WEBVTT - Radio Better Offline: Edward Ongweso Jr. & Allison Morrow 0:00:02.400 --> 0:00:03.560 Also media. 0:00:05.240 --> 0:00:07.800 Hello and welcome to Better Offline. I'm your host Ed Zichron, 0:00:19.640 --> 0:00:21.920 and we are recording here in beautiful New York City, 0:00:22.079 --> 0:00:24.040 and I have a wonderful pair of guests today. I have, 0:00:24.360 --> 0:00:28.520 of course, Alison Morrow from CNN Nightcap Newsletter and Edward 0:00:28.600 --> 0:00:32.560 Anguiso Junior, the hater himself from the Tech Bubble newsletter. 0:00:32.800 --> 0:00:34.400 Thank you so much for joining me today. 0:00:34.880 --> 0:00:35.760 Great to be here as all. 0:00:36.680 --> 0:00:38.839 So I think we should start with exactly what we 0:00:38.840 --> 0:00:41.000 were just talking about. The open AI claims that they 0:00:41.040 --> 0:00:44.000 have worked out what causes hallucinations. Alison, do you want 0:00:44.000 --> 0:00:44.519 to go over this? 0:00:45.640 --> 0:00:47.479 I should have read the paper a bit more carefully, 0:00:47.520 --> 0:00:50.280 but I can you know. The highlights were getting digested 0:00:50.479 --> 0:00:54.640 yesterday on X and Blue Sky, and it seems like 0:00:55.200 --> 0:00:58.240 it's kind of the test taking problem of when you 0:00:58.360 --> 0:01:01.440 encourage students when they're taking anderdized tests and you don't 0:01:01.480 --> 0:01:04.200 know the answer, you guess, And that's exactly how these 0:01:04.280 --> 0:01:07.360 models are trained, and they don't. You don't get a 0:01:07.400 --> 0:01:10.279 point if you say, I don't know, come up with something, 0:01:10.360 --> 0:01:16.560 and the models are meant to keep guessing until they 0:01:16.560 --> 0:01:19.520 get something close to right. Right, So that's why you 0:01:19.560 --> 0:01:24.839 get a lot of nonsense and hallucinations and open AI 0:01:25.000 --> 0:01:27.399 at least in their reading of it, says oh, this 0:01:27.480 --> 0:01:30.200 is a simple solution. We'll just encourage the models to 0:01:30.560 --> 0:01:34.080 understand better when it's like a binary question and when 0:01:34.319 --> 0:01:36.280 you can say I don't know the answer to that, 0:01:36.760 --> 0:01:37.360 so we'll see. 0:01:38.000 --> 0:01:40.759 I read the paper, albeit not today because my brain 0:01:40.840 --> 0:01:43.440 just immediately read it as marketing copy. Just put that, 0:01:43.680 --> 0:01:47.560 replaced it with another anime theme song. I went through 0:01:47.600 --> 0:01:49.680 it and I was like, okay, so it's going to 0:01:49.720 --> 0:01:52.400 encourage them to say I don't know. This feels like 0:01:52.480 --> 0:01:54.920 a very flat view of what hallucinations are, though, because 0:01:55.000 --> 0:01:58.600 hallucinations as people know them authoritatively stating something that isn't true, 0:01:58.880 --> 0:02:02.040 but hallucinations in life coding model are. It will just 0:02:02.040 --> 0:02:04.680 say yeah, I did that when it didn't. This is 0:02:04.800 --> 0:02:06.680 very common you going through the Cursor and the Claude 0:02:06.680 --> 0:02:09.360 code forums, you can see this or subredits at least, 0:02:09.560 --> 0:02:12.400 and it's not that they say something that's true. They 0:02:12.400 --> 0:02:15.919 don't know that something's not true also, so they might 0:02:16.000 --> 0:02:19.239 say they don't know, and it's just very silly because 0:02:20.000 --> 0:02:22.520 they claim that they're going to fix this problem with 0:02:22.600 --> 0:02:25.359 this solution, but they've done years and hundreds of millions 0:02:25.360 --> 0:02:28.440 of dollars of reinforcement learning. Do they unreinforce do they 0:02:28.480 --> 0:02:33.040 reinforce it? More so that I don't know. I'm fucking 0:02:33.080 --> 0:02:34.799 tired of these companies are going to be completely honest. 