WEBVTT - Radio Better Offline: Gare Davis, Victoria Song, Allison Morrow 0:00:02.800 --> 0:00:03.560 Zone Media. 0:00:05.600 --> 0:00:08.480 This is water extermination, fights cell by cell, through bodies 0:00:08.480 --> 0:00:10.879 and mindscreens of the earth. Soul's rotten from the orgasm, 0:00:10.960 --> 0:00:13.720 drug flesh shuddering from the ovens. Prisoners of the Earth, 0:00:13.760 --> 0:00:16.239 come out storm the studio. This is better offline and 0:00:16.280 --> 0:00:31.080 I'm at Zetron. We have an incredible studio guest assortment. 0:00:31.120 --> 0:00:33.920 Today we have the wonderful Victoria song from the Verge, 0:00:33.880 --> 0:00:37.760 hating Victoria. I'm good geez great, and we of course 0:00:37.760 --> 0:00:42.520 of Alison Morrow of CNN CNN Nightcap Newsletter as well. Yes, wonderful, 0:00:42.720 --> 0:00:46.480 and of course Gare Davis, the wonderful Gear Davis who 0:00:46.479 --> 0:00:48.800 didn't insult me on Bluesky for being late because I 0:00:48.960 --> 0:00:51.519 was on time. Of Cool Zone Media, Gear, thank you 0:00:51.560 --> 0:00:52.279 for joining us. 0:00:52.440 --> 0:00:55.639 Thank you for having me again, despite our despite our 0:00:55.640 --> 0:00:57.960 brief fight on Blue Sky tiff. 0:00:58.160 --> 0:01:00.280 It was not even a tiff. It was a friendly 0:01:00.400 --> 0:01:02.640 thing by some merchandise though, if you're listening to it. 0:01:02.680 --> 0:01:04.319 We have new hoodies, we have new t shirts, we 0:01:04.319 --> 0:01:06.840 have new hats. We have an upcoming challenge coin that 0:01:06.920 --> 0:01:08.640 you can spend your money on and it will flow 0:01:08.720 --> 0:01:11.839 somewhat to me. That's what's great. But today we're talking 0:01:11.840 --> 0:01:14.399 about artificial intelligence. There have been a few stories in 0:01:14.440 --> 0:01:17.040 the media. Allison is freshback from vacation, so she has 0:01:17.080 --> 0:01:18.760 to learn about all the good things that have been happening. 0:01:18.880 --> 0:01:20.800 Don't want to start with one of my favorite stories 0:01:20.800 --> 0:01:23.560 at the moment, and this is the negotiation between Microsoft 0:01:23.640 --> 0:01:27.160 and open Ai. Now, the negotiation is just to run 0:01:27.200 --> 0:01:29.319 this down because open ai, by the end of the 0:01:29.360 --> 0:01:32.000 year needs to become a for profit entity. It's a 0:01:32.000 --> 0:01:34.080 little more complex than that. It's the for profit part 0:01:34.080 --> 0:01:37.160 of a nonprofit needs to convert. That alone would be difficult, 0:01:37.200 --> 0:01:40.480 but Microsoft owns forty nine percent of this company's future 0:01:40.520 --> 0:01:42.320 profits and a bunch of other stuff. They get a 0:01:42.360 --> 0:01:46.120 revenue share, they get right store of their IP through 0:01:46.120 --> 0:01:48.480 twenty thirty and all these other things. And open ai 0:01:48.600 --> 0:01:51.840 has said, Okay, what if we give you thirty three 0:01:51.840 --> 0:01:54.960 percent equity less revenue share and you don't get access 0:01:54.960 --> 0:01:59.720 to ORIP And understandably Microsoft has said no. So we 0:01:59.760 --> 0:02:03.480 are in the funniest possible scenario here in the Microsoft 0:02:03.520 --> 0:02:06.120 could literally just fold their arms and let open ai die. 0:02:06.920 --> 0:02:10.160 And I feel like at the moment, this is an 0:02:10.320 --> 0:02:15.440 underdiscussed topic because this is a gun to Sam Worman's head, 0:02:16.080 --> 0:02:18.040 and everyone's just kind of acting like it's fine. I 0:02:18.040 --> 0:02:21.400 guess maybe it's just two complex and it's confusing to 0:02:21.440 --> 0:02:25.880 me why more people aren't a little bit worried. Anyone 0:02:26.160 --> 0:02:27.440 numbers hard numbers. 