WEBVTT - Will AI Radically Change the World by 2027? 0:00:15.564 --> 0:00:16.044 Pushkin. 0:00:20.964 --> 0:00:24.124 Hey everybody, Nate here, jumping in before the short so 0:00:24.164 --> 0:00:25.884 that we've been having a lot of fun answering your 0:00:25.884 --> 0:00:29.484 listener questions. So far, we've covered things like Pascal's Wager, 0:00:29.684 --> 0:00:32.164 the hot hand fallacy, or the fallacy of the hot 0:00:32.204 --> 0:00:36.044 hand fallacy actually, and the expected value of learning new languages. 0:00:36.684 --> 0:00:38.444 We want to keep doing this kind of thing, so 0:00:38.484 --> 0:00:43.004 send us all your questions about risk decision making, game theory, poker, 0:00:43.444 --> 0:00:45.444 you name it. Reach out to us on social media 0:00:45.564 --> 0:00:49.804 or email at Risky Business at pushkin dot fm. Even 0:00:49.804 --> 0:00:51.804 if you're not a premium subscriber, this is a great 0:00:51.804 --> 0:00:54.004 way to support the show, so you can keep sharing 0:00:54.044 --> 0:00:56.444 it free of charge. We look forward to hearing from you. 0:01:03.324 --> 0:01:06.724 Welcome back to Risky Business, a show about making better decisions. 0:01:06.844 --> 0:01:08.524 I'm Maria Kanakova. 0:01:08.084 --> 0:01:08.924 And I'm Nate Silver. 0:01:09.604 --> 0:01:11.204 Today on the show is going to be a little 0:01:11.204 --> 0:01:12.204 bit doomtastic. 0:01:12.924 --> 0:01:15.404 Yeah, I mean, I don't know if it's like worse 0:01:15.524 --> 0:01:19.564 than thinking about like the global economy going into a 0:01:19.604 --> 0:01:23.804 recession because of dumbfuck tariff policies. This is all about 0:01:23.804 --> 0:01:25.844 how we're all going to die in seven years instead. No, 0:01:25.884 --> 0:01:29.084 I'm just kidding. This is a very very intelligent and 0:01:29.124 --> 0:01:32.644 well written and thoughtful report called AI twenty twenty seven 0:01:32.644 --> 0:01:34.924 that we're going to spend the whole show on because 0:01:34.924 --> 0:01:37.604 think it's such an interesting to talk about, but that 0:01:37.684 --> 0:01:42.044 you know, includes some dystopian possibilities. 0:01:42.044 --> 0:01:45.324 I would say it does indeed, so let's get into 0:01:45.364 --> 0:01:47.764 it and hope that you guys are all still here 0:01:47.804 --> 0:01:52.684 to listen to us in seven years. 0:01:52.884 --> 0:01:57.884 The contrast is interesting between like all the chaos we're 0:01:57.924 --> 0:02:01.684 seeing with tariff policy in terms of starting a trade 0:02:01.724 --> 0:02:05.684 war with China and then other types of chaos. It's 0:02:05.724 --> 0:02:07.844 interesting to kind of look at this. I mean, I 0:02:07.844 --> 0:02:12.244 wouldn't call it a more optimistic future exactly, but like 0:02:12.684 --> 0:02:14.644 but like on a different trajectory of like a future 0:02:14.684 --> 0:02:18.844 that's going to change very fast, according to these authors, 0:02:18.844 --> 0:02:23.964 with profound implications for you know, everything, the human species. 0:02:25.084 --> 0:02:28.324 These researchers and authors are saying that everything is going 0:02:28.364 --> 0:02:30.404 to change profoundly. 0:02:30.164 --> 0:02:30.804 And. 0:02:32.524 --> 0:02:36.524 Even though there is some hedging hair, this is kind 0:02:36.564 --> 0:02:40.484 of their base case scenario, and like you know, base 0:02:40.524 --> 0:02:43.244 case number one and base case number two differ. There's 0:02:43.244 --> 0:02:44.844 like kind of a choose your own adventure at some 0:02:44.924 --> 0:02:47.764 point in this report. But they're both very different than 0:02:47.764 --> 0:02:50.604 the status quo, right, and the notionately hear from that, 0:02:50.764 --> 0:02:56.564 like everything becomes different. If ais become substantially more intelligent 0:02:56.604 --> 0:02:58.564 than human beings, people can debate and we will debate 0:02:58.604 --> 0:03:00.484 on this program what that means. But yeah, do you 0:03:00.564 --> 0:03:02.564 want to do you want to contextualize this more? Do 0:03:02.564 --> 0:03:04.444 you want to tell people to actors are of this report? 