0:02:34.800 --> 0:02:36.440 I realized this is kind of cliche for the show, 0:02:36.480 --> 0:02:38.920 but it's the fact that these things get written up 0:02:38.919 --> 0:02:41.320 as very serious things. US is saying, you, guys still 0:02:41.320 --> 0:02:43.360 haven't worked this out. It's kind of frustrating to me. 0:02:44.400 --> 0:02:46.720 And they don't seem to be the models. Some getting 0:02:46.760 --> 0:02:49.120 bad dimension returns and all that, but this is the 0:02:49.120 --> 0:02:49.960 best they've got. 0:02:50.280 --> 0:02:54.400 I mean, does this kind of thrust feel like, you know, 0:02:54.400 --> 0:02:56.400 do you guys feel like it's a downstream of the 0:02:56.480 --> 0:02:59.000 attempts by some of these firms to say, oh, actually, 0:02:59.000 --> 0:03:03.120 like if we dialor bad the sick a fancy, you know, 0:03:03.280 --> 0:03:05.520 then we'll be able to have a much more engaging 0:03:05.600 --> 0:03:08.800 consumer product. We'll be able to have it hallucinate less, 0:03:09.280 --> 0:03:11.040 you know, and do psychosis less. 0:03:11.480 --> 0:03:11.680 You know. 0:03:11.880 --> 0:03:13.639 But does it do they feel linked in that way? 0:03:13.720 --> 0:03:16.440 Does it just feel like a they you know, another 0:03:16.639 --> 0:03:17.760 maybe another dead end. 0:03:18.080 --> 0:03:20.080 I feel like it's just them trying to work shit 0:03:20.160 --> 0:03:22.760 out and trying to like this also feels like a 0:03:22.840 --> 0:03:26.079 very rudimentary answer that they probably already had. It doesn't. 0:03:26.720 --> 0:03:28.880 Anytime someone comes up with an idea that I can 0:03:28.919 --> 0:03:31.200 that's technical, I'm like, okay, mate, you cannot be cooking 0:03:31.240 --> 0:03:33.760 with gas. This is this cannot be that good an idea. 0:03:33.800 --> 0:03:35.640 And the thing is the sick of fancy problem I 0:03:35.640 --> 0:03:38.520 don't think is solvable through solving hallucinations. The problem is 0:03:38.680 --> 0:03:41.880 it should stop. It needs to just It should not 0:03:41.920 --> 0:03:44.440 say I don't understand, it should say no. Actually, you 0:03:44.520 --> 0:03:46.680 sound like you think I'm God King and that you 0:03:46.720 --> 0:03:48.480 are God King, and that as God kings together we 0:03:48.480 --> 0:03:50.440 will destroy the world. Yeah. I mean, I mean my 0:03:50.560 --> 0:03:51.200 case is true. 0:03:51.280 --> 0:03:54.160 But do you think like the emphasis or the attempt 0:03:54.200 --> 0:03:56.440 to overcorrect on that is leading them to go down 0:03:56.440 --> 0:03:58.640 solutions where they think, oh, like if we just like 0:03:58.760 --> 0:03:59.600 this and that, then we. 0:03:59.800 --> 0:04:02.880 Yeah, yes, yes, actually I think that that's right because 0:04:02.880 --> 0:04:04.360 they have the I don't know if you saw the 0:04:04.400 --> 0:04:07.640 Attorney general. Attorney's general from Delaware and California sent a 0:04:07.720 --> 0:04:11.000 letter to open AI last week saying, hey, look, if 0:04:11.040 --> 0:04:13.520 you need to fix these safety protocols, you need to 0:04:13.520 --> 0:04:15.440 actually have them, because what you have right now doesn't work, 0:04:15.440 --> 0:04:18.400 and we will block your nonprofit otherwise. I was really 0:04:18.440 --> 0:04:20.240 happy to read that right up until the bit they 0:04:20.279 --> 0:04:23.480 say we wish you well with your quest for AI dominance. 0:04:23.520 --> 0:04:26.680 I'm just like, these are the fucking people protecting us 0:04:26.720 --> 0:04:28.800 from the They're like, no, it's great, you want to 0:04:28.839 --> 0:04:32.359 dominate everyone with AI. It's just you drove some people 0:04:32.400 --> 0:04:34.000 to a murder suicide situation. 0:04:34.120 --> 0:04:36.880 And wasn't that part of the problem with GPT five 0:04:36.960 --> 0:04:39.279 as they tried to dial back the SICKA fancy and 0:04:39.320 --> 0:04:43.160 then it took away the character and like the humanness 0:04:43.160 --> 0:04:46.240 that people had gotten attached to in GPT four, yeah, 0:04:46.279 --> 0:04:49.280 in earlier models, and like it all seems to come 0:04:49.320 --> 0:04:52.440 back to open. AI doesn't realize that what it's selling 0:04:53.040 --> 0:04:57.600 often is like a companion and a therapist, and it doesn't. 0:04:57.720 --> 0:05:00.000 It reminds me of like Q tips. You're not supposed 0:05:00.040 --> 0:05:02.240 but them in your ear right, but that's all anyone 0:05:02.320 --> 0:05:03.080 uses qtubes. 