WEBVTT - Cal Newport 0:00:03.160 --> 0:00:07.760 Media. Hello, and welcome to Better Offline. I'm, of course 0:00:07.840 --> 0:00:23.000 your host ed Zetron. As ever, support your neighborhood Zetron 0:00:23.040 --> 0:00:25.360 by subscribing to the premium newsletter discount link in the 0:00:25.400 --> 0:00:27.840 episode notes, of course, by a T shirt, download a 0:00:27.880 --> 0:00:30.240 blog where whatever it is you want to do, Okay, 0:00:30.240 --> 0:00:32.519 it's not up to me what you do. But today 0:00:32.800 --> 0:00:35.960 I'm joined by the incredible commside professor and commentator col 0:00:36.040 --> 0:00:37.600 newportcl Thank you for joining me. 0:00:37.880 --> 0:00:38.640 Oh, it's a pleasure. 0:00:38.760 --> 0:00:41.720 Ed. So I kind of wanted to start with I 0:00:41.760 --> 0:00:43.920 asked you for a quote a few like a week ago, 0:00:44.000 --> 0:00:46.440 maybe two weeks ago. I can't remember how time works anymore, 0:00:46.880 --> 0:00:51.160 but it was around the way the reporters cover AI 0:00:51.360 --> 0:00:53.760 and how it seems that a lot of the reporting 0:00:53.960 --> 0:00:56.360 is kind of directionally true rather than actually true. 0:00:57.880 --> 0:00:59.720 Yes, and I want to add something to it. Sense, 0:01:00.440 --> 0:01:02.920 So I've been thinking about that quote. Yeah, I've been 0:01:02.920 --> 0:01:04.880 thinking about it. So what I said, if I remember 0:01:04.920 --> 0:01:07.040 that quote properly, what I was saying is I was 0:01:07.120 --> 0:01:08.920 I was picking up a lot in the reporting on 0:01:09.040 --> 0:01:13.119 AI that you would lean into a story without having 0:01:13.160 --> 0:01:17.360 necessarily verified that the details are true, and that this 0:01:17.440 --> 0:01:19.600 is what's actually going on. Say, with a new AI model, 0:01:19.640 --> 0:01:21.960 you would lean into it anyways, because it was what 0:01:22.000 --> 0:01:25.480 I called directionally correct. It makes the general point that 0:01:25.520 --> 0:01:27.560 you see it as your job as reported to make, 0:01:27.600 --> 0:01:29.240 which is, hey, you need to be worried about this 0:01:29.360 --> 0:01:32.360 or this is a big deal, right, And so I 0:01:32.360 --> 0:01:34.800 think that is a problem. There's another issue I'm seeing, 0:01:34.840 --> 0:01:36.880 though I've sort of been refining my thinking on this, 0:01:37.800 --> 0:01:40.039 I'm also wondering if some of what I'm seeing in 0:01:40.080 --> 0:01:44.880 some of the reporting on this is just a embrace 0:01:45.000 --> 0:01:48.040 of the form of I'm going to give you a 0:01:48.160 --> 0:01:52.120 stress wave with no relief, just like we're all going 0:01:52.200 --> 0:01:55.240 to take turns. Just I will choose an era you 0:01:55.240 --> 0:01:58.840 haven't thought about. How about mathematics are going to go away. 0:01:58.840 --> 0:02:01.080 Mathematicians are going to be okay, I'll take that one. Yeah, 0:02:01.120 --> 0:02:05.920 let's stress like yeah, and they But there there's this 0:02:06.000 --> 0:02:08.480 weird sort of passivity to it where it's like I'm 0:02:08.520 --> 0:02:10.440 just gonna sort of it's it's uh, I call it 0:02:10.480 --> 0:02:14.160 like head shaking numerism. You're just like, it's this, this 0:02:14.240 --> 0:02:16.560 feels just going away. What can we do like just 0:02:16.639 --> 0:02:19.000 this sort of like passive head shaking. It's a very 0:02:19.000 --> 0:02:22.119 specific style you don't see a lot of other reporting historically. 