0:03:04.764 --> 0:03:10.604 Absolutely? Absolutely? So the report is authored officially by five people, 0:03:10.764 --> 0:03:14.364 and I think unofficially there's also a sixth We've got 0:03:14.844 --> 0:03:19.924 Eli Leifelund, who's a super forecaster and he was ranked 0:03:20.004 --> 0:03:24.324 first on RAN's forecasting initiative, so he is someone who 0:03:24.404 --> 0:03:27.484 is very good at kind of looking at the future 0:03:27.524 --> 0:03:30.524 trying to predict what's going to happen. You have Jonas 0:03:30.564 --> 0:03:35.684 Palmer who's a VC at Macroscopic Ventures. Thomas Larson was 0:03:35.724 --> 0:03:40.004 a former executive director of the Center for AI Policy, 0:03:40.764 --> 0:03:44.644 so that is a center that advises both sides of 0:03:44.684 --> 0:03:47.404 the aisle on you know, how how AI is going 0:03:47.484 --> 0:03:51.884 to go? And Romeo Dean, who is part of Harvard's 0:03:51.884 --> 0:03:56.044 AI Safety Student Team, so someone who is still a student, 0:03:56.124 --> 0:03:59.884 still learning, but kind of the next generation of people 0:04:00.564 --> 0:04:04.884 looking at AI. And finally, we have Daniel coch Taylor, 0:04:05.484 --> 0:04:10.244 who basically had written a report back in twenty twenty one. 0:04:10.284 --> 0:04:14.484 He was a AI researcher and he looked at predictions 0:04:14.524 --> 0:04:17.644 for AI for twenty twenty six, and it turns out 0:04:17.764 --> 0:04:20.924 that his predictions were pretty spot on, and so open 0:04:20.924 --> 0:04:24.844 AI actually hired Daniel as a result of this report. 0:04:24.844 --> 0:04:26.804 Obviously he now left, yes. 0:04:26.684 --> 0:04:29.004 And then and then he left exactly. 0:04:28.644 --> 0:04:32.324 And importantly, so there's also Scott Alexander exactly. 0:04:32.404 --> 0:04:34.284 So I was saying, he's the he's the person who 0:04:34.764 --> 0:04:37.644 kind of is in the background, and he's you guys 0:04:37.764 --> 0:04:40.964 might know him as the author behind astral Codux. 0:04:41.804 --> 0:04:43.404 And I know Scott. He's one of the kind of 0:04:43.444 --> 0:04:50.604 fathers of what you might call rationalism. I think Scott, 0:04:50.644 --> 0:04:52.884 when I interview him from a book, was happy enough 0:04:52.884 --> 0:04:55.284 with that term and accused me or co opted me 0:04:55.324 --> 0:05:00.284 into also being a rationalist. These people are somewhat adjacent 0:05:00.324 --> 0:05:02.884 to the effective altruist, but not quite right. They're just 0:05:02.884 --> 0:05:07.964 trying to apply a sort of thoughtful, rigorous, quantitative lens 0:05:08.164 --> 0:05:12.244 to big picture problem, including existential risk, of which most 0:05:12.284 --> 0:05:15.124 people in this community believe that AI is both an 0:05:15.124 --> 0:05:18.964 existential risk and also kind of an existential opportunity, right, 0:05:18.964 --> 0:05:21.004 that it could transform things. You talk to Sam Alt 0:05:21.044 --> 0:05:23.564 and he'll say We're going to cure cancer and illuminate 0:05:23.604 --> 0:05:28.084 poverty and whatever else. Right, And Scott's also an excellent writer. 0:05:28.324 --> 0:05:31.524 And so let me disclose something which is slightly important here. 0:05:31.564 --> 0:05:34.524 So I actually was approached by some of the authors 0:05:34.564 --> 0:05:37.324 of this report a couple of months ago, I guess 0:05:37.364 --> 0:05:40.644 it was in February ish, just to give feedback and 0:05:40.764 --> 0:05:44.204 chat with them. So I'm working off the draft version, right, 0:05:44.244 --> 0:05:46.724 which I do not believe they changed very much. So 0:05:46.764 --> 0:05:48.964 my notes pertained to an earlier draft. I did not 0:05:49.044 --> 0:05:50.204 have time this morning to go back and re. 0:05:50.244 --> 0:05:54.084 Read it, so I was not on the inside loops. 