0:05:03.360 --> 0:05:06.160 I was gonna say they're not meant to do, of. 0:05:06.120 --> 0:05:09.359 Course not, but like that's the consumer, that's how the 0:05:09.400 --> 0:05:13.440 consumer has chosen to use this product. And they're saying like, well, 0:05:13.480 --> 0:05:15.680 we don't condone that, we don't think it's like the 0:05:15.720 --> 0:05:18.240 best use of our product, and you know, we know 0:05:18.320 --> 0:05:20.359 better than the than the consumer, of course. 0:05:20.480 --> 0:05:22.760 I actually I think it's one abstraction high, which is 0:05:22.800 --> 0:05:25.360 I don't think they know what chat GPT is. I don't. 0:05:25.480 --> 0:05:28.800 I did AI boaster piece came out last week, but 0:05:28.920 --> 0:05:30.880 it was I had this whole thing where it was. 0:05:31.360 --> 0:05:33.840 It's so they don't describe what chat GPT does. If 0:05:33.839 --> 0:05:36.200 you're on the website, like it's like, yeah, analyzes data 0:05:36.240 --> 0:05:40.520 and it's it's brainstorming and charts and stuff. The agent 0:05:40.560 --> 0:05:44.200 does things too, Please buy it and then you try 0:05:44.200 --> 0:05:46.680 and like actually look for any use cases and there's nothing. 0:05:46.760 --> 0:05:48.840 I think that they're just guessing. But my favorite thing 0:05:48.839 --> 0:05:51.560 I'm seeing, my favorite thing is that now people are 0:05:51.600 --> 0:05:54.560 like GPT four h isn't the same because they brought 0:05:54.560 --> 0:05:57.159 it back and now people are just freaking out. People like, no, 0:05:57.240 --> 0:06:00.359 it's not the same, it's different. Somehow it's on I 0:06:00.360 --> 0:06:02.359 honestly don't know if it's true. I just think that 0:06:02.400 --> 0:06:04.760 they've entered gamer mode. This is just what. This is 0:06:04.760 --> 0:06:06.880 how gamers react. It's like the gun isn't the same, 0:06:06.920 --> 0:06:10.040 it's literally the same code. No, you've changed something, You've 0:06:10.360 --> 0:06:13.200 I know, and it's what happens when you release an 0:06:13.240 --> 0:06:18.080 imprecise glazing bot onto the world. It's also really funny 0:06:18.120 --> 0:06:20.839 how literally any of these companies could have been open AI. 0:06:21.040 --> 0:06:23.200 There doesn't seem to be anything special about this company 0:06:23.200 --> 0:06:24.880 at all anymore or ever. 0:06:24.800 --> 0:06:27.720 Well you know, I mean, not everybody can lose as 0:06:27.800 --> 0:06:32.360 much money that feels a very special or have the 0:06:32.400 --> 0:06:33.840 connection to you. 0:06:34.200 --> 0:06:36.280 Oh my god. What I haven't heard from Maco's son 0:06:36.279 --> 0:06:39.480 in the minute. We haven't had an announcement of tomorrow 0:06:39.640 --> 0:06:42.320 or tomorrow is gonna be? I like that he bought 0:06:42.360 --> 0:06:45.039 a he bought a Fox count Lab, an old Fox 0:06:45.080 --> 0:06:48.480 con lab too turned into an AI server building place 0:06:48.520 --> 0:06:51.599 in like Ohio. I think it is. It's like, mate, 0:06:51.839 --> 0:06:54.839 what are you doing? Man? Are you okay? Someone needs 0:06:54.880 --> 0:06:57.279 to check in on mass We need to hold space 0:06:57.320 --> 0:06:58.640 for Masso Shei's son. 0:06:58.839 --> 0:06:59.560 Yeah, bring them on. 0:07:00.640 --> 0:07:03.640 I have asked him. I have genuinely emailed soft Banks 0:07:03.680 --> 0:07:07.520 PR and I've emailed Anthropics PR for Dario and han't 0:07:07.560 --> 0:07:09.680 heard back from either. I assume there's an email issue 0:07:10.000 --> 0:07:12.760 because there's no other reason they wouldn't come on this show. 0:07:13.400 --> 0:07:16.640 It's just