0:02:27.480 --> 0:02:29.040 That's scary, that's. 0:02:28.760 --> 0:02:30.800 Scary, that's yeah. 0:02:30.840 --> 0:02:34.320 But I think I think why I'm going so insane 0:02:34.360 --> 0:02:37.760 about it is this could kill open AI one hundred percent. 0:02:37.840 --> 0:02:40.320 Like this is they don't turn into a for profit, 0:02:40.480 --> 0:02:43.799 they're dead dead, And I'm just wondering why everyone's just 0:02:43.880 --> 0:02:47.160 kind of chilling walking around about it. I feel crazy. 0:02:47.639 --> 0:02:51.800 I think it's just the sense that like that seems 0:02:51.840 --> 0:02:54.520 implausible to most people, like if you look on it, 0:02:54.919 --> 0:02:57.720 like maybe, but you know, chat GBT is synonymous with 0:02:57.760 --> 0:03:01.280 AI in the sense that kleanex is synonymous with tissues 0:03:01.360 --> 0:03:03.560 right now, So you know, we're at a point in 0:03:03.600 --> 0:03:07.960 time where when you see the big one of of 0:03:08.120 --> 0:03:11.359 like a tech thing, you kind of feel that they're infallible. 0:03:11.400 --> 0:03:14.160 It's sort of like saying, well, if Apple doesn't get 0:03:14.160 --> 0:03:16.640 its ducks in a row with these tariffs, they're fucked, 0:03:17.000 --> 0:03:19.880 in which case you're like, are they are they? It's 0:03:19.960 --> 0:03:22.200 like a Marvel movie at the end, Are they dead. 0:03:22.320 --> 0:03:24.280 Are they are they really dead? Or will they come 0:03:24.320 --> 0:03:28.480 back in some mutated form in Avengers like Part seventy two, 0:03:28.880 --> 0:03:30.639 The avengein ing right. 0:03:31.000 --> 0:03:34.600 And I think everyone's still under the spell of Sam Altman, right, 0:03:34.880 --> 0:03:38.320 there's a sense that he's the visionary who's going to 0:03:38.440 --> 0:03:42.800 lead us into this AI utopia. And you know, I 0:03:43.000 --> 0:03:45.240 at others, you know, a bunch of us have reported 0:03:45.280 --> 0:03:47.840 that a lot of it is smoke and mirrors. But 0:03:48.640 --> 0:03:53.280 I think investors and shareholders and people who like frankly, 0:03:53.760 --> 0:03:56.640 the people who work for him, desperately want to see succeed. 0:03:56.720 --> 0:04:00.400 I think Silicon Valley wants him to succeed. So maybe 0:04:00.400 --> 0:04:02.320 it's just a willful blindness. 0:04:02.400 --> 0:04:05.480 Yeah, it's the most strange time in history because in 0:04:06.040 --> 0:04:08.480 Victoria you've written quite a lot about this. When you 0:04:08.520 --> 0:04:11.280 look at the actual things that this shit does, it 0:04:11.360 --> 0:04:14.040 don't do that much. Right now, you've been on the 0:04:14.080 --> 0:04:17.000 lead a on the verge for a few months now, 0:04:17.240 --> 0:04:19.400 have you seen anything exciting at all? 0:04:20.640 --> 0:04:22.640 Define exciting, Anything. 0:04:22.240 --> 0:04:24.240 That you looked at and you felt delighted by in 0:04:24.320 --> 0:04:27.760 any way, Because I'm genuinely curious by delighted. 0:04:27.960 --> 0:04:29.920 I think delighted is a strong word. 0:04:30.279 --> 0:04:33.120 Have I seen things that have been surprising. Yeah, with 0:04:33.240 --> 0:04:35.120 some of the Like I wrote a story not that 0:04:35.160 --> 0:04:37.800 long ago about the what I call the hug and 0:04:37.920 --> 0:04:39.120 Kiss generators. 0:04:39.600 --> 0:04:40.960 Uh yeah, exactly. 0:04:41.920 --> 0:04:46.840 So there are these apps there are called Hugging and 0:04:47.000 --> 0:04:51.240 Kiss AI generators. So you take a picture and like 0:04:51.279 --> 0:04:54.120 they were advertised in a kind of skivy way where 0:04:54.160 --> 0:04:56.400 it's just like, oh, you take a picture of you 0:04:56.520 --> 0:04:59.200 and your crush and make them kiss, and like that's 0:04:59.360 --> 0:05:02.919 good that you can do. AI doesn't understand what to 0:05:02.960 --> 0:05:06.760 do with tongues yet, so you know, I was generating 0:05:06.920 --> 0:05:09.200 very cursed content for the Verge dot com. 0:05:09.240 --> 0:05:11.000 That was what those horrible things you're sending me? 0:05:11.120 --> 0:05:13.760 Right, yeah, yeah, yeah, I sent you some horrible things. 