0:02:22.160 --> 0:02:25.440 I think it takes on this resignation of I'm just 0:02:25.480 --> 0:02:30.000 gonna make the case that like you're screwed and then 0:02:30.400 --> 0:02:32.240 kind of give you a shoulder shrug and then we're 0:02:32.240 --> 0:02:34.200 and then then we're gonna drop the mic and walk off. 0:02:34.240 --> 0:02:36.000 And I'm kind of getting tired of this, Like I 0:02:36.000 --> 0:02:39.760 think there is a cost to stressing the hell out 0:02:39.760 --> 0:02:41.880 of people. I mean, I'm getting letters all the time 0:02:41.919 --> 0:02:43.799 now from people. They'll say things like I feel like 0:02:43.840 --> 0:02:45.839 I'm trapped in a cage, just being hit with wave 0:02:45.880 --> 0:02:48.360 after wave of stress, and there's no outlet. There's no 0:02:48.520 --> 0:02:51.160 door or possibility in making things better. And I think 0:02:51.160 --> 0:02:53.560 the CEOs are doing it, and I think increasingly we're 0:02:53.560 --> 0:02:57.000 seeing commentators doing it as well. This is not good 0:02:57.360 --> 0:02:59.359 in many different ways, So I don't know. I'm adding 0:02:59.360 --> 0:03:01.320 that to my list. Some of it's directional true reporting, 0:03:01.360 --> 0:03:03.800 like they really are worried that people aren't worried enough, 0:03:04.120 --> 0:03:06.280 and I think it's just sport. Now, can you find 0:03:06.280 --> 0:03:09.359 an area to come in and just write a headshaking 0:03:09.480 --> 0:03:12.480 article that's only trying to undermine the existence of this 0:03:12.560 --> 0:03:15.640 like important human activity or this job or our lives 0:03:15.720 --> 0:03:18.800 or whatever. It's a it's a very unusual style that 0:03:18.880 --> 0:03:20.079 quickly became a standard. 0:03:20.680 --> 0:03:22.920 And I see a lot with anything to do with 0:03:23.000 --> 0:03:26.880 AI and job studies, like I've been sent this Tuft's 0:03:26.960 --> 0:03:32.040 report where it's like, oh yeah, AI affected or a. 0:03:32.360 --> 0:03:36.280 They find these weird weasel words where it's like jobs 0:03:36.320 --> 0:03:39.920 that could be at risk from AI at some point 0:03:40.280 --> 0:03:42.120 and we put them in one bucket, and then jobs 0:03:42.120 --> 0:03:45.320 that might one day be we'll put that in another bucket. 0:03:45.400 --> 0:03:47.520 And there you go. Don't know what we're like you said, 0:03:47.600 --> 0:03:49.280 don't know what we're meant to do with this, don't 0:03:49.320 --> 0:03:50.960 know what anyone's meant to do with this information. But 0:03:51.000 --> 0:03:54.240 it's just like, well there you haven't they have. We're 0:03:54.280 --> 0:03:56.880 all fugged. It's it's the it's the end the job, 0:03:57.120 --> 0:03:59.520 even though the data does not say that. Like I've 0:03:59.520 --> 0:04:03.600 read I think every AI jobs report now, every single one, 0:04:04.080 --> 0:04:06.760 and they're all the same. They are all right, now, 0:04:07.440 --> 0:04:09.760 AI can do this, and then you look at what 0:04:09.800 --> 0:04:12.080 it says. It's like it can do law, well, it 0:04:12.200 --> 0:04:15.080 can't really do law. It can do one sigma within 0:04:15.240 --> 0:04:18.400 law kind of, and even then it isn't really obvious, 0:04:18.440 --> 0:04:20.520 and the people saying it can do that are partners 0:04:20.520 --> 0:04:23.279 at law firms that don't write motions but don't do 0:04:23.400 --> 0:04:27.760 like the grunt work. So it's it's almost it feels 0:04:27.800 --> 0:04:30.440 like the reporters have either given up or are just 0:04:30.480 --> 0:04:33.320 looking for clicks, and it's hard to tell. Sometimes. 0:04:33.440 --> 0:04:35.600 This is what I'm trying to figure out, because I'm 0:04:35.640 --> 0:04:40.159 realizing if it's entirely just I think this is directionally 0:04:40.200 --> 0:04:43.279 true and that's good enough, then they should be way 0:04:43.279 --> 0:04:46.680 more upset and in the streets and sparking a revolution, 0:04:46.839 --> 0:04:50.120 right Like, if you actually really believe fifty percent of 0:04:50.120 --> 0:04:52.000 the economy was going to be automated, that we're going 0:04:52.080 --> 0:04:55.240 to have to have government checks just so we can 0:04:55.279 --> 0:04:57.320 afford to buy the cat food to eat after all 0:04:57.320 --> 0:05:00.280 the jobs are gone. If you really thought that our 0:05:00.400 --> 0:05:03.480 entire infrastructure is about the collapse, that superintelligence was going 0:05:03.520 --> 0:05:06.560 to emerge suddenly and be a threat to human existence, 0:05:07.320 --> 0:05:09.880 you want to just right sort of too