it now. I what's fun to watch though 0:07:16.680 --> 0:07:18.920 is considering how many times we've had like AI bubble 0:07:18.960 --> 0:07:22.040 conversations on here, everyone seems to be kind of waking 0:07:22.120 --> 0:07:24.559 up to it. It's kind of fun. Must be clear, 0:07:25.400 --> 0:07:28.280 everyone here was early on this. I was new, actually 0:07:28.400 --> 0:07:30.680 very early. You've been early on every day as early 0:07:30.680 --> 0:07:32.920 as you No, I give you credit. You're one of 0:07:33.000 --> 0:07:35.360 like the three people who, when the metaverse was happening, 0:07:36.080 --> 0:07:38.680 was actually calling it out. So it's good. It's good 0:07:38.680 --> 0:07:41.920 to see in here, but also insane that it's still going. 0:07:42.160 --> 0:07:44.320 That's what I don't like. This open a one hundred 0:07:44.320 --> 0:07:47.280 and fifteen billion spurned within by twenty twenty eight, I 0:07:47.280 --> 0:07:50.320 think it is. I don't understand how people keep publishing 0:07:50.360 --> 0:07:52.200 this and being like and that will happen. 0:07:52.360 --> 0:07:54.760 I will say, to my relief, the metaverse went away 0:07:54.880 --> 0:07:57.640 quickly because when it first was announced and I wrote 0:07:57.640 --> 0:07:59.280 a piece just being like what. 0:07:59.600 --> 0:08:00.760 Yeah, what is what? 0:08:01.840 --> 0:08:04.440 And then like everything about it seemed really dumb in 0:08:04.480 --> 0:08:07.679 every iteration coming after, and so I was like, oh, okay, 0:08:07.920 --> 0:08:10.400 I wasn't crazy, but with AI and I know you 0:08:10.400 --> 0:08:14.600 can relate to this, I feel crazy because the lack 0:08:14.680 --> 0:08:18.000 of utility is still there and like the absurdity of 0:08:18.040 --> 0:08:22.200 the investment is still there, and it does seem like that. 0:08:22.200 --> 0:08:24.160 That's why I wrote about the vibe shitt like it 0:08:24.240 --> 0:08:27.360 has been like going in Edzitron's favor in the last 0:08:27.360 --> 0:08:27.960 few weeks. 0:08:28.560 --> 0:08:32.120 Fucking vindication finally. But it's it's funny as well because 0:08:32.280 --> 0:08:36.079 even with them, this open Ai story comes out and 0:08:36.120 --> 0:08:37.800 people are still like, yeah, they're gonna burn one hundred 0:08:37.800 --> 0:08:41.000 and fifteen billion dollars. People reported about a month ago 0:08:41.040 --> 0:08:43.280 that they're going to spend thirty billion dollars a year 0:08:43.280 --> 0:08:48.480 on Oracles starting twenty twenty eight. How how how? I 0:08:48.640 --> 0:08:51.600 just That's what I don't understand. How these articles keep 0:08:51.600 --> 0:08:53.560 coming out. And I understand that reporters have to report 0:08:53.760 --> 0:08:55.839 the news and things they have discovered. I get it, 0:08:56.000 --> 0:08:57.600 But can no one just be like and no one 0:08:57.679 --> 0:08:59.839 has any idea where this money's coming from. No one, 0:09:00.240 --> 0:09:04.640 thirty billion dollars a fucking year from Oracle for service 0:09:04.679 --> 0:09:06.200 that are yet to be built, and all of this. 0:09:06.400 --> 0:09:08.240 They don't need to worry about the fact that the 0:09:08.280 --> 0:09:11.079 servers aren't actually built. In Applene Texas is not finished, 0:09:11.080 --> 0:09:13.760 and the money doesn't exist, and it isn't obvious how 0:09:13.840 --> 0:09:16.040 Oracle affords it they will have to take on debt 0:09:16.080 --> 0:09:19.120 to build this, and Crusoe and primary digital Infrastructure have 0:09:19.160 --> 0:09:21.520 already done that. And I mean, other than that, it 0:09:21.559 --> 0:09:24.439 could happen any day. I just wonder if the media 0:09:24.559 --> 0:09:26.319 is actually not prepared for this. And I don't mean 0:09:26.320 --> 0:09:29.400 in a conspiratle way, I mean just in a is 0:09:29.440 --> 0:09:32.760 the media actually set up for companies just lying or 0:09:32.880 --> 0:09:34.200 just projecting. 