0:05:13.520 --> 0:05:14.680 Really awful. 0:05:15.000 --> 0:05:17.200 It turns to the AI bots. A lot of humans 0:05:17.200 --> 0:05:18.239 don't know what to do with tongue. 0:05:18.520 --> 0:05:20.760 But no, they but they really don't know what to 0:05:20.800 --> 0:05:22.920 do with the songs, like you know, you're supposed to 0:05:22.920 --> 0:05:25.320 make people kissing, and so like, I generated a couple 0:05:25.320 --> 0:05:28.320 of videos of me and Edward Collin because not because 0:05:28.360 --> 0:05:30.640 I'm like a twyhard but because he was like a 0:05:30.680 --> 0:05:34.480 preset in the app, right, and you know, you just 0:05:34.560 --> 0:05:37.800 you just watch yourself kiss Edward Cullen with full on 0:05:37.880 --> 0:05:38.680 tongue and you're just. 0:05:38.600 --> 0:05:40.880 Like, but little wrong tongue. 0:05:40.960 --> 0:05:44.279 It's wrong tongue because honestly, it's it's like if you 0:05:44.360 --> 0:05:46.960 told a toddler what kissing looks like and they just 0:05:47.160 --> 0:05:52.120 you know, imagined two faces smooshing together and like things 0:05:52.240 --> 0:05:54.919 coming out of the mouths and odd rhythms. That's what 0:05:54.960 --> 0:05:55.400 it looked like. 0:05:55.440 --> 0:05:57.200 I mean, I could be wrong here, but I assume 0:05:57.240 --> 0:05:59.279 most of these products are made by straight man who 0:05:59.480 --> 0:06:02.119 do not know what to do with the tongue. 0:06:02.240 --> 0:06:05.359 These apps are just very like, yeah, weird, but you 0:06:05.360 --> 0:06:07.839 know so I tested them. I deep faked my parents 0:06:09.160 --> 0:06:10.880 parents at my wedding, and I was like, oh, this 0:06:11.000 --> 0:06:16.960 makes me feel weird. I have emotions that mom's teeth 0:06:17.000 --> 0:06:20.360 are not correct in this story. But you know so 0:06:20.520 --> 0:06:22.560 that was just like one of the I think one 0:06:22.560 --> 0:06:25.960 of the things that I've tested most recently and I went, oh, okay. 0:06:25.760 --> 0:06:27.240 This is something. 0:06:27.560 --> 0:06:29.400 I mean, the thing that I've seen this with is 0:06:29.440 --> 0:06:32.320 like this this woman who made like a video of 0:06:32.880 --> 0:06:35.640 like her mom hugging her as a kid, based on 0:06:35.720 --> 0:06:37.880 like an old picture, right, and it's like this is 0:06:38.320 --> 0:06:40.960 I've never seen a video of my mom before, and 0:06:41.000 --> 0:06:44.599 it's you start obsessing over this, this like artificially generated video. 0:06:44.960 --> 0:06:47.200 When you're ignoring you you actually have a picture of 0:06:47.240 --> 0:06:49.840 your mom hugging yeah, Like you can look at like 0:06:49.880 --> 0:06:53.280 that and that actually is her, that is like what 0:06:53.360 --> 0:06:54.120 she looks. 0:06:53.920 --> 0:06:58.560 Like, yeah, versus watching something imagine, not imagine just generate. 0:06:58.400 --> 0:07:00.839 Put on like a skin suit of your mom hugging you, 0:07:00.920 --> 0:07:03.240 which it just isn't like it's weird. The video is 0:07:03.279 --> 0:07:04.640 not real, but the picture is. 0:07:04.960 --> 0:07:07.760 I just it's a bizarre I feel those I felt 0:07:07.839 --> 0:07:10.120 very judgmental of it before I tried it, and then 0:07:10.160 --> 0:07:11.200 I tried it and I sobbed. 0:07:11.240 --> 0:07:13.960 I like genuinely sobbed because that's interesting. 