cool for 0:05:09.920 --> 0:05:12.440 school headshaking resignation article. You would be like, we got 0:05:12.560 --> 0:05:16.400 to where are the Don Connors? Right Like, we need 0:05:16.440 --> 0:05:19.000 to get on the cool trench coats and get out 0:05:19.000 --> 0:05:21.520 there and go against the Skynet revolution, like you would 0:05:21.560 --> 0:05:24.880 be on your feet, you'd be you know, nothing would 0:05:24.880 --> 0:05:27.039 be more important to you. So this is the this 0:05:27.120 --> 0:05:29.120 is my case about the text CEOs. I think there 0:05:29.360 --> 0:05:31.040 there's a moral hazard that I don't think that we're 0:05:31.040 --> 0:05:33.520 putting our finger on properly here. Right, So you have 0:05:33.640 --> 0:05:35.839 the text CEOs in the AI space that will just 0:05:35.839 --> 0:05:39.360 they'll just come out and just drop these bombs like, yeah, 0:05:39.480 --> 0:05:41.320 white collar blood. You know, I never actually said that 0:05:41.320 --> 0:05:42.680 that's Axios putting words in people. 0:05:42.760 --> 0:05:45.159 That was actually I thought that that was a He 0:05:45.240 --> 0:05:49.640 definitely Dario Amatay Warrio. He did say fifty, but I said, 0:05:50.480 --> 0:05:52.760 but not so he set the blood both. That's my bad. 0:05:53.000 --> 0:05:55.560 Well, I trust I. This is the New York fact 0:05:55.600 --> 0:05:57.920 checkers figured that out for me. But uh, Axios does 0:05:57.920 --> 0:06:00.240 a lot of this where they put like these really 0:06:00.360 --> 0:06:04.359 quotable quotes in the headlines about articles on interviews or 0:06:04.640 --> 0:06:07.279 speeches given my AI people, And it turns out the 0:06:07.320 --> 0:06:09.359 thing and the headline wasn't what they said. It was 0:06:09.400 --> 0:06:12.320 directly what they said. But anyways, so they're out there 0:06:12.360 --> 0:06:15.560 making these big statements. These jobs are going away. The 0:06:15.680 --> 0:06:17.719 Internet as we know it is about the all fall 0:06:17.720 --> 0:06:20.120 apart because of mythos is going to have this, this 0:06:20.160 --> 0:06:23.520 new capability, the super intelligence is coming. I'm I don't 0:06:23.520 --> 0:06:26.600 even know what's going to happen. There's two possible things 0:06:26.600 --> 0:06:28.359 going on here, and both of them are morally bad. 0:06:28.760 --> 0:06:30.600 One is which is the one I think is true, 0:06:30.640 --> 0:06:34.440 which is this is largely marketing. I look, this works, 0:06:34.480 --> 0:06:36.880 it gets reported, It keeps us seeming inevitable and important, 0:06:36.880 --> 0:06:38.839 in which case that's a huge moral hazard because you 0:06:38.920 --> 0:06:43.520 are making many many people normal people stressed the hell out, 0:06:43.760 --> 0:06:47.120 actively scaring them, actively scary them. The other option is 0:06:47.160 --> 0:06:49.400 you actually believe it's true. Well, this is an even 0:06:49.640 --> 0:06:52.960 larger moral trap that you've just fallen into because you 0:06:53.000 --> 0:06:57.080 are now perpetuating something that's going to cause exponentially more harm. 0:06:57.120 --> 0:06:59.240 You should be the very first person shutting down your 0:06:59.240 --> 0:07:00.640 company and trying to get the other ones that do 0:07:00.640 --> 0:07:03.159 it as well. So it's this weird world trap they've 0:07:03.160 --> 0:07:05.919 set up where whatever is actually going on here, if 0:07:05.960 --> 0:07:08.960 they're coming out here saying these things, it is bad. 0:07:09.560 --> 0:07:13.160 This can't possibly normalively speak. He be the right ethical 0:07:13.200 --> 0:07:15.040 behavior to be out there saying these scary things all 0:07:15.040 --> 0:07:17.360 the time, because either you need to be building the barricade, 0:07:17.800 --> 0:07:20.200 or you're just scaring people for the marketing. Neither of 0:07:20.200 --> 0:07:22.240 these I think is something that's defensible. 0:07:22.800 --> 0:07:25.800 I have a third and worse option, which is I 0:07:25.880 --> 0:07:29.520 choose Acxio. I think Axios there are some good reporters there. 