> 0:16:14.800 And of course gave the right answer. It's read the 0:16:14.880 --> 0:16:17.120 damn problem lots of times. 0:16:17.480 --> 0:16:19.600 What's interesting, this is the calculation. 0:16:19.200 --> 0:16:22.520 I ran recently, is you know, if you look at 0:16:22.560 --> 0:16:27.040 the common crawl, this corpus of the data that all 0:16:27.080 --> 0:16:32.800 these large language models crawl and absorb that's, according to 0:16:32.800 --> 0:16:36.760 my calculation, one thousand times larger than what you could 0:16:36.800 --> 0:16:41.840 read in a lifetime. And so the fact is, many, 0:16:41.920 --> 0:16:45.680 many times we'll pose a question to let's say, chat YOUPT, 0:16:45.920 --> 0:16:48.000 it'll give some answer. I'll think, my god, it's got 0:16:48.040 --> 0:16:50.760 theory of mind. It's really done something strange. But in fact, 0:16:50.760 --> 0:16:53.240 you're just hearing this intelligence echo, by which I mean 0:16:54.000 --> 0:16:55.120 people already knew this. 0:16:55.360 --> 0:16:57.000 You just didn't know that, or you didn't know that 0:16:57.040 --> 0:16:57.800 other people do that. 0:16:59.240 --> 0:17:04.119 Sure, but I mean we're right back in this territory 0:17:04.160 --> 0:17:07.760 of mysticism and ontology. Because if somebody wished to disagree 0:17:07.840 --> 0:17:09.320 with you, and that person would not be me. 0:17:09.840 --> 0:17:10.880 But if somebody wished to. 0:17:10.800 --> 0:17:13.920 Disagree with you, they'd say that all those people were 0:17:14.000 --> 0:17:17.000 just reflecting things like that anyway, So there's no difference 0:17:17.080 --> 0:17:19.719 that the AI and the people are trained on are 0:17:19.920 --> 0:17:23.800 in the same status, the same category. And so what 0:17:23.960 --> 0:17:26.040 if the AI got to it this way with people 0:17:26.119 --> 0:17:27.840 in the way, it's still the same thing that people 0:17:27.920 --> 0:17:31.280 are doing. And why are you making this distinction? Why 0:17:31.280 --> 0:17:33.160 are you trying to be all mystical in love people? 0:17:33.200 --> 0:17:35.919 What's wrong with you? And so I'm just going to 0:17:35.960 --> 0:17:39.120 declare you should be mystical and elevate people because it's 0:17:39.160 --> 0:17:43.280 the only thing. Weirdly, this little bit of faith, which 0:17:43.400 --> 0:17:46.200 I seem irrational, is the only way to save rationality 0:17:46.280 --> 0:17:51.240 because otherwise, once again we have no beneficiary for technology. 0:17:51.720 --> 0:17:55.520 You can have science without elevating people, but you can't 0:17:55.560 --> 0:18:00.479 have technology. Technology has to have a beneficiary. The problem 0:18:00.520 --> 0:18:05.360 to be solved is serving people, and technology isn't sensible 0:18:05.359 --> 0:18:08.440 without a problem to be solved. Science is you can 0:18:08.440 --> 0:18:10.520 have a science theory without knowing what it's for. You 0:18:10.560 --> 0:18:14.480 can't have a technology without knowing what it's for. Or 0:18:14.520 --> 0:18:16.480 maybe I mean one could argue that you can have 0:18:16.520 --> 0:18:19.680 some kind of underlying, very fundamental technology that might be 0:18:19.680 --> 0:18:22.440 applied in different ways, but ultimately to actually make any 0:18:22.480 --> 0:18:24.960 instance of it, it has to be for something. There 0:18:25.040 --> 0:18:26.440 has to be a human at the end of the 0:18:26.520 --> 0:18:29.639 chain for it to be sensible. So I totally agree 0:18:29.640 --> 0:18:31.480 with you, but you have to recognize that there's a 0:18:31.520 --> 0:18:34.159 figure ground reversal there, And somebody who disagrees with you 0:18:34.280 --> 0:18:36.160 might just say, well, the AI and the people are 0:18:36.200 --> 0:18:38.800 the same and they're both just getting this information. But 0:18:39.160 --> 0:18:43.679 I still think our only choice to remain rational technologists 0:18:43.840 --> 0:18:45.200 is to accept that we have to be kind of 0:18:45.240 --> 0:18:50.320 mystical humanists to frame it, and that's not comfortable for 0:18:50.320 --> 0:18:52.200 a lot of people. But I think if you really 0:18:52.240 --> 0:18:54.880 examine the logic of the situation, you'll come to agree 0:18:54.960 --> 0:18:55.159 with me. 0:18:55.600 --> 0:18:57.920 Oh, I already agree with you, actually, And that's why, 0:18:58.119 --> 0:19:01.000 that's why I really care about this intelligence echo illusion, 0:19:01.040 --> 0:19:05.120 which is to say, you're hearing the echoes of people 0:19:05.160 --> 0:19:08.040 who've said this thing before, but you mistake it for 0:19:08.119 --> 0:19:11.280 the voice of an AI that has theory of mind. 0:19:11.400 --> 0:19:15.160 Yeah, and so why can't you Let's say the AI 0:19:15.320 --> 0:19:18.200 gets the puzzle with the bag, right, Okay, let's say 0:19:18.280 --> 0:19:23.639 those now in my preferred future, where you then get 0:19:24.240 --> 0:19:27.159 a list of the top twelve or twenty five or 0:19:27.200 --> 0:19:30.400 whatever it is people who contributed. What it might say 0:19:30.520 --> 0:19:34.040 is I would tell you, but actually there are a 0:19:34.119 --> 0:19:36.800 million people interchangeably in the first slot here because this 0:19:36.840 --> 0:19:40.840 thing is so common, and that has to be an 0:19:40.880 --> 0:19:43.920 acceptable and honest answer. So like if you say, hey, 0:19:43.960 --> 0:19:47.000 could you do something like I want a cat that's 0:19:47.400 --> 0:19:50.520 playing the banjo or whatever, and they'll say, got to 0:19:50.560 --> 0:19:53.680 tell you a lot of cats out there. It's the Internet, remember, 0:19:54.040 --> 0:19:57.520 lots of cats, So it really could have been any cat. 0:19:57.600 --> 0:19:59.920 In this case, it was this cat. But that's not special. 