0:09:34.559 --> 0:09:37.880 I mean, I feel like it reminds me of kind 0:09:37.880 --> 0:09:40.480 of the relationship that maybe our critics or you know, 0:09:40.559 --> 0:09:43.520 our coverage might have for the medium, where there's not 0:09:43.600 --> 0:09:49.240 an inherent antagonism or skepticism of claims that are being offered, 0:09:49.920 --> 0:09:54.320 an assumption of good faith that continually gets betrayed or 0:09:54.400 --> 0:09:57.400 you know, punished, but gets carried on with over and 0:09:57.440 --> 0:10:00.520 over and over again. I feel like the AI bubble 0:10:01.360 --> 0:10:03.880 discussions that we're seeing. Part of me also feels like 0:10:04.080 --> 0:10:06.720 they are going to disappear the second we start to 0:10:06.800 --> 0:10:10.439 see some of the firms maybe announce some favorable metrics, 0:10:10.480 --> 0:10:13.319 even though, as we've been talking about for a long time, right, 0:10:13.360 --> 0:10:16.080 revenues are not there, profits are not there now, the 0:10:16.120 --> 0:10:20.920 burn is only increasing, right, and there's no way forward 0:10:22.000 --> 0:10:23.880 in the short term. Right, you know that I can 0:10:23.920 --> 0:10:26.640 think of where these companies start to actually do the 0:10:26.640 --> 0:10:28.120 things that they're claiming they're going to be doing with 0:10:28.240 --> 0:10:31.440 transforming the world. But I can see a scenario where 0:10:31.679 --> 0:10:34.360 you know, someone has a favorable quarter for adoption even 0:10:34.400 --> 0:10:37.000 though you know, we just I just saw yesterday Apollo 0:10:37.040 --> 0:10:40.640 Global Management was talking about large firms are actually scaling 0:10:40.679 --> 0:10:45.000 back AI adoption, which already wasn't even providing returns and 0:10:45.080 --> 0:10:46.800 was hurting productivity in the first place. 0:10:46.880 --> 0:10:47.080 Right. 0:10:47.160 --> 0:10:49.440 It's and it's really weird as well, because there was 0:10:49.440 --> 0:10:52.040 the MIT study where it's like they said, ninety five 0:10:52.080 --> 0:10:55.480 percent of generative AI integrations don't have any return on investment. 0:10:55.640 --> 0:10:58.360 There are some people critiquing that number, but something that 0:10:58.360 --> 0:10:59.840 comes out of that study that I really like was 0:10:59.880 --> 0:11:03.120 it was saying that enterprise adoption is high, but the 0:11:03.160 --> 0:11:06.480 actual transformation is low because this shit doesn't work. And 0:11:06.600 --> 0:11:10.400 it's it's so strange, and I think that the only 0:11:10.480 --> 0:11:13.080 reason that things won't immediately unrebel with a good quarter 0:11:13.240 --> 0:11:16.559 is because the media has chosen to follow a direction. 0:11:16.679 --> 0:11:20.120 Now you've got the when you've got the Atlantic, the 0:11:20.200 --> 0:11:23.839 goddamn Atlantic publishing a story saying yeah, it turns out 0:11:23.880 --> 0:11:26.080 that AI isn't really doing much, but it's holding up 0:11:26.120 --> 0:11:29.480 our economy. It's like, holy shit, the Atlantic's willing to 0:11:29.480 --> 0:11:32.760 admit something that happened happened. I didn't even know this 0:11:32.880 --> 0:11:35.640 was in there. I thought that they just wrote up 0:11:35.640 --> 0:11:37.559 whatever was email to them by and dohold. 0:11:37.320 --> 0:11:39.080 Your bread, they just published Mike Solana. 0:11:40.200 --> 0:11:43.520 Yeah, actually I retract all my statements to But it's 0:11:43.760 --> 0:11:45.800 I think with the media. 