0:07:29.880 --> 0:07:33.400 I think the leadership over there is disgusting. I think 0:07:33.480 --> 0:07:36.240 that they are aligning themselves with the companies. I think 0:07:36.240 --> 0:07:38.040 that what like if you watched it was a gym 0:07:38.280 --> 0:07:42.000 what's his name interviewing Sam Altman these I think that 0:07:42.040 --> 0:07:44.400 there is a level of and I would put this 0:07:44.440 --> 0:07:46.800 across people like Kevin Bruce and Casey Newton. These are 0:07:46.800 --> 0:07:51.360 my words, not cows that they're aligning them that they're saying, 0:07:51.480 --> 0:07:53.080 we think this is going to happen, and we're here 0:07:53.080 --> 0:07:55.520 to tell you great news. This is good news for 0:07:55.600 --> 0:07:58.960 me the writer, because I will be safe somehow I 0:07:59.000 --> 0:08:01.800 will be fine. You will not. You should be scared. 0:08:01.840 --> 0:08:06.200 But it's also a good thing because economy, marketing market good, 0:08:06.520 --> 0:08:09.200 and it is. It's a very incoherent message because it's like, 0:08:09.280 --> 0:08:12.600 to your point, Yeah, if this was a virus like 0:08:12.640 --> 0:08:16.840 a pandemic, you wouldn't be writing, Hey, millions of people 0:08:16.840 --> 0:08:21.000 are gonna die. What pretty good right, hey, or be 0:08:21.040 --> 0:08:23.720 good we'll have less people. That'd be good, right, It 0:08:23.760 --> 0:08:25.080 would be seen as peculiar. 0:08:25.560 --> 0:08:28.080 Someone did write that. Someone did write that, by the way, 0:08:28.280 --> 0:08:31.640 someone did They did say I remember early pandemics. Someone 0:08:31.720 --> 0:08:33.640 did write, Hey, you know what, this is good for 0:08:33.679 --> 0:08:36.160 the planet. Did it go on like, hey, we're driving lives, 0:08:36.240 --> 0:08:37.960 this is great and we're overpopulated. 0:08:38.040 --> 0:08:40.680 I go oh, I mean, I mean that's a different 0:08:40.760 --> 0:08:44.280 conversation that maybe, But in all seriousness, you didn't have 0:08:44.320 --> 0:08:49.280 mainstream media being like, well, COVID's gonna kill everyone the end, 0:08:49.360 --> 0:08:51.600 I get. I guess you know, maybe we'll just be 0:08:51.679 --> 0:08:55.000 inside forever. You didn't have this kind of stray. In fact, 0:08:55.040 --> 0:08:56.760 you had the direct propicy was we need to get 0:08:56.800 --> 0:08:59.200 outside again. Who cares about this thing? Well? 0:08:59.200 --> 0:09:01.480 And it's just yeah, go on, yeah. I think that's 0:09:01.480 --> 0:09:02.840 an interesting otle. I want to just pull on that 0:09:02.880 --> 0:09:04.840 thread a little bit because I think cod COVID gives 0:09:04.880 --> 0:09:07.280 an interesting I think it gives two different interesting observations 0:09:07.280 --> 0:09:10.760 to go in both directions. Right, So I think you're 0:09:10.840 --> 0:09:14.640 definitely right. What you're saying is when the pandemic was 0:09:14.679 --> 0:09:17.160 coming or it was getting bad, Really, a lot of 0:09:17.160 --> 0:09:20.760 the coverage was about what should we be doing or 0:09:21.440 --> 0:09:23.520 who are the people doing the wrong thing. But it 0:09:23.559 --> 0:09:26.200 was very much coming from this angle of like, Okay, 0:09:26.200 --> 0:09:27.760 we need to do whatever it is, like, we need 0:09:27.800 --> 0:09:29.800 to be better about this. It's got to be vaccines, 0:09:29.800 --> 0:09:31.960 it's got to be masks, it's got to be pick 0:09:32.000 --> 0:09:34.560 your mitigation, whether you like it or not. It was 0:09:34.800 --> 0:09:37.920 very focused on what should we be doing or who 0:09:38.000 --> 0:09:39.560 is it that's getting in the way of a plan 0:09:39.640 --> 0:09:41.920 that maybe would get us out of this. Which is 0:09:41.920 --> 0:09:44.079 where I think you're very right is that you did 0:09:44.080 --> 0:09:46.679 not see a lot of COVID pieces that were just well, 0:09:46.679 --> 0:09:48.160 I'm just going to kind of walk through like all 0:09:48.160 --> 