0:05:54.084 --> 0:05:56.764 So I did not get an earlier draft. And I've 0:05:56.804 --> 0:06:00.644 read this draft and basically just to kind of big 0:06:00.684 --> 0:06:06.204 picture sum it up, it outlines two scenarios, right, two 0:06:06.284 --> 0:06:10.644 major scenarios for how AI might change the world as 0:06:10.924 --> 0:06:16.284 soon as twenty thirty. Now important note like that that 0:06:16.564 --> 0:06:19.404 date is kind of hedged. It might be sooner, it 0:06:19.484 --> 0:06:22.844 might be later. There's kind of a there's a confidence 0:06:22.884 --> 0:06:27.924 interval there. But the two different scenarios. One basically we 0:06:28.124 --> 0:06:32.524 do twenty thirty, humanity disappears and is taken over by AI. 0:06:33.084 --> 0:06:37.244 The positive report is in twenty thirty. Basically we get 0:06:38.004 --> 0:06:41.084 AIS that are aligned to our interests, and we get 0:06:41.164 --> 0:06:44.764 kind of this AI utopia where AI is actually help 0:06:44.884 --> 0:06:47.764 make life much better for everyone and make the standard 0:06:47.844 --> 0:06:50.484 of living much higher. But the crucial turning point is 0:06:50.524 --> 0:06:53.844 before twenty thirty, and the crucial kind of question at 0:06:53.844 --> 0:06:58.524 the center of this is will we be able to 0:06:58.644 --> 0:07:04.244 design AIS that are truly aligned to human interests rather 0:07:04.284 --> 0:07:07.164 than just appear to be aligned and kind of lying 0:07:07.204 --> 0:07:11.684 to us while actually following their own agenda. And how 0:07:11.724 --> 0:07:14.924 we handle that is kind of the lynchpin. And it's 0:07:14.924 --> 0:07:17.684 actually interesting, Nate, that you started out with China because 0:07:17.964 --> 0:07:20.204 a lot of the policy choices and a lot of 0:07:20.204 --> 0:07:23.524 what they see as kind of the decision points that 0:07:23.564 --> 0:07:27.924 will affect the future of humanity actually hinge on the 0:07:28.124 --> 0:07:32.364 US China dynamic, how they compete with each other, and 0:07:32.404 --> 0:07:37.644 how that sometimes might basically clash against safety concerns because 0:07:38.324 --> 0:07:41.004 no one wants to be left behind. Can we manage 0:07:41.044 --> 0:07:44.964 that effectively and can kind of that transition work in 0:07:45.004 --> 0:07:47.524 our favor as opposed to against us. I think that 0:07:47.564 --> 0:07:49.204 this is kind of one of the big questions here, 0:07:49.364 --> 0:07:51.044 and so it's funny that we're seeing all of this 0:07:51.164 --> 0:07:54.564 trade war right now as the sideport is coming out. 0:07:55.404 --> 0:08:00.804 Yeah, look, I think this exercise is partly just a 0:08:00.884 --> 0:08:04.044 forecasting exercise, right, I mean, obviously it's just kind of 0:08:04.084 --> 0:08:08.204 like fork at the bottom where we learn to have 0:08:08.284 --> 0:08:11.404 an AI slow down or we kind of are pressing 0:08:11.404 --> 0:08:14.044 fully on the accelerator, right, Like, in some ways, scenarios 0:08:14.084 --> 0:08:20.084 are like not that different, right, Either one assumes remarkable 0:08:20.164 --> 0:08:23.604 rates of technological growth that I think even AI I'm 0:08:23.604 --> 0:08:26.364 never quite sure who to call an optimist or a pessimist, right, 0:08:26.404 --> 0:08:30.324 even AI believers you know, might think is a little 0:08:30.324 --> 0:08:32.164 bit aggressive. Right. But what they want to do is 0:08:32.204 --> 0:08:36.884 they want to have like a specific, fleshed out scenario 0:08:36.964 --> 0:08:38.804 for how the world would look. Like it's kind of 0:08:38.884 --> 0:08:42.644 like a modal scenario. And like I think they'd say that, like, 0:08:44.164 --> 0:08:46.564 we're not totally sure about either of these necessarily, right, 0:08:46.564 --> 0:08:48.884 And I don't think they'd be as like pedantic as 0:08:48.884 --> 0:08:50.964 to say, if you do X, Y, and Z, then 0:08:51.844 --> 0:08:54.004 we'll save the world and have utopia, and if you don't, 0:08:54.084 --> 0:08:56.884 then we'll all die. Right. I think they probably say 0:08:56.964 --> 0:08:59.884 it's unclear and there's kind of like risk either way. 0:09:00.004 --> 0:09:02.204 