0:07:14.240 --> 0:07:18.320 That reminds me of like the the VR thing where 0:07:18.400 --> 0:07:23.000 you're like reconnecting with like dead family members only VR headsets, 0:07:23.000 --> 0:07:25.040 which which yeah, people were very skeptical of. And then 0:07:25.080 --> 0:07:28.040 I saw saw some people try it in Japan and 0:07:28.080 --> 0:07:29.960 they like like just totally broke down. 0:07:30.040 --> 0:07:32.960 It's just like when I like, I think the phrasing 0:07:33.000 --> 0:07:35.080 that I used was like, I know it's fake. It 0:07:35.120 --> 0:07:37.640 didn't look anything like my dad gave him hair. My 0:07:37.720 --> 0:07:40.040 dad had never had hair in his life, but like 0:07:40.240 --> 0:07:44.400 the shape of it was enough to like sketch this 0:07:44.560 --> 0:07:46.520 like part of me that was very much longing for 0:07:46.600 --> 0:07:48.560 my father to have been able to go to my wedding. 0:07:49.120 --> 0:07:51.440 So like doing it, I was like, this is weird. 0:07:51.480 --> 0:07:53.760 I don't feel this is not comforting. I mean it's 0:07:53.760 --> 0:07:55.280 not comforting, but it's. 0:07:55.360 --> 0:07:56.360 Makes you feel bad. 0:07:56.600 --> 0:08:00.800 That nausea is like like indicative of like the hyperreality problem. Yeah, 0:08:00.840 --> 0:08:03.400 which which mean people like like Altman in the whole 0:08:03.440 --> 0:08:05.560 the whole industry or can you break them rapidly pushing 0:08:05.640 --> 0:08:06.280 us towards. 0:08:06.080 --> 0:08:07.760 On the hyperreality problem? Can you break that down? 0:08:07.920 --> 0:08:09.560 Well, I mean, I guess if turn gets used in 0:08:09.560 --> 0:08:12.160 a few different ways, but it's like the more something 0:08:12.200 --> 0:08:14.760 is so fake that it's more real than real hmm. 0:08:15.680 --> 0:08:17.920 And we see this problem with a lot of like 0:08:18.480 --> 0:08:21.600 the VR stuff, But now now you see this a 0:08:21.600 --> 0:08:25.640 lot with with a AI generated images which are like 0:08:25.720 --> 0:08:30.200 quote unquote photorealistic, but they're like two photorealistic because they're 0:08:30.200 --> 0:08:32.000 being trained on on on a data set of like 0:08:32.040 --> 0:08:34.800 highly photoshoped images, So it looks like reality, but it 0:08:34.800 --> 0:08:38.559 looks more than reality. It's it's it's it's stronger than 0:08:38.559 --> 0:08:41.480 what reality actually is. Supposed to be and that's like 0:08:41.679 --> 0:08:44.720 completely poisoning the data set and like this can affect 0:08:44.760 --> 0:08:45.880 you like emotionally too. 0:08:46.000 --> 0:08:48.679 Yeah, and what But the thing is that gets me 0:08:48.720 --> 0:08:51.480 about this is you look at everything. You look at 0:08:51.520 --> 0:08:54.560 all the AI stuff and this is a fairly old 0:08:54.600 --> 0:08:57.000 example at this point. And I'm not I'm not insulting 0:08:57.000 --> 0:08:59.400 you would. This is nothing bad about your point. It's 0:08:59.440 --> 0:09:02.199 just they've not been able to find a dude at 0:09:02.280 --> 0:09:05.600 or a gizmo that at least brings you cheer. 0:09:07.800 --> 0:09:09.840 I mean you can if you are twisted, then my 0:09:09.920 --> 0:09:13.040 mind is twisted. And you'll like come up with some 0:09:13.120 --> 0:09:16.280 prompts that are truly cursed and no editor will let 0:09:16.360 --> 0:09:19.120 you publish in good faith, Like yeah, you could have 0:09:19.240 --> 0:09:20.199 a little fun with it. 0:09:20.320 --> 0:09:22.679 I'm talking about a thing that people used to do 0:09:22.800 --> 0:09:23.600 something normal. 0:09:23.960 --> 0:09:27.720 No, yeah, no one, it's I mean it helped that 0:09:27.760 --> 0:09:29.320 one kid graduate from UCLA. 0:09:29.480 --> 0:09:30.160 So there you go. 0:09:30.440 --> 0:09:31.240 Which one was that? 0:09:31.640 --> 0:09:32.760 Oh, this has just been. 0:09:32.600 --> 0:09:35.360 A viral video the past like two weeks of this 0:09:35.400 --> 0:09:38.280 guy like showing the prompts he used to graduate from college, 0:09:38.360 --> 0:09:40.320 like during his graduation ceremony. 