0:09:50.280 the different ways you know, you might die and the 0:09:50.320 --> 0:09:52.439 morgues are going to fill up, and uh, you know 0:09:52.800 --> 0:09:58.319 that's COVID. But I also think what the other thing 0:09:58.360 --> 0:10:00.320 we saw on a lot of COVID coverage is something 0:10:00.360 --> 0:10:03.280 that we are seen in the AI coverage. That's where 0:10:03.520 --> 0:10:06.360 I saw a lot of the directionally true, not factually, 0:10:06.360 --> 0:10:09.120 but directly true. There's definitely a period early on in 0:10:09.160 --> 0:10:13.080 COVID because I was following that coverage quite carefully where 0:10:13.080 --> 0:10:15.920 the papers were thinking, Okay, this is the right behavior 0:10:17.120 --> 0:10:18.600 and they're probably right about a lot of these things, 0:10:18.640 --> 0:10:20.080 but I just would notice this. There'd be a lot 0:10:20.080 --> 0:10:22.640 of like, Okay, we need people to buy into, for example, 0:10:22.760 --> 0:10:25.280 the lockdowns or whatever. And there'd be a lot of 0:10:25.280 --> 0:10:27.800 directionally true reporting where maybe they would like put on 0:10:27.880 --> 0:10:30.520 a photo of a mass grave that was sort of 0:10:30.600 --> 0:10:33.400 unrelated the COVID, or you would see a lot of 0:10:34.160 --> 0:10:38.080 there'd be pushed back from like conservatives about schools, and 0:10:38.120 --> 0:10:40.120 then they put a lot of articles in the paper 0:10:40.520 --> 0:10:42.920 about teachers dying to COVID, even though it was they 0:10:42.960 --> 0:10:45.480 weren't in school, they got COVID elsewhere, And if you 0:10:45.600 --> 0:10:49.880 really pushed on it, it was because it's directionally true. 0:10:49.920 --> 0:10:53.080 Like the general or more general truth here is like 0:10:53.160 --> 0:10:55.719 we need to be worried about this or these mitigations work. 0:10:55.760 --> 0:10:59.200 It doesn't matter if this if this photo is actually right, 0:10:59.360 --> 0:11:01.960 or if this teaching who died in Orlando, the fact 0:11:02.000 --> 0:11:04.080 that they hadn't yet been back in a school building 0:11:04.160 --> 0:11:08.000 yet it doesn't it's serving the directional truth. So it's 0:11:08.040 --> 0:11:11.240 like it highlights something. COVID highlights something we're seeing now 0:11:11.280 --> 0:11:13.560 that the reporters that are doing directional reporting like we 0:11:13.559 --> 0:11:15.680 should be scared about it. I dare you not to 0:11:15.679 --> 0:11:17.120 be scared now, I dare not to be scared now. 0:11:17.200 --> 0:11:18.880 That's trying to ratchet it up. Yeah, but then you 0:11:18.920 --> 0:11:23.120 also get the contrast, which is this new style of 0:11:23.200 --> 0:11:26.000 just like headshaking resignation. And actually, I don't think the 0:11:26.040 --> 0:11:27.880 reporters think they're going to be safe. They're also like, 0:11:27.920 --> 0:11:29.760 writing's going to go away, the media is going to 0:11:29.800 --> 0:11:33.400 go away. So it's it's an almost like a nihilistic 0:11:33.920 --> 0:11:35.960 type of approach to this, like, yeah, I'm screwed, We're 0:11:35.960 --> 0:11:37.480 all screwed. What are we going to do? And that 0:11:37.600 --> 0:11:40.880 is definitely different than we saw during that last crisis, 0:11:40.880 --> 0:11:44.679 which was obviously much more actually severe than what's happening now. 0:11:44.720 --> 0:11:46.360 So it's really confusing me, to be honest. 0:11:46.679 --> 0:11:50.040 Well, the directionally reporting during COVID, yeah, probably shouldn't have, 0:11:50.120 --> 0:11:52.920 but at the same time it was in it was 0:11:52.960 --> 0:11:57.280 actually in pursuit of something good. That's an attempt to 0:11:57.320 --> 0:11:59.840 make people take this seriously, because that's ultimately what it was. 0:12:00.080 --> 0:12:03.360 Take this seriously. Don't go outside state, like, don't don't 0:12:03.360 --> 0:12:05.240 meet with people, don't be indoors with peop, blah blah 0:12:05.280 --> 0:12:08.320 blah blah blah. Great in this case, it's like, Yep, 