We wanted to go through the scenario like fleshing out 0:09:02.324 --> 0:09:05.604 like what the world might look like. Right. I do 0:09:05.644 --> 0:09:08.644 think one thing that's important is that whatever decisions are 0:09:08.724 --> 0:09:13.044 made now could get locked in, right that you pass 0:09:13.084 --> 0:09:18.244 certain points in overturn and it becomes very hard to accelerate, 0:09:18.364 --> 0:09:20.804 like an arms race. This is you know, what we 0:09:20.884 --> 0:09:23.924 found during the Cold War for example. I mean, one 0:09:23.924 --> 0:09:25.684 of the big things I look at is like, do 0:09:25.764 --> 0:09:29.884 we force the AI to be transparent in its thinking 0:09:30.444 --> 0:09:32.844 with humans? Right? Like, now there's been a movement toward 0:09:33.444 --> 0:09:36.484 the AI will actually explicate it's thinking more. I'll ask 0:09:36.524 --> 0:09:39.924 you a query open AI. The Chinese models to this too, right, 0:09:39.964 --> 0:09:42.204 and I'll say, I am thinking about X, Y and Z, 0:09:42.324 --> 0:09:44.484 and I'm looking up PD and Q and now I'm 0:09:44.484 --> 0:09:47.644 reconsidering this. It actually has this chain of thought process, right, 0:09:48.044 --> 0:09:51.284 which is explicated in English. You know. One concern is 0:09:51.284 --> 0:09:53.604 that what if the AI just kind of communicates to 0:09:53.644 --> 0:09:59.164 another in these implicit vectors that's inferring from all the 0:09:59.164 --> 0:10:02.644 texts it has. It's kind of unintelligible to human beings, right, 0:10:02.644 --> 0:10:04.604 and maybe kind of quote unquote thinking in that way 0:10:05.204 --> 0:10:07.044 in the first place, and then does us the favor 0:10:07.124 --> 0:10:10.884 of like translating back to it goes from to kind 0:10:10.884 --> 0:10:13.284 of this big bag of numbers is one day I 0:10:13.284 --> 0:10:17.044 researcher called it, right, and then it translates back into 0:10:17.084 --> 0:10:19.764 English or whatever language you want. Really, in the end, 0:10:20.844 --> 0:10:23.524 what if it just skip cuts out that last step right, 0:10:23.604 --> 0:10:26.644 then we can't kind of like check what AI is doing. 0:10:26.724 --> 0:10:30.604 Then it can behave deceptively more easily. So you know, 0:10:30.804 --> 0:10:34.004 so that part seems to be important. I want to 0:10:34.004 --> 0:10:37.084 hear your your first impressions before I kind of poison 0:10:37.124 --> 0:10:38.164 the well too much. 0:10:39.284 --> 0:10:43.204 Well, my first impressions is that the alignment problem is 0:10:43.204 --> 0:10:46.804 a very real one and an incredibly important one to solve. 0:10:47.124 --> 0:10:49.684 And what I got from this is that actually, the 0:10:49.724 --> 0:10:53.884 problem that I've had with like these initial AI lms 0:10:54.364 --> 0:10:56.924 is the kernel of what they're seeing there. Right. So 0:10:57.124 --> 0:10:58.844 you and I have talked about this on the show 0:10:58.924 --> 0:11:01.964 in the past, and I've said, well, my problem is 0:11:01.964 --> 0:11:04.884 that when I'm a domain expert, right, I start seeing 0:11:05.004 --> 0:11:08.684 some inaccuracies, and I start seeing like places where like 0:11:08.724 --> 0:11:12.284 it either just didn't do well or made shit up 0:11:12.364 --> 0:11:15.044 or or whatever it is. Now I think it's very 0:11:15.084 --> 0:11:17.444 clear that those problems are going to go away, right, 0:11:17.524 --> 0:11:20.804 that that is going to get much much better. However, 0:11:21.364 --> 0:11:24.804 the kernel of it's showing me something, but that might 0:11:24.884 --> 0:11:27.404 just be you know, I have no way of verifying 0:11:27.484 --> 0:11:30.364 if that's what's going on, what it's reading, like, how 0:11:30.404 --> 0:11:32.764 it's I don't want to say thinking about it, even 0:11:32.804 --> 0:11:34.644 though in the report they do use thinking, but. 0:11:34.644 --> 0:11:38.404 It's normalizing too much, I think is correct. 