0:11:46.120 --> 0:11:49.720 I see it in political reporting and business reporting in tech. 0:11:50.559 --> 0:11:55.640 There's a deference to authority that I think American media, 0:11:55.720 --> 0:11:59.800 but all media, have an issue with. And I think 0:11:59.840 --> 0:12:02.880 that sort of speaks to the underlying economics of being 0:12:02.960 --> 0:12:06.840 in media right now, where there's a general chill both 0:12:06.880 --> 0:12:12.240 economically and politically. Reporters are worried about their bylines being 0:12:12.320 --> 0:12:14.679 out there and getting stuff wrong. And I'm not saying 0:12:14.679 --> 0:12:16.640 that that's an excuse, but I do think that is 0:12:16.679 --> 0:12:20.319 an institutional mindset that has taken route, especially in the 0:12:20.400 --> 0:12:23.200 last ten years, that's just become like really hard to 0:12:23.200 --> 0:12:26.400 be a journalist and to do it right. But you 0:12:26.520 --> 0:12:29.199 are starting to see that MIT report was so important 0:12:29.200 --> 0:12:34.040 because it caused people on Wall Street authority figures to 0:12:34.480 --> 0:12:37.439 say hmm, I don't know about this, And then that 0:12:37.480 --> 0:12:41.040 got a lot of mainstream financial media to kind of 0:12:41.080 --> 0:12:44.680 like do that questioning headline about AI that they maybe 0:12:44.720 --> 0:12:46.160 wouldn't have done six months ago. 0:12:46.440 --> 0:12:48.440 I do like that there's a rumor about the next 0:12:48.440 --> 0:12:52.080 deep Seek model doing agents. They're not going to, but 0:12:52.200 --> 0:12:54.120 even just like if that comes out and they even 0:12:54.120 --> 0:12:56.080 claim they can, I think that might have a market 0:12:56.080 --> 0:12:58.439 panic just because they'll be like, ah, China, I. 0:12:58.360 --> 0:13:00.800 Mean yeah, I mean that's I. I actually think there's 0:13:00.840 --> 0:13:03.000 something to that, right, because you know, we did agents here, 0:13:03.360 --> 0:13:06.760 didn't do shed for sales powers really the market did 0:13:06.760 --> 0:13:07.439 the market even? 0:13:07.480 --> 0:13:11.920 Really? Actually and a wonderful story from this information open 0:13:11.960 --> 0:13:15.559 AI in their in their whole projections through twenty twenty nine, 0:13:15.559 --> 0:13:17.600 they've reduced the amount of money of the micromagents by 0:13:17.600 --> 0:13:23.880 twenty six billion dollars. Why I want anyway? So, but yeah, 0:13:23.920 --> 0:13:27.200 it's like that this deep Seek thing could inspire people 0:13:27.240 --> 0:13:30.319 to get scared horribly because no one actually believes that 0:13:30.360 --> 0:13:33.400 agents exist. Looks they don't, but they think they do, 0:13:33.440 --> 0:13:36.120 but they will, but they won't. It's this I don't 0:13:36.120 --> 0:13:38.240 think I've ever actually seen anything like this in tech. 0:13:38.240 --> 0:13:40.760 I'm gonna be honest, it's worse than crypto, even worse 0:13:40.800 --> 0:13:42.720 than just the general generative AI. I think is this 0:13:42.800 --> 0:13:45.719 concept of agents though, because I saw some fucking thing 0:13:45.800 --> 0:13:49.440 about some one political blog saying that Donald Trump would 0:13:49.440 --> 0:13:52.080 do some sort of act he would do the I 0:13:52.120 --> 0:13:53.840 forget the exact thing, but it would be an act 0:13:53.880 --> 0:13:57.839 that would make AI copyright holders just handshit over to 0:13:57.880 --> 0:14:00.800 AI due to us needing to beat China. And it 0:14:00.880 --> 0:14:04.119 mentioned with you know, like, yeah, the growing agency capabilities 0:14:04.160 --> 0:14:06.280 of AI. It's just what the fuck is that. I've 0:14:06.280 --> 0:14:08.199 never seen a tech thing in my life now that 0:14:08.320 --> 0:14:10.600 has not existed like this, and people talk about it 0:14:10.640 --> 0:14:11.200 like it's real. 0:14:11.760 --> 0:14:14.280 And I think also it's interesting to see the more 0:14:14.440 --> 0:14:18.080 it doesn't manifest, the more some of the recommendations to 0:14:18.160 --> 0:14:21.360 make them happen just sound more and more like these 0:14:21.360 --> 0:14:24.440 are also things that might somehow bend the cost curve 0:14:24.440 --> 0:14:27.560 in our direction. Maybe we should make the Internet unusable 0:14:27.600 --> 0:14:30.360 to anything other than some of these programs. Yeah right, 0:14:30.840 --> 0:14:35.160 And you know, I'm curious which one is going to 0:14:35.240 --> 0:14:38.640 