0:12:08.360 --> 0:12:10.319 you should be scared of this, and what should you do? 0:12:10.640 --> 0:12:16.400 Fuck knows? Use jet GPT, I guess. And what's what's 0:12:16.480 --> 0:12:18.800 really confusing to me as well is you say, oh, 0:12:18.840 --> 0:12:21.520 these people don't think they'll be safe. For the most part, 0:12:21.559 --> 0:12:23.520 I just don't. I actually take back what I said. 0:12:23.559 --> 0:12:25.200 I think a lot of them just don't acknowledge it. 0:12:25.400 --> 0:12:28.720 They don't acknowledge the core ridiculousness of being like, well, 0:12:28.760 --> 0:12:32.120 everyone's jobs are going to get replaced, don't know, Like 0:12:32.440 --> 0:12:35.760 the Garfield meme with him looking at the Garfield with 0:12:35.400 --> 0:12:41.200 the crossout and the TV. Yeah, flawlessly described there. It's 0:12:41.280 --> 0:12:44.160 just it's frustrating as well, because it is terrifying people 0:12:44.240 --> 0:12:47.480 without Like I'm not saying literally axios or however, but 0:12:47.559 --> 0:12:50.840 stories like this are what made that mentally unstable person 0:12:50.920 --> 0:12:53.120 throw a Molotov cocktail at Sam Mormon's house. Like it's 0:12:53.160 --> 0:12:56.080 obvious that these people are scared of the AI doom, 0:12:56.280 --> 0:12:58.520 partly because to your point, what the fuck are we 0:12:58.559 --> 0:13:00.480 meant to do about it? Because using the these tools 0:13:00.520 --> 0:13:03.680 is not I don't really see how that works. Because 0:13:04.600 --> 0:13:07.520 if going along that line of logic, if the answer 0:13:07.600 --> 0:13:10.000 is you need to use this stuff now, but the 0:13:10.040 --> 0:13:12.679 eventual end point is that it's intelligent enough to do 0:13:12.760 --> 0:13:16.480 everything for you. How does using it now matter at all? 0:13:17.240 --> 0:13:20.960 Like what's the Surely chat GPT would be seen as 0:13:21.040 --> 0:13:23.760 like a rock versus a shotgun at that point, Like 0:13:23.840 --> 0:13:28.600 it's just technologically irrelevant if they get to AGI, which 0:13:28.600 --> 0:13:31.959 they probably won't, and it's just naturally illogical stuff. 0:13:32.440 --> 0:13:34.760 Yeah, and I'm with you. I've been making that same argument. 0:13:34.800 --> 0:13:36.960 This idea that you need to learn how to prompt 0:13:37.920 --> 0:13:41.560 some generation of a chatbot that exists right now is 0:13:41.600 --> 0:13:43.200 going to be the key to your long term I mean, 0:13:43.240 --> 0:13:45.920 even if, as you say, AI ends up playing a 0:13:46.200 --> 0:13:49.400 major sustained role in the economy, it's not going to 0:13:49.440 --> 0:13:52.080 be everyone typing on a web interface to a chat 0:13:52.080 --> 0:13:55.120 bot that sinka fantic and has a personality like I think. 0:13:55.320 --> 0:13:57.120 I think I've heard you say this recently and I 0:13:57.160 --> 0:13:58.920 agree with it as well, Like I don't think we 0:13:58.920 --> 0:14:02.840 should be chatting with technology. He should not be chatting 0:14:02.880 --> 0:14:05.600 in a sort of anthropomorphized, humanized way. Doesn't mean you 0:14:05.600 --> 0:14:08.840 can't do natural language processing. I mean Google's natural language processing. 0:14:08.880 --> 0:14:13.280 You're writing your Google searches in natural language, but no 0:14:13.280 --> 0:14:17.000 one's having a conversation with Google's. You list the keywords 0:14:17.040 --> 0:14:19.600 as quickly as possible, and Google's pretty good at figuring out, 0:14:19.760 --> 0:14:23.760 you know, population Spain nineteen eighty two. Any presenter, and 0:14:23.800 --> 0:14:25.640 you get that information. You're not like, hey, so I'm 0:14:25.640 --> 0:14:28.040 wondering what the population is of Spain in nineteen eighty two, 0:14:28.040 --> 0:14:30.560 can you help me find that question mark? There's something 0:14:30.600 --> 0:14:36.200 odd about that anthropomorphized conversational interface. I guess we saw 0:14:36.200 --> 0:14:37.960 a lot of Star Trek growing up, and that's, you know, 0:14:38.000 --> 0:14:39.960 what we think the future is supposed to be like, 0:14:40.040 --> 0:14:41.800 but it has all sorts of problems. 