0:11:38.404 --> 0:11:39.004 I think it's. 0:11:39.164 --> 0:11:42.724 Okay, we'll stick to that language. Yeah, okay, So, so 0:11:42.844 --> 0:11:47.644 how it's thinking about it that those little problems and 0:11:47.724 --> 0:11:51.244 like the little the glitches and the things that it 0:11:51.324 --> 0:11:54.364 might be doing where it starts actually glitching on purpose, 0:11:54.804 --> 0:11:57.924 are not going to be visible to the human eye. 0:11:58.124 --> 0:11:59.884 And so one of the main things that they say 0:11:59.924 --> 0:12:04.404 here is that as AI internal R and D gets 0:12:04.684 --> 0:12:09.924 rapidly faster, so that means basically AI's researching AI, right, 0:12:10.204 --> 0:12:13.684 and so internally they start developing new models, and as 0:12:13.724 --> 0:12:17.924 they kind of surpass human ability to monitor it, it 0:12:17.964 --> 0:12:21.804 becomes progressively more difficult to figure out. Okay, is the 0:12:21.844 --> 0:12:24.564 AI actually doing what I want it to do? Is 0:12:24.644 --> 0:12:28.964 the output that it's giving me its actual thought process? 0:12:29.284 --> 0:12:32.084 And is it accurate or is it like trying to 0:12:32.124 --> 0:12:34.804 deceive me? But it's actually kind of inserting certain things 0:12:35.164 --> 0:12:37.844 on purpose because it has different goals, right, because it 0:12:37.884 --> 0:12:41.244 is actually secretly misaligned, but it's very good at persuading 0:12:41.284 --> 0:12:43.484 me that it's aligned. Because one of the things that 0:12:43.844 --> 0:12:45.964 actually came out of this report, and I was like, huh, 0:12:46.044 --> 0:12:48.684 you know this is interesting is if we get this 0:12:50.644 --> 0:12:55.844 remarkable improvement in AI, it will also remarkably improve at 0:12:55.884 --> 0:13:00.684 persuading us right as part of it. Making Yeah, so 0:13:00.724 --> 0:13:03.044 this is but I've never even thought about that. I 0:13:03.084 --> 0:13:05.884 was like, okay, fine, But one of the things that 0:13:05.924 --> 0:13:08.524 I do buy is that it's going to be very 0:13:08.524 --> 0:13:11.044 difficult for us to monitor it and to figure out 0:13:11.084 --> 0:13:14.604 like is it truly aligned with human wants, with human desires, 0:13:14.644 --> 0:13:18.324 with human goals, And the experts who are capable of 0:13:18.364 --> 0:13:20.884 doing that, I think are actually going to dwindle right 0:13:21.004 --> 0:13:24.604 as AI starts proliferating in society, And so to me, 0:13:24.724 --> 0:13:27.004 that is something that is actually quite worrisome and that 0:13:27.124 --> 0:13:29.524 is something that we really need to be paying attention to. 0:13:29.724 --> 0:13:29.964 Now. 0:13:30.204 --> 0:13:33.444 Just to fast forward a little bit, in their doomsday 0:13:33.484 --> 0:13:38.284 scenario in twenty thirty one, AI takes over it basically 0:13:38.324 --> 0:13:42.204 like suddenly releases some chemical agents, right, and humanity ties 0:13:42.724 --> 0:13:44.804 and the rest of the stragglers are taken care of 0:13:44.844 --> 0:13:45.844 by drones, et cetera. 0:13:46.444 --> 0:13:49.044 I don't even like it. It doesn't even quick and 0:13:49.084 --> 0:13:52.764 painless death. I will say, let's hope we. 0:13:52.764 --> 0:13:55.044 Don't know what the chemical agents are. Might not be 0:13:55.124 --> 0:13:58.364 quick and painless. Some chemical agents are actually a very 0:13:58.404 --> 0:14:01.404 painful death thing. So let's hope. Let's hope it's quick 0:14:01.444 --> 0:14:06.484 and painless. Quickly, quick, yes, stuff quick, Okay, hopefully some 0:14:06.604 --> 0:14:11.164 chemical agents are not quick. I hope it's quick and painless. 