give out first, Like the really savvy ability a lot 0:14:38.640 --> 0:14:43.040 of these firms have had in spinning, sputtering or you know, 0:14:43.320 --> 0:14:47.920 spinning you know, a crisis that might drain the markets 0:14:48.000 --> 0:14:50.560 or deter investors and to oh, actually we just need 0:14:50.600 --> 0:14:53.320 even more capital, similar to what they did with deep Seek. 0:14:53.640 --> 0:14:57.120 The solution is even more compute intensive models. You know, 0:14:57.160 --> 0:14:58.880 whether they're going to be able to do that faster 0:14:59.000 --> 0:15:02.120 than people wising up and saying, you know, maybe we 0:15:02.160 --> 0:15:05.960 shouldn't misallocate trillions of dollars a capital over the next 0:15:05.960 --> 0:15:06.840 few years towards this. 0:15:07.000 --> 0:15:09.880 But this is the thing, though, I don't think there's 0:15:09.920 --> 0:15:13.400 anything stopping this, because the suggested thing was the okay, 0:15:13.440 --> 0:15:16.040 so all of these modial companies can steal from everyone, 0:15:16.160 --> 0:15:18.880 which is what happened already even with this anthropic settlement. 0:15:18.960 --> 0:15:21.400 It's not it's great people are getting three thousand dollars. 0:15:21.440 --> 0:15:22.880 I love that, but. 0:15:22.680 --> 0:15:24.840 It's also what the companies are offering that's similar to 0:15:24.840 --> 0:15:27.600 what the your payout would have been if you said, yeah, right, 0:15:28.640 --> 0:15:31.040 I didn't know that. I think like some publishers were offering, 0:15:31.360 --> 0:15:33.800 or you know, trying to ask authors, Hey, if we 0:15:33.880 --> 0:15:36.440 pay you this, would you allow your book to be 0:15:36.520 --> 0:15:39.320 trained on would you allow us, you know, be put 0:15:39.360 --> 0:15:42.640 into a data set. The payment that you're getting from 0:15:42.640 --> 0:15:45.120 the settlement feels or reminds me of the amount of 0:15:45.120 --> 0:15:50.720 money that people are being offered and those sorts of deals. 0:15:58.600 --> 0:16:02.480 I think the thing is, Okay, they already steal everything. Well, okay, 0:16:02.520 --> 0:16:04.120 we need to give them as much money as possible. 0:16:04.120 --> 0:16:06.320 We've already done that. Are we just going to do 0:16:06.360 --> 0:16:08.120 this forever? Because even if we do this for rever, 0:16:08.200 --> 0:16:11.040 nothing's going to change, even if I'm completely wrong and 0:16:11.120 --> 0:16:15.360 open ai keeps going another five years. Okay, so we're 0:16:15.400 --> 0:16:17.400 just going to annihilate one hundred and fifteen billion dollars. 0:16:17.560 --> 0:16:19.840 There is no there are no more things here like 0:16:20.320 --> 0:16:24.480 they're projections open ays projections from the information their chart. 0:16:24.520 --> 0:16:26.200 I don't even need to show you this because this 0:16:26.360 --> 0:16:29.680 is just fan fiction at this point. There is in 0:16:29.720 --> 0:16:32.120 twenty twenty six starting there is this growth of this 0:16:32.200 --> 0:16:36.560 orange thing that is other revenue. Who knows what it is? 0:16:37.200 --> 0:16:39.360 I don't know. Open Ai doesn't seem to And that's 0:16:39.360 --> 0:16:41.200 really important because they're going to make what looks like 0:16:41.280 --> 0:16:45.480 several billion dollars from this next year. What the fuck 0:16:45.600 --> 0:16:48.280 is going Every time I look at this company, I 0:16:48.280 --> 0:16:50.520 feel a little more insane because they've now lowered their 0:16:50.560 --> 0:16:53.600 expectations of selling their access to their models by five 0:16:53.600 --> 0:16:56.240 billion dollars over the next few years. What even is 0:16:56.280 --> 0:16:58.720 open ai at this point? Is it just a wrap? 0:16:58.760 --> 0:17:01.600 Have they become a wrapper company of their own models? 