0:14:42.400 --> 0:14:45.720 Remembering Star Trek when he would go, computer do this, 0:14:45.920 --> 0:14:48.920 the computer didn't go. That's a great idea, Jean Luke, 0:14:49.200 --> 0:14:51.360 what a great idea. Thank you. If it's the computer 0:14:51.440 --> 0:14:54.000 just did the thing, that's like. I don't have any 0:14:54.040 --> 0:14:57.400 trouble with natural language grease because I think the whole reason, 0:14:57.480 --> 0:15:01.200 let's say, Chut GPT has grown comes from search. I 0:15:01.240 --> 0:15:03.800 think it is the core of it, because Chad GPT 0:15:04.120 --> 0:15:08.200 and Claude and all them are better at understanding what 0:15:08.280 --> 0:15:12.240 you asked for, not saying that they output is necessarily great, 0:15:12.280 --> 0:15:16.120 but just they understand the inference they make from what 0:15:16.200 --> 0:15:19.640 you say is better than Google, or at least better 0:15:19.680 --> 0:15:22.120 than Google has been. I feel like it was better before, 0:15:22.680 --> 0:15:25.720 and I think that had Google not kind of boothed 0:15:25.800 --> 0:15:27.960 it on this one, we wouldn't be in this spot. 0:15:28.000 --> 0:15:30.960 But even then, using Google now it forces you. It 0:15:31.000 --> 0:15:33.320 forces the AI summariz and you could do minus AI 0:15:33.400 --> 0:15:35.840 and all that. But sometimes I don't remember too, and 0:15:35.880 --> 0:15:40.360 it's just it's just turned search into this nightmare. But nevertheless, 0:15:41.080 --> 0:15:43.640 back to what you were saying, I agree. I think 0:15:43.680 --> 0:15:46.680 the anthropomorphization needs to go. I think that these things 0:15:46.720 --> 0:15:49.520 need to respond like terminal windows or what have you. 0:15:49.560 --> 0:15:51.800 They need to respond like computers and go, okay, here 0:15:51.800 --> 0:15:54.920 you go. Just don't need all that clutch. I don't 0:15:54.920 --> 0:15:57.920 need to be told, oh, what a great idea. I 0:15:57.960 --> 0:16:01.480 know I had it, or indeed, if I'm being told that, 0:16:01.560 --> 0:16:03.240 I need to be told if it's a bad idea. 0:16:03.320 --> 0:16:06.000 But I don't even necessarily need a nonswer. I just 0:16:06.040 --> 0:16:07.840 need stuff to look at so that I can come 0:16:07.840 --> 0:16:09.040 to my own conclusions. 0:16:09.160 --> 0:16:11.800 I think it's hard, actually, I think it's actually hard 0:16:11.800 --> 0:16:15.160 to get a language model to do that, right, Because 0:16:15.160 --> 0:16:17.720 if you think about when you go back to the 0:16:17.720 --> 0:16:19.640 base layer of what's happening in the pre training is 0:16:19.640 --> 0:16:21.480 that you're building a language model that's trying to win 0:16:21.520 --> 0:16:23.600 at the token guessing game. So it's I'm trying to 0:16:23.600 --> 0:16:25.560 guess what word or part of word actually comes next 0:16:25.600 --> 0:16:27.560 to what I assume to be a real piece of text. 0:16:27.600 --> 0:16:29.640 And then if you do that auto aggressively, so you 0:16:29.680 --> 0:16:31.640 call it again and again and again, adding the answer 0:16:31.720 --> 0:16:33.960 to the input so it grows out an answer, what 0:16:34.000 --> 0:16:37.480 you're going to get is a text expansion. You've given 0:16:37.480 --> 0:16:39.280 me a text that I'm trying to expand as if 0:16:39.320 --> 0:16:41.240 there is a real text that exists and I'm trying 0:16:41.240 --> 0:16:43.480 to match it. You get that like kind of indirectly. 