0:14:12.004 --> 0:14:15.364 But if they're actually capable of deception at that high level, 0:14:15.684 --> 0:14:19.124 then you technically don't even need them to do it. 0:14:19.164 --> 0:14:22.004 If we're trusting medicine and all sorts of things to 0:14:22.084 --> 0:14:25.044 the AIS, it's pretty easy for it to actually manipulate 0:14:25.124 --> 0:14:30.764 something and actually insert something into codes et cetera that 0:14:30.804 --> 0:14:33.284 will fuck up humanity in a way that we can't 0:14:33.804 --> 0:14:37.004 that we can actually figure out at the moment. Right, 0:14:37.084 --> 0:14:38.364 Like the way I think of it, and this is 0:14:38.404 --> 0:14:40.644 not from the this is not from the paper, but 0:14:40.684 --> 0:14:43.524 this is just the way that my mind processed. It 0:14:43.564 --> 0:14:46.804 is like think about DNA, right, like you have these 0:14:47.004 --> 0:14:52.724 remarkably complex, huge strands of data, and as we've found out, 0:14:53.084 --> 0:14:57.884 but it's taken forever, one tiny mutation can actually be fatal, right, 0:14:57.924 --> 0:15:01.564 but you can't spot that mutation. Sometimes that mutation isn't 0:15:01.564 --> 0:15:04.524 fatal immediately, but we'll only manifest at a certain point 0:15:04.564 --> 0:15:06.764 in time. That's the way that my mind tried to 0:15:06.924 --> 0:15:11.204 kind of try to conceptualize what this actually means. And 0:15:11.284 --> 0:15:13.804 so I think that you know, that would be easy 0:15:13.844 --> 0:15:17.964 for a deceptive AI to do, and to me like that. 0:15:17.964 --> 0:15:20.444 That's kind of the big takeaway from this report is 0:15:20.484 --> 0:15:23.524 that we need to make sure that we are building 0:15:23.684 --> 0:15:29.564 AIS that will not deceive right that they're capabilities, they 0:15:29.924 --> 0:15:33.124 explain them in an honest way, and that honesty and 0:15:33.164 --> 0:15:36.564 trust is actually prioritized over other things, even though it 0:15:36.644 --> 0:15:39.284 might slow down research, it might slow down other things, 0:15:39.324 --> 0:15:43.644 but that that kind of alignment step is absolutely crucial 0:15:43.804 --> 0:15:47.324 at the beginning because otherwise humans are human, right, they're 0:15:47.364 --> 0:15:52.324 easily manipulated. And we often trust that computers are quote 0:15:52.404 --> 0:15:55.484 unquote rational because they're computers, but they're not. They have 0:15:55.564 --> 0:15:57.644 their own inputs, they have their own weights, they have 0:15:57.724 --> 0:16:01.884 their own values, and that could just lead us down 0:16:02.164 --> 0:16:02.844 a dark path. 0:16:02.924 --> 0:16:06.244 Yeah, so let me follow up with this pushback. I 0:16:06.244 --> 0:16:08.964 guess right, Like, first of all, I don't know that 0:16:09.044 --> 0:16:13.084 human so easily persuaded. This is my big critique with 0:16:13.204 --> 0:16:18.124 like all the misinformation people who say, well, misinformation is 0:16:18.164 --> 0:16:20.724 the biggest problem on society basis. It's like people are 0:16:20.724 --> 0:16:25.484 actually pretty stubborn and they're kind of sound pretentious. They're 0:16:25.564 --> 0:16:28.084 kind of basian and how they formulate their beliefs, right, 0:16:28.084 --> 0:16:30.204 they have some notion of reality. They're looking at the 0:16:30.204 --> 0:16:34.724 credibility of the person who is telling them these remarks. 0:16:35.084 --> 0:16:37.324 If it's an unpersuasive source, it might make them less 0:16:37.364 --> 0:16:39.884 likely to believe they're balancing with other information with their 0:16:39.924 --> 0:16:44.124 lived experience so called. Right. You know, part of the 0:16:44.124 --> 0:16:47.484 reason that like I am skeptical of AIS being super 0:16:47.484 --> 0:16:50.764 persuasive is like you know that it's an AI. You know, 0:16:50.804 --> 0:16:52.484 it's trying to persuade you, you know what I mean. So, like, 0:16:52.524 --> 0:16:54.764 if you go and play poker against like a really 0:16:54.844 --> 0:16:57.404 chatty player I Phil hemm With or Scott Seeber or 0:16:57.444 --> 0:16:59.924 someone like that, right, you know, on some level of 0:16:59.924 --> 0:17:04.004 the best play is just to totally ignore it. Right, 0:17:04.124 --> 0:17:06.124 You know that they are trying to sweet talk you 0:17:06.284 --> 0:17:09.804 into doing exactly what they want you to do. And 0:17:09.884 --> 0:17:13.564 so the best play is to disengage, or literally you 0:17:13.564 --> 0:17:15.284 