0:09:40.440 --> 0:09:41.000 So cool. 0:09:41.440 --> 0:09:44.000 Yeah, that's the thing I've pointed out a lot of that. 0:09:44.200 --> 0:09:46.640 AI has found a lot of use cases. They're just 0:09:46.920 --> 0:09:48.400 all kind of bad. 0:09:48.480 --> 0:09:50.880 But they don't generate money, right. 0:09:50.800 --> 0:09:55.920 They don't generate money, and they take away This will 0:09:55.960 --> 0:09:58.080 be a metaphor that you relate to. I can't remember 0:09:58.080 --> 0:09:59.440 who said it first. 0:10:00.320 --> 0:10:01.040 It was a tech. 0:10:00.840 --> 0:10:05.560 Columnist anyway, it was about letting kids use AI to 0:10:05.600 --> 0:10:08.120 do their homework is like going to the gym and 0:10:08.160 --> 0:10:11.439 having a machine lift the weights for you. Like, the 0:10:11.600 --> 0:10:15.360 point is the process and the learning, and AI just 0:10:15.440 --> 0:10:18.520 kind of subverts that, which can be useful if you're 0:10:19.040 --> 0:10:23.080 coding or doing some high level technical stuff. I suppose yeah, 0:10:23.240 --> 0:10:23.920 not my area. 0:10:24.080 --> 0:10:27.080 But but even then, it's like with the coding stuff, 0:10:27.080 --> 0:10:30.040 they massively overstate what coding like, coding is only one 0:10:30.080 --> 0:10:33.440 part of the software engineering stack. And even then you 0:10:33.480 --> 0:10:35.280 can't like help They're like, oh, I could build an 0:10:35.360 --> 0:10:38.680 entire application? Could it has anyone? Actually, And this is 0:10:38.679 --> 0:10:40.440 Matt Hughes Murtis that brought this up the other day. 0:10:40.640 --> 0:10:43.439 How did it all this bullshit Kevin Rucy and bullshit 0:10:43.480 --> 0:10:45.960 about oh, vibe coding is taking up. I've got seen 0:10:46.000 --> 0:10:47.760 one fucking vibe coded company. 0:10:48.080 --> 0:10:51.360 Man. The phrase vibe code, it's just I haven't heard 0:10:51.400 --> 0:10:51.880 this before. 0:10:51.960 --> 0:10:53.480 Vibe coding is it's horrible. 0:10:53.720 --> 0:10:56.840 It's it's it's it's horrible. Like at Google Io they're 0:10:56.880 --> 0:10:59.520 like vibe coating and I was like, please kill me, 0:11:00.400 --> 0:11:02.440 because what does it even mean. 0:11:02.640 --> 0:11:05.840 It's so, here's what it's meant to mean. It's meant 0:11:05.840 --> 0:11:08.440 to mean that you, as a person do not understand software. 0:11:09.040 --> 0:11:09.240 True. 0:11:09.440 --> 0:11:11.880 True, you are able to use the coding thing to 0:11:11.920 --> 0:11:15.559 build software. And the idea which the massive liberty they 0:11:15.559 --> 0:11:17.880 take from there is that because someone can do a 0:11:18.040 --> 0:11:20.880 thing like this, whether it works or not, whether it's 0:11:20.920 --> 0:11:24.960 secure or not, who cares, it means that someone who 0:11:24.960 --> 0:11:27.560 doesn't understand coding at all could build a huge company 0:11:27.559 --> 0:11:30.440 that does whatever. It's kind of like if you saw 0:11:30.600 --> 0:11:34.280 the watch The Simpsons anyone. Yeah, there's an episode of 0:11:34.320 --> 0:11:37.880 The Simpsons where they rebuild Ned Flanders his house and 0:11:37.920 --> 0:11:39.679 like the rooms get smaller and they're like, this room 0:11:39.720 --> 0:11:42.200 has no electricity, this room has too much electricity, and 0:11:42.240 --> 0:11:44.320 all the hair stands up. It's like seeing that and 0:11:44.360 --> 0:11:46.440 being like, holy shit, these people could build a city. 0:11:46.800 --> 0:11:49.600 It's fucking insane. And of course there is a Kevin 0:11:49.640 --> 