0:16:43.760 --> 0:16:48.320 So really it's idiom is the type of text that's 0:16:48.360 --> 0:16:50.640 trained on, which for the most part is more sort 0:16:50.680 --> 0:16:53.520 of pro style text. So you can tune it away 0:16:53.560 --> 0:16:56.200 from it like you can tune its mood, you can 0:16:56.240 --> 0:16:59.960 tune its synofancy. But it might be hard to actually 0:17:00.120 --> 0:17:03.880 tune an LLM because it deals with human written pros 0:17:04.000 --> 0:17:06.240 as its main training data. It might be harder than 0:17:06.240 --> 0:17:10.240 we think to tune that away from being verbose and 0:17:10.280 --> 0:17:12.320 that just give a table. Now, I guess you could 0:17:12.320 --> 0:17:15.080 take its output and then maybe run that through another 0:17:15.160 --> 0:17:17.280 thing that then strips away the other piece. It's like, 0:17:17.320 --> 0:17:21.720 it's possible. But I think the anthropomorphized verbosity we see 0:17:21.760 --> 0:17:24.280 in language models is also that's kind of the native 0:17:24.359 --> 0:17:28.119 tongue of this particular, which is why we still have 0:17:28.280 --> 0:17:32.000 a lot of chatbots being emphasized, and tools that are 0:17:32.000 --> 0:17:35.160 built upon LLM as the digital brain are still way 0:17:35.200 --> 0:17:38.000 more scarce than you would imagine outside of maybe computer 0:17:38.040 --> 0:17:40.920 programming en coding harnesses, we just don't have a lot 0:17:40.920 --> 0:17:43.160 of other examples where we just use the LLM as 0:17:43.160 --> 0:17:46.560 a general person's digital brain, because I think this verbosity 0:17:46.960 --> 0:17:49.879 is okay, humans can interpret that, but it's not great. 0:17:50.320 --> 0:17:52.720 If the LM is just a digital brain that's interfacing 0:17:52.720 --> 0:17:55.760 between you and another computer, it doesn't need to hear 0:17:55.800 --> 0:17:57.560 that their idea is great or wants to try to 0:17:57.600 --> 0:17:59.280 parse the different types of text. So there's some interesting 0:17:59.280 --> 0:18:01.720 things going on there about the fundamental nature of these things. 0:18:02.000 --> 0:18:06.200 But even then with Google Aiye mode, it's still seems 0:18:06.440 --> 0:18:09.800 kind it's still like, actually seems like it can give 0:18:09.880 --> 0:18:13.119 fairly short answers, but if you mess, if you argue 0:18:13.119 --> 0:18:16.520 with it, as I have. It will just provide you it. 0:18:16.600 --> 0:18:20.000 Even Googles will provide you with just hot dogshit. Yeah, 0:18:20.040 --> 0:18:22.600 like it will just claim something is true. Why one, 0:18:22.640 --> 0:18:25.159 I just did a private EQ thing on private credit 0:18:25.160 --> 0:18:27.600 even And my favorite thing is being like, what fund 0:18:27.680 --> 0:18:29.040 is this part of? And it goes, it's part of 0:18:29.040 --> 0:18:32.240 this fun. That fund was fund that was founded after 0:18:32.320 --> 0:18:35.879 this happened, And it goes, okay, well maybe it's this 0:18:35.920 --> 0:18:38.720 one different fun three years old. Doesn't it not involved? 0:18:38.800 --> 0:18:39.680 Do you have proof of that? 0:18:39.840 --> 0:18:41.800 Well, this is what you don't know. This is what 0:18:41.840 --> 0:18:44.600 you don't see in Star Trek. Is you know Captain 0:18:44.680 --> 0:18:47.360 Kirk or whoever I'm going to mix up the episodes here, Yeah, 0:18:47.440 --> 0:18:50.560 say like, hey, computer, we are approaching deep space nine 0:18:50.640 --> 0:18:54.600 prepared docking procedures and computer is like photon torpedo fired, 0:18:54.840 --> 0:18:57.119 sea destroyed. And you're like, well no, I said we're 0:18:57.119 --> 0:18:59.680 supposed to dock. Oh you're right, Kirk, I should enough 0:18:59.720 --> 0:19:04.679 fire the thanking me accountable. That was I did the 0:19:04.960 --> 0:19:07.400 opposite thing, you know, Yeah, that didn't happen in Star Trek. 0:19:18.680 --> 0:19:22.440 So one thing that's really been driving me insane, by 0:19:22.440 --> 0:19:24.560 which I mean going on Twitter, is looking at people 0:19:24.600 --> 0:19:28.240 like Aaron Levy of Box and Brian Armstrong of coinbase 0:19:28.480 --> 0:19:31.320 talking about like agents spending money in the agentic web 0:19:31.320 --> 0:19:33.960 and how we need to prepare the web for agents 0:19:34.000 --> 0:19:38.720 doing stuff and the agents will do this fantastical doesn't exist. 0:19:38.760 --> 0:19:41.159 Agents don't do that, just like not they don't have 0:19:41.240 --> 0:19:44.640