can randomize your movis VI have some notion of what 0:17:15.324 --> 0:17:20.124 the game, theoretical optimal play might be, right, or salesmen 0:17:20.484 --> 0:17:23.724 or politicians have reputations for being Oh he's a little 0:17:23.724 --> 0:17:26.924 too smooth, Gavin Newsome a little too fucking smooth, right, 0:17:27.004 --> 0:17:29.244 I don't find Gavin news some persuasive at all, Right, 0:17:29.284 --> 0:17:31.044 use a little too from the hair gel to the 0:17:31.204 --> 0:17:34.564 constantly shifting vibes. I mean, I don't really find Gavin 0:17:34.604 --> 0:17:36.444 new Some persuasive at all, even though like an aim, 0:17:36.484 --> 0:17:39.164 I'd say, boy, Gavin new some good looking guys look 0:17:39.204 --> 0:17:41.924 gravelly throated, But you know whatever, I mean. Look the 0:17:41.924 --> 0:17:45.764 big critique I have of this project, and by the way, 0:17:45.804 --> 0:17:48.684 I think this is an amazing project in addition to 0:17:48.804 --> 0:17:51.884 like wonderful writing if you viewed on the web not 0:17:51.924 --> 0:17:54.364 your phone, so these very cool like little infographic that 0:17:54.484 --> 0:17:57.644 updates everything from like the market value. If they don't 0:17:57.644 --> 0:17:59.644 call it open AI, they call it open brain, I 0:17:59.644 --> 0:18:01.324 guess is what they settle on for a substitui. 0:18:01.404 --> 0:18:05.244 Yeah, they call everything something else just to make sure 0:18:05.284 --> 0:18:07.884 that they're not stepping on any toes. So they have 0:18:08.004 --> 0:18:11.844 open brain and and they have deep scent from China. 0:18:12.124 --> 0:18:16.404 I wonder which one that could be. I wonder. But 0:18:16.524 --> 0:18:20.564 it's beautifully presented and written, and like I appreciate they're 0:18:20.604 --> 0:18:23.764 going out on a limb here, you know, I mean, 0:18:23.804 --> 0:18:26.044 I think they have. It's been fairly well received. They've 0:18:26.044 --> 0:18:29.084 gotten some push back both from inside and I think 0:18:29.084 --> 0:18:32.444 outside the AI safety community. Right, but they're putting their 0:18:32.524 --> 0:18:35.284 necks in the line, hear. They will look if things 0:18:35.324 --> 0:18:37.084 look pretty normal, if or it looks pretty normal in 0:18:37.084 --> 0:18:39.804 twenty thirty two or whatever, right, then they will look 0:18:41.124 --> 0:18:43.484 dumb for having published this. 0:18:43.564 --> 0:18:47.044 Well, and they actually have that right as some scenario 0:18:47.204 --> 0:18:49.924 that you end up looking stupid if everything goes well. 0:18:50.164 --> 0:18:51.004 But that's okay. 0:18:51.364 --> 0:18:53.644 Now, can I push back on the persuasion thing a 0:18:53.644 --> 0:18:57.724 little bit, just on two things. So, first of all, 0:18:57.724 --> 0:19:02.004 the poker example is not actually a particularly applicable one here, 0:19:02.364 --> 0:19:05.404 because you know that you're playing poker and you know 0:19:05.484 --> 0:19:07.684 that someone is trying to get information and deceive you. 0:19:08.324 --> 0:19:10.804 The tricky thing. So this is kind of when I 0:19:10.844 --> 0:19:13.524 spend time with con artists. The best con artists aren't 0:19:13.564 --> 0:19:16.804 Gavin Newso like, they're not car salesmen. You have no 0:19:16.964 --> 0:19:19.524 idea they're trying to persuade you to do something. They 0:19:19.564 --> 0:19:23.244 are just like nice, affable people who are incredibly charismatic, 0:19:23.564 --> 0:19:25.764 and even in the poker community, by the way, like 0:19:25.804 --> 0:19:27.844 some of the biggest grifters who like it comes out 0:19:28.484 --> 0:19:31.084 later on, we're just like stealing money and doing all 0:19:31.164 --> 0:19:34.724 of these things. Are charming, right, They're not sleazy looking 0:19:35.004 --> 0:19:37.324 like they have no signs of, oh, I'm a salesman. 