0:11:51.840 Ruth's article in New York Times, so he's like, that's arm. 0:11:51.920 --> 0:11:54.840 Oh my god, I made a recipe application. It's just 0:11:55.840 --> 0:11:59.640 real like peekaboo moments in AI. And I know I've 0:11:59.679 --> 0:12:02.199 been thinking about AI for what feels like seven years now, 0:12:02.240 --> 0:12:06.360 but it's only one. But it's just I don't know 0:12:06.400 --> 0:12:09.480 why more people again, I get the Microsoft Open Air, 0:12:09.520 --> 0:12:11.040 I think I do, but I don't get why more 0:12:11.080 --> 0:12:13.960 people aren't more alarm. There's nothing. There's not a thing. 0:12:14.080 --> 0:12:15.880 There's not a thing that you can point and be like, Wow, 0:12:15.880 --> 0:12:17.520 this is actually kind of fucking cool. That costs too 0:12:17.600 --> 0:12:20.360 much money, it requires stealing, it boils legs with all 0:12:20.400 --> 0:12:22.840 these things. But at least we have this. Not really, 0:12:23.880 --> 0:12:27.840 I genuinely, I'm not even asking the question sarcastically anymore. 0:12:27.840 --> 0:12:32.880 I'm just like anything, anything, anything, one thing that I 0:12:32.880 --> 0:12:35.520 don't mean kind of works. I mean, this is a 0:12:35.559 --> 0:12:38.240 tool that I use every day, like a drop box 0:12:38.320 --> 0:12:40.120 style thing. I don't even mean a filse. I just 0:12:40.120 --> 0:12:42.560 mean a useful piece of software you can point to 0:12:42.600 --> 0:12:44.400 and go, I use this, and it's good. 0:12:45.679 --> 0:12:47.440 I have I have one, but I don't know how 0:12:47.520 --> 0:12:50.400 much of it is LM based. Do you use otter 0:12:50.640 --> 0:12:51.200 in your way? 0:12:51.320 --> 0:12:51.600 I do? 0:12:51.760 --> 0:12:54.400 Yeah, Like Otter is a dictation service. 0:12:54.480 --> 0:12:55.599 Yea transcription. 0:12:55.720 --> 0:12:57.360 I was just gonna say transcription. 0:12:57.480 --> 0:12:59.000 I was just gonna say clips captioning. 0:12:59.040 --> 0:13:02.840 That's like the only thing like Dropbox has integrated like 0:13:02.960 --> 0:13:06.400 of auto transcriptions for almost all their uploads and then 0:13:06.440 --> 0:13:08.600 mix My job really easy because I have to do 0:13:08.640 --> 0:13:10.400 a lot of interviews and now I can just refer 0:13:10.440 --> 0:13:12.240 to that. If I need a more complex transcription, I 0:13:12.280 --> 0:13:14.280 could set it to one of our services. But but no, 0:13:14.400 --> 0:13:17.560 like that's that's it, But like that's LLL Hulm's have 0:13:17.559 --> 0:13:18.959 been doing that for a long time. 0:13:19.280 --> 0:13:23.080 Yeah, it's just transcription. I'm not even being a misanthrope. 0:13:23.160 --> 0:13:26.640 I'm just so much money is going into this, so 0:13:26.800 --> 0:13:29.560 much money is not going into other things. And the 0:13:29.600 --> 0:13:31.880 only thing we're having popped out is, hey, we've got 0:13:31.920 --> 0:13:35.199 a Google search that kind of works but doesn't and 0:13:35.240 --> 0:13:37.719 we can do transcriptions, which they did almost immediately. I 0:13:37.760 --> 0:13:40.200 feel I feel like that like rev had their AI 0:13:40.200 --> 0:13:42.600 transcriptions almost immediately. Yeah, they've been around for it at 0:13:43.080 --> 0:13:46.400 There were various companies I worked with an AI transcription 0:13:46.440 --> 0:13:49.320 coming dead now A couple of years back with the 0:13:49.600 --> 0:13:51.280 like twenty twenty three, I think it was. It was 0:13:51.320 --> 0:13:53.320 like meeting transcription zooms had it. 0:13:53.600 --> 0:13:55.400 I was I was using where I was AI transcriptions 0:13:55.400 --> 0:13:56.160 back in twenty twenty. 