0:20:00.240 --> 0:20:02.880 And so we have to have a fungibility measure, not 0:20:02.960 --> 0:20:06.399 just an influence measure. And that that's so, this whole 0:20:06.800 --> 0:20:09.919 future world of acknowledging people is a bit more subtle 0:20:09.960 --> 0:20:11.160 and complicated than. 0:20:11.280 --> 0:20:12.119 I'd like it to be. 0:20:12.280 --> 0:20:15.080 But I've never seen anything about it that doesn't make 0:20:15.119 --> 0:20:16.400 sense or is unachievable. 0:20:17.400 --> 0:20:20.199 Right, So one idea that you've mentioned me in the 0:20:20.240 --> 0:20:22.719 past would be, I can't remember if I got this 0:20:22.800 --> 0:20:24.720 term digital dividends right, if I made that up some 0:20:24.760 --> 0:20:26.960 whe along the way. But the idea is that a 0:20:27.040 --> 0:20:29.760 creator gets paid, He gets a few nickels here and 0:20:29.800 --> 0:20:34.080 there because on you know, Dolly, it's used his painting 0:20:34.600 --> 0:20:35.960 as part of the influence. 0:20:36.320 --> 0:20:37.280 Yeah. 0:20:37.560 --> 0:20:41.200 So the thing about this idea, which currently the term 0:20:41.240 --> 0:20:44.960 I like is is dated dignity. Many will object and 0:20:45.040 --> 0:20:48.600 correctly that in a lot of cases, if somebody's output 0:20:48.680 --> 0:20:51.160 happens to include a picture of your shoe for some 0:20:51.240 --> 0:20:54.760 image that has a shoe or something, or maybe they 0:20:54.840 --> 0:20:57.239 used your drone footage of waves for the wave thing 0:20:57.359 --> 0:21:00.880 or something, what you'll get literally would be very small. 0:21:00.920 --> 0:21:03.320 It might be quite a bit less than Nichols, right, 0:21:03.720 --> 0:21:04.720 But that's not the point. 0:21:04.760 --> 0:21:08.640 See, the question is whether we think of the. 0:21:08.640 --> 0:21:11.720 Future as a linear extension of what we already know 0:21:11.960 --> 0:21:17.480 or something really expansive and fantastic and unpredictable. And a 0:21:17.520 --> 0:21:20.240 lot of the people who tell me that they believe 0:21:20.240 --> 0:21:22.760 in this radical future where there's a singularity, where this 0:21:22.840 --> 0:21:25.280 AI transformers reality and a blink of an eye and 0:21:25.400 --> 0:21:28.440 humans are obsolete, and they're actually thinking in a very 0:21:28.480 --> 0:21:31.440 linear way because what they believe is the AI as 0:21:31.480 --> 0:21:35.159 that exists is capable of creating the AI of the future, 0:21:35.160 --> 0:21:37.639 and that we already know in a sense everything that 0:21:37.680 --> 0:21:39.719 needs to be known, and all we have to do 0:21:39.760 --> 0:21:42.080 is turn the on switch. In future generations won't be 0:21:42.160 --> 0:21:45.719 more creative than us. Where the final creative, you know? 0:21:46.240 --> 0:21:49.000 And I think that might not be the future. 0:21:49.000 --> 0:21:49.320 I want. 0:21:49.359 --> 0:21:53.760 What if the future is one in which there's incredible, creative, 0:21:53.800 --> 0:21:59.399 productive things happening, and I'm not creative enough perhaps to 0:21:59.400 --> 0:22:00.600 come up with the examples. 0:22:00.640 --> 0:22:02.360 I'll just use the one I used last. 0:22:02.119 --> 0:22:05.920 Week, which is, in the future, we have some way 0:22:05.920 --> 0:22:10.080 of extending our bodies physiologically, so we can fly around 0:22:10.080 --> 0:22:12.800 an open space in the vacuum and do all sorts 0:22:12.840 --> 0:22:16.720 of things and propel ourselves. And what if in order 0:22:16.760 --> 0:22:19.280 to do those body extensions there's some sort of bioprinter 0:22:19.400 --> 0:22:22.240 we get into. And what if they're aisystems that run 0:22:22.280 --> 0:22:24.680 that because it's the only way. And what if there's 0:22:24.720 --> 0:22:27.480 a whole new creative class of people who've contributed data 0:22:27.480 --> 0:22:29.920 to that thing. All right, so that's the whole new thing, 0:22:30.240 --> 0:22:32.320 and I think the best of those people might get 0:22:32.400 --> 0:22:35.560 rich from contributing data that then is able to be 0:22:35.560 --> 0:22:38.320 beneficial to people through a large model that their data 0:22:38.359 --> 0:22:42.120 helped train. And what if there's more and more examples 0:22:42.240 --> 0:22:44.400 like that. What if there's thousands and millions of tens 0:22:44.440 --> 0:22:47.920