1 00:00:01,760 --> 00:00:17,480 Speaker 1: Wuz Media. Hello, I'm ed Zitron, and this is Better Offline. 2 00:00:19,360 --> 00:00:21,480 Speaker 1: We finally reached the end of our three part how 3 00:00:21,520 --> 00:00:24,640 Speaker 1: to Argue with an AI Booster series. Big strong men 4 00:00:24,680 --> 00:00:29,120 Speaker 1: are standing outside of the Better Offline studio, suspended two 5 00:00:29,240 --> 00:00:31,880 Speaker 1: hundred feet above the Las Vegas Strip, with tears in 6 00:00:31,920 --> 00:00:34,240 Speaker 1: their eyes, and they're saying, sir, sir, it's the most 7 00:00:34,240 --> 00:00:37,479 Speaker 1: beautiful podcast I've ever heard. Please stop recording it. Everyone 8 00:00:37,560 --> 00:00:40,280 Speaker 1: else will feel insufficient when they hear it. No, No, 9 00:00:40,479 --> 00:00:43,360 Speaker 1: I have to continue. I'm afraid my listeners need me. 10 00:00:44,440 --> 00:00:47,720 Speaker 1: But okay, seriously, folks, if there's anything I want you 11 00:00:47,760 --> 00:00:50,280 Speaker 1: to take home so far, is that the arguments that 12 00:00:50,360 --> 00:00:54,320 Speaker 1: these people, these AI sicker fans make, they crumble under 13 00:00:54,320 --> 00:00:57,120 Speaker 1: the slightest bit of scrutiny. And yet these arguments work 14 00:00:57,200 --> 00:00:59,440 Speaker 1: because they either exploit the lack of knowledge of those 15 00:00:59,480 --> 00:01:01,920 Speaker 1: of us who do and understand the rotten economics of AI, 16 00:01:02,600 --> 00:01:04,920 Speaker 1: or because they forced the opposite party to surrender their 17 00:01:04,959 --> 00:01:08,080 Speaker 1: reason and to exit the planes of reality, to which 18 00:01:08,120 --> 00:01:10,720 Speaker 1: I say, fuck that, absolutely not. This is a hell 19 00:01:10,720 --> 00:01:12,760 Speaker 1: I'm prepared to die on, and I'll hope that you'll 20 00:01:12,760 --> 00:01:15,399 Speaker 1: all be standing with me, side by side Now, in 21 00:01:15,440 --> 00:01:17,360 Speaker 1: the first episode, we shook away the claims that it's 22 00:01:17,400 --> 00:01:19,160 Speaker 1: the early days for AI and we just need to 23 00:01:19,160 --> 00:01:21,480 Speaker 1: give it more time. Then I took the idea that 24 00:01:21,560 --> 00:01:24,480 Speaker 1: generative AI is like Uber or fiber optic networking to 25 00:01:24,600 --> 00:01:26,240 Speaker 1: industries that both burned a lot of money at the 26 00:01:26,240 --> 00:01:29,920 Speaker 1: start but are otherwise nothing like generative AI. And now 27 00:01:29,959 --> 00:01:32,160 Speaker 1: it's time to deal with the dregs of the arguments. 28 00:01:32,240 --> 00:01:33,960 Speaker 1: These are the worst of the worst. Boost to quips. 29 00:01:34,240 --> 00:01:38,880 Speaker 1: Here we go, ultra booster quip. I thought about recording 30 00:01:38,880 --> 00:01:40,880 Speaker 1: that with a bunch of reever, but it didn't really 31 00:01:40,920 --> 00:01:42,640 Speaker 1: work out for me, but I wanted to do it once. 32 00:01:42,680 --> 00:01:45,840 Speaker 1: But anyway, their argument is hum AI is just like 33 00:01:45,880 --> 00:01:48,640 Speaker 1: Amazon Web Services, a massive investment that took a while 34 00:01:48,640 --> 00:01:52,000 Speaker 1: to go profitable, and everybody hated Amazon for it. Now 35 00:01:52,000 --> 00:01:54,280 Speaker 1: I actually covered this in depth and the Hater's Guide 36 00:01:54,280 --> 00:01:56,160 Speaker 1: to the AI Bubble, But the long and short of 37 00:01:56,200 --> 00:01:58,640 Speaker 1: it is that Amazon Web Services is a platform and 38 00:01:58,720 --> 00:02:02,080 Speaker 1: necessity with an obvious choice and has burned about ten 39 00:02:02,120 --> 00:02:04,560 Speaker 1: percent of what Amazon and all of them have burned 40 00:02:04,600 --> 00:02:08,960 Speaker 1: chasing generative AI, and had also proven demand before building it. Also, 41 00:02:09,120 --> 00:02:12,160 Speaker 1: Amazon web Services was break even within three years. An 42 00:02:12,200 --> 00:02:14,600 Speaker 1: open aye was founded in fucking twenty fifteen, and even 43 00:02:14,600 --> 00:02:17,320 Speaker 1: if you start from November twenty twenty two, by Amazon 44 00:02:17,360 --> 00:02:20,360 Speaker 1: web Services, standards should be break even by now. But 45 00:02:20,440 --> 00:02:23,720 Speaker 1: now i'll quote myself, Amazon, weent no way. Sorry, that's 46 00:02:23,720 --> 00:02:27,200 Speaker 1: the boosters. Amazon web Services was created out of necessity. 