0:13:56.240 --> 0:14:00.280 Yeah, it's just like it feels it feels like I'm 0:14:00.280 --> 0:14:02.440 going insane sometimes. It feels like when I read these 0:14:02.440 --> 0:14:05.080 stories and they're like in the revolutionary power of Ai, 0:14:05.600 --> 0:14:07.640 but you look at it, it's like you don't even 0:14:07.679 --> 0:14:09.800 have a funny You have a funny thing. I guess 0:14:10.120 --> 0:14:11.880 a joker level thing. Yeah. 0:14:12.040 --> 0:14:15.280 I think the problem is is just like we are 0:14:15.360 --> 0:14:20.360 promised one thing. Yeah, you're promised this personalization, this automat, 0:14:20.680 --> 0:14:23.320 this automation that it's gonna know you. And like we've 0:14:23.360 --> 0:14:27.280 been fed through so many generations of like science fiction 0:14:27.440 --> 0:14:30.000 what we think uh ai is going to be. This 0:14:30.120 --> 0:14:33.440 is not that this requires so much work from you 0:14:33.520 --> 0:14:36.360 to train it. Like you have to understand the language 0:14:36.360 --> 0:14:39.320 which with to prompt, chat cept or any of these 0:14:39.360 --> 0:14:42.440 other ais in order to get something remotely useful. So 0:14:42.480 --> 0:14:45.400 you're actually having to learn a new language. And like 0:14:45.480 --> 0:14:48.600 if you look at the most successful chat ChiPT prompts, 0:14:48.960 --> 0:14:52.400 they're like four four paragraphs along, they're insane. You have 0:14:52.480 --> 0:14:56.880 to like you have to be preempting what like this 0:14:56.960 --> 0:14:59.240 thing could be like I had. I was just like, 0:14:59.320 --> 0:15:02.200 you know, to your point about vibe coding, I'm not 0:15:02.280 --> 0:15:04.320 a I'm like a spreadsheet girly, but I'm not like 0:15:04.360 --> 0:15:06.760 an advanced spreadsheet early and I was trying to pull 0:15:06.800 --> 0:15:10.120 some data insights from this set I was looking at, 0:15:10.120 --> 0:15:13.360 and I was like, Okay, I don't know fuck all 0:15:13.520 --> 0:15:17.720 about spreadsheet formulas. Besides like the really basic ones. How 0:15:17.720 --> 0:15:20.040 am I going to do this conditional logic program? Let 0:15:20.120 --> 0:15:24.440 me ask chat CHETPT And it took me so long 0:15:25.000 --> 0:15:27.680 just to figure out the stuff, and it was always wrong. 0:15:28.120 --> 0:15:30.960 And because I could, like, because I understand math, I 0:15:31.000 --> 0:15:34.960 could parse out how to fix the completely wrong formulas 0:15:34.960 --> 0:15:35.680 it was giving me. 0:15:36.320 --> 0:15:39.960 But that was such a painful process, and. 0:15:39.920 --> 0:15:41.960 So was the output even that good? 0:15:42.360 --> 0:15:44.480 Oh no, it was an excellent output. I got a 0:15:44.480 --> 0:15:45.440 beautiful spreadsheet. 0:15:45.440 --> 0:15:47.480 That's cool. I had much time. Would you say you 0:15:47.560 --> 0:15:50.400 invested two and a half hours? Nice? 0:15:51.440 --> 0:15:54.640 You're you're a journalist who understands math, like you're a 0:15:54.760 --> 0:15:56.200 rare and special Yeah, you know. 0:15:56.320 --> 0:15:56.520 It was. 0:15:56.880 --> 0:15:59.400 I was just basically like this, the fact that this 0:15:59.440 --> 0:16:02.440 conditional if and statement is not working, or it's working 0:16:02.440 --> 0:16:04.800 opposite to what I want for to find this one 0:16:04.840 --> 0:16:09.640 particular data set is driving me cuckoo for cocoa puffs. 0:16:10.360 --> 0:16:13.680 It's very simple. I know how to do the math manually. 0:16:13.840 --> 0:16:16.040 Why can't the computer tell me how to write it 0:16:16.080 --> 0:16:19.840 to the other fucking computers two and a half hours. 