0:17:01.880 --> 0:17:05.359 They're no better than Cursor. It's just it's so weird, 0:17:05.480 --> 0:17:07.359 and I realize them kind of going in circles at 0:17:07.359 --> 0:17:12.520 this point, but it's I even the metaverse, even Crypto, 0:17:12.760 --> 0:17:15.800 even Crypto functioned. It was bad. It's still bad. It 0:17:15.880 --> 0:17:19.000 is bad cloud software, but it still did the thing. Hey, 0:17:19.040 --> 0:17:22.480 I doesn't even seem to be doing it, and they 0:17:22.480 --> 0:17:24.480 need more money to prove that it can't do it. 0:17:24.520 --> 0:17:27.200 And actually they don't have enough right now, but they're 0:17:27.240 --> 0:17:30.160 going to need even more. I don't even know how 0:17:30.200 --> 0:17:32.440 people are still taking this seriously because on top of that, 0:17:32.840 --> 0:17:36.879 did you hear about the Microsoft negotiations over the nonprofit. 0:17:36.800 --> 0:17:37.520 I've been hearing. 0:17:37.760 --> 0:17:41.199 Well, they're delaying it to next year. They have to. 0:17:41.320 --> 0:17:42.760 They need to convert by the end of the year 0:17:42.760 --> 0:17:45.480 otherwise SoftBank comes their round in half. And everyone's just like, yeah, 0:17:45.480 --> 0:17:48.199 I'll be fine, mate, it'll work it out. What the 0:17:48.240 --> 0:17:50.919 fuck have you have either of you ever? I know 0:17:51.240 --> 0:17:53.400 ed you covered Uber a lot. I don't even think 0:17:53.400 --> 0:17:55.000 the economics match with that either. 0:17:55.200 --> 0:17:58.400 No. I mean, you know, it's interesting because I think 0:17:59.480 --> 0:18:05.119 Uber's strategy, central strategy from the beginning was, you know, 0:18:05.160 --> 0:18:08.040 we have a few existing playbooks that we need to reference, 0:18:08.400 --> 0:18:12.199 you know, the coke deregulation of the taxi industry in 0:18:12.280 --> 0:18:16.960 the nineties, as well as Regula deregulatory campaigns that they led, 0:18:18.440 --> 0:18:22.440 you know, in Seattle and historic campaigns in San Francisco. 0:18:22.960 --> 0:18:26.399 There's a lot that we can reference. And if we 0:18:26.440 --> 0:18:30.480 can figure out a way to bootstrap ourselves onto the 0:18:30.520 --> 0:18:36.400 model and onto those previous histories of deregulation while delaying 0:18:36.680 --> 0:18:39.920 scrutiny long enough our economics to actually, you know, get 0:18:39.960 --> 0:18:41.639 to a profitable place, we'll get there, which is what 0:18:41.680 --> 0:18:44.480 they did right. But even then, I mean, I feel 0:18:44.560 --> 0:18:47.280 like from the beginning, as much as I hated a 0:18:47.320 --> 0:18:50.000 lot of the coverage of Uber for years, the people 0:18:50.000 --> 0:18:53.199 who were always correct about it, We're like the labor 0:18:53.920 --> 0:18:56.200 reporters who had actually, you know, if you spend time 0:18:56.200 --> 0:18:58.480 talking to the drivers, that will lead you to be 0:18:58.520 --> 0:19:00.600 a little bit more interested in the you know, what 0:19:00.760 --> 0:19:04.399 can justify the suffering behind this, and then you almost 0:19:04.440 --> 0:19:08.440 always will see that there's no act. At that point, 0:19:08.480 --> 0:19:10.520 there were no way the unit economics worked unless you 0:19:10.520 --> 0:19:11.320 subsidize everything. 0:19:11.400 --> 0:19:11.560 Yeah. 0:19:11.560 --> 0:19:11.960 I feel like. 0:19:11.960 --> 0:19:14.960 Similarly, there's something going on with with artificial intelligence firms 0:19:14.960 --> 0:19:19.080 and the global AI value chain where if you start 0:19:19.160 --> 0:19:22.080 with a labor analysis and you look at you know, 0:19:22.119 --> 0:19:25.359 invisible workers or ghost workers that are integitical. 0:19:24.920 --> 0:19:27.520 The Kenyan people training the models m M yeah. 0:19:27.440 --> 0:19:30.720 Or labeling or you know, an any any laborer that's 0:19:30.840 --> 0:19:33.639 out of sight, out of mind. You know, then the 0:19:34.080 --> 0:19:36.520 you know, starting there and going up, it becomes hard 0:19:36.560 --> 0:19:40.239 to ask, Okay, how are we supposed to allocate all 0:19:40.280 --> 0:19:44.040