47 00:02:27,360 --> 00:02:30,239 Speaker 1: Amazon's infrastructure needs were so great that he had effectively 48 00:02:30,320 --> 00:02:33,680 Speaker 1: had to build both the software and hardware necessary to 49 00:02:33,760 --> 00:02:37,840 Speaker 1: deliver a store that sold theoretically everything, the theoretically everywhere, 50 00:02:38,639 --> 00:02:41,440 Speaker 1: handling both the traffic from customers, delivering the software that 51 00:02:41,520 --> 00:02:45,160 Speaker 1: runs Amazon dot Com quickly and reliably, and well making 52 00:02:45,200 --> 00:02:47,640 Speaker 1: sure things were stable. It didn't need to come up 53 00:02:47,639 --> 00:02:50,320 Speaker 1: with a reason for people to run web applications. They 54 00:02:50,320 --> 00:02:52,760 Speaker 1: were already doing so themselves, but in ways that cost 55 00:02:52,800 --> 00:02:56,079 Speaker 1: a lot more, were inflexible and required specialist server skills. 56 00:02:57,200 --> 00:03:00,400 Speaker 1: Amazon web Services took something that people already did and 57 00:03:00,440 --> 00:03:02,600 Speaker 1: what there was already a proven demand for, and made 58 00:03:02,639 --> 00:03:05,200 Speaker 1: it better and scaled it. Eventually Google and Microsoft would 59 00:03:05,240 --> 00:03:07,560 Speaker 1: join the frame. I editorialized a bit there, but I 60 00:03:07,560 --> 00:03:10,919 Speaker 1: can do that with my own work now. A common 61 00:03:10,960 --> 00:03:12,960 Speaker 1: booster quip, by the way, is for them to say, well, 62 00:03:13,120 --> 00:03:16,519 Speaker 1: this AI company they've got high annualized revenues. And as 63 00:03:16,560 --> 00:03:19,680 Speaker 1: I've discussed in the past, this metric is basically month 64 00:03:19,720 --> 00:03:22,960 Speaker 1: times twelve. And while it's a fine measure for normal 65 00:03:23,160 --> 00:03:26,280 Speaker 1: high gross margin businesses like software as a service companies, 66 00:03:26,320 --> 00:03:29,040 Speaker 1: it isn't for AI. It doesn't account for churn, which 67 00:03:29,080 --> 00:03:31,600 Speaker 1: is when people leave. It also is a number intentionally 68 00:03:31,720 --> 00:03:34,280 Speaker 1: used to make a company sound more successful. So you 69 00:03:34,280 --> 00:03:37,160 Speaker 1: can say two hundred million dollars annualized revenue instead of 70 00:03:37,160 --> 00:03:39,760 Speaker 1: sixteen point six million dollars a month. And you're also 71 00:03:39,880 --> 00:03:43,360 Speaker 1: meant to budship. You heard they said two hundred million annualized, 72 00:03:43,400 --> 00:03:46,000 Speaker 1: but you heard two hundred million. Your mind did that 73 00:03:46,080 --> 00:03:49,200 Speaker 1: was a bad planket reference. But I'll continue. They want 74 00:03:49,280 --> 00:03:52,000 Speaker 1: you to think two hundred million. They want you to 75 00:03:52,000 --> 00:03:54,560 Speaker 1: think that's what they'll make. More often than not, if 76 00:03:54,560 --> 00:03:56,800 Speaker 1: they mentioned an annualized number, that number will not be 77 00:03:56,800 --> 00:04:00,760 Speaker 1: how much they make that year. Also, if they're using 78 00:04:00,760 --> 00:04:04,200 Speaker 1: this number, it's likely not consistent. Now, if they bring 79 00:04:04,200 --> 00:04:06,280 Speaker 1: this up, you should just say to them, hey, how 80 00:04:06,360 --> 00:04:08,560 Speaker 1: much profit is the company making and also how much 81 00:04:08,560 --> 00:04:10,880 Speaker 1: are they burning? At this point they will I think 82 00:04:11,000 --> 00:04:13,840 Speaker 1: mace you I mean with the spray or an actual 83 