0:16:19.880 --> 0:16:22.360 I love innovation, and I think I think that that 0:16:22.400 --> 0:16:25.560 speaks to the larger problem, which is generative. AI isn't 0:16:25.600 --> 0:16:28.160 completely useless. If it had been sold as it is, 0:16:28.200 --> 0:16:33.280 which is kind of niche cloud software, like cloud compute stuff, 0:16:33.640 --> 0:16:35.080 they wouldn't have been able to fund any of the 0:16:35.160 --> 0:16:36.880 data centers. If they would have been like, all right, 0:16:36.920 --> 0:16:38.120 we're going to be able to in two and a 0:16:38.120 --> 0:16:40.840 half hours give you the world's best spreadsheet, they would 0:16:41.200 --> 0:16:41.440 it was. 0:16:41.600 --> 0:16:44.680 It was a good spread okay, lot the best, Okay, 0:16:45.520 --> 0:16:46.520 a decent spreadsheet. 0:16:46.600 --> 0:16:50.440 The thing even the asterisks have asterisks, and it's it's 0:16:50.520 --> 0:16:53.800 just I feel like in the feedback I get from 0:16:53.880 --> 0:16:56.720 listeners is very much that everyone is like a lot. 0:16:57.280 --> 0:16:59.680 I don't get any emails from people being like, hey, 0:16:59.680 --> 0:17:01.960 atu man, I have the most useful thing I get 0:17:02.000 --> 0:17:04.480 the occasional bright spark is like I have, over the 0:17:04.480 --> 0:17:06.840 course of hours, created a very useful thing that I 0:17:06.960 --> 0:17:15.119 use Sometimes it's like cool, Okay, it's fine. Bidet sounds 0:17:15.119 --> 0:17:17.359 more useful than that, Like I'm trying to think of 0:17:17.440 --> 0:17:20.520 like other innovations that exist that could be Yeah, everything 0:17:20.560 --> 0:17:23.040 is more useful, like Apple pay is more useful than 0:17:23.080 --> 0:17:25.720 any of the shit that they've built. But it's this 0:17:25.800 --> 0:17:28.080 thing where we are being told again and again and 0:17:28.119 --> 0:17:31.240 again that it's the future, and we're being told that 0:17:31.280 --> 0:17:34.080 it's this ultracomplex thing that will never understand, which leads 0:17:34.119 --> 0:17:36.639 really neatly into my favorite story of the week, which 0:17:36.720 --> 0:17:39.439 is all of you, I assume have heard about this 0:17:40.200 --> 0:17:43.280 Meta offering one hundred million dollars to Open AI staff 0:17:43.680 --> 0:17:46.399 and how there's this big talent more and four people 0:17:46.480 --> 0:17:49.720 just left Open AI to go to Meta. Alison, you're 0:17:49.760 --> 0:17:52.000 on vacation, so you missed some of this, which is 0:17:52.000 --> 0:17:53.399 probably best for your mental health. 0:17:53.680 --> 0:17:57.959 Oh I saw a bit. The seven figure bonus story 0:17:58.080 --> 0:18:00.359 came out right before I went on vacations, so I 0:18:00.440 --> 0:18:01.199 was getting that. 0:18:02.000 --> 0:18:05.400 I can't blame Oh no, get way to a four 0:18:05.480 --> 0:18:09.520 oh one k that's fat and sizable, Like yeah, no, no. 0:18:09.520 --> 0:18:12.800 I think it rocks. I think it's great. I think 0:18:12.840 --> 0:18:15.040 they should demand There was a part of the erin 0:18:15.080 --> 0:18:16.879 Woo from the information of a great story about this, 0:18:17.480 --> 0:18:20.199 where she she was talking about how some people are like, 0:18:20.240 --> 0:18:22.960 oh yeah, I just threatened to quit and they gave 0:18:23.000 --> 0:18:25.160 me more money. I would join one of these companies 0:18:25.160 --> 0:18:28.240 and day two slug fucking quit. Mark, What are you 0:18:28.320 --> 0:18:30.160 gonna do about it? Mark, I'm gonna go right back 0:18:30.160 --> 0:18:32.480 to Open Ai. They're gonna give me this money and 0:18:32.520 --> 0:18:34.760 then Mark Zuckerbo will give you whatever you He's been 0:18:34.760 --> 0:18:37.160 flying them to his house in Tahoe and they're still 0:18:37.160 --> 0:18:40.400 saying no. That's the best part that people just like, nah, 0:18:40.400 --> 0:18:42.720 I don't want to, don't want to mate, But it 0:18:42.840 --> 0:18:45.280 might be because they're all in a weird I'm gonna 0:18:45.280 --> 0:18:48.600