0:19:37.324 --> 0:19:39.564 I'm trying to sell you something. The people who are 0:19:39.604 --> 0:19:43.444 actually good at persuasion, you do not realize you are 0:19:43.444 --> 0:19:47.164 being persuaded. And I think people are incredibly easy to 0:19:47.244 --> 0:19:51.684 kind of to subtly lead in a certain direction if 0:19:51.724 --> 0:19:54.204 you know how to do it, and I think ais 0:19:54.324 --> 0:19:56.924 could could do that, and they might persuade you when 0:19:56.924 --> 0:19:58.804 you don't even think they're trying to persuade you. You might 0:19:58.844 --> 0:20:01.724 just ask like, can you please summarize this research report 0:20:02.084 --> 0:20:03.964 and the way that it frames it right, The way 0:20:03.964 --> 0:20:07.884 that it summarizes it just subtly changes the way that 0:20:07.964 --> 0:20:10.244 you think of this issue. We see that in psych 0:20:10.284 --> 0:20:12.404 studies all the time, by the way, where you have 0:20:12.484 --> 0:20:16.564 different articles presented in slightly different orders, slightly different ways, 0:20:16.604 --> 0:20:18.964 and people from the same political beliefs, you know, same 0:20:19.004 --> 0:20:24.084 starting point, come away with different impressions of what kind 0:20:24.124 --> 0:20:26.204 of the right course of action is or what this 0:20:26.324 --> 0:20:28.484 is actually trying to tell you. Because the way the 0:20:28.484 --> 0:20:32.604 information is presented actually influences how you think about it. 0:20:32.604 --> 0:20:36.524 It's very very easy to do subtle manipulations like that. 0:20:36.884 --> 0:20:39.564 And if we're relying on AI on a large scale 0:20:39.684 --> 0:20:42.044 for a lot of our lives, I think that if 0:20:42.044 --> 0:20:44.804 it has like a quote unquote master plan, you know, 0:20:44.844 --> 0:20:47.604 the way that they present in this report, then persuasion 0:20:47.644 --> 0:20:49.204 in that sense is actually going to be. 0:20:49.404 --> 0:20:52.164 You'll know you're being manipulated, right, that's the issue. 0:20:52.244 --> 0:20:53.444 No, you don't know. That's the thing. 0:20:53.604 --> 0:20:56.444 People will be manipulated because it's AI and that I 0:20:56.444 --> 0:20:58.244 don't but I don't know. 0:20:58.324 --> 0:21:02.164 I honestly, like Nate, I applaud your belief in human's 0:21:02.204 --> 0:21:05.284 ability to adjust to this, but I don't know that 0:21:05.364 --> 0:21:08.244 they will because I've just seen enough people who are 0:21:08.324 --> 0:21:14.524 incredibly intelligent fall for cons and then be very unpersuadable 0:21:14.644 --> 0:21:17.404 that they have been conned, right, instead doubling down and 0:21:17.404 --> 0:21:20.564 saying no, I have not so humans are stubborn, but 0:21:20.604 --> 0:21:23.044 they're also stubborn and saying I have not been deceived, 0:21:23.044 --> 0:21:25.324 I have not been manipulated, when in fact they have 0:21:25.844 --> 0:21:28.444 to protect their ego and to protect their view of 0:21:28.484 --> 0:21:31.484 themselves as people who are not capable of being manipulated 0:21:31.564 --> 0:21:34.524 or deceived. And I think that that is incredibly powerful. 0:21:34.844 --> 0:21:37.964 And I think that that's going to push against your optimism. 0:21:38.044 --> 0:21:40.204 I hope you're right, but from what I know, I 0:21:40.204 --> 0:21:40.884 don't think you are. 0:21:41.844 --> 0:21:43.804 I'm not quite sure to call it optimism some I 0:21:43.804 --> 0:21:45.884 guess maybe we do likely different views of human nature. 0:21:45.884 --> 0:21:48.284 But like there's not yet a substantial market for like 0:21:48.364 --> 0:21:53.244 AI driven art we're writing, and I'm sure there will 0:21:53.244 --> 0:21:58.604 be one eventually, right, But like people understand that context matters, right, 0:21:58.644 --> 0:22:00.124 that you could heavy I create a rip off the 0:22:00.164 --> 0:22:01.724 Mona Lisa, but you can also buy rip off the 0:22:01.724 --> 0:22:03.844 Mona Lisa and Canal Street for five bucks, right, and 0:22:03.884 --> 0:22:07.724 like it? You know, So it's it's the intentionality of 0:22:08.284 --> 0:22:11.924 the act and the context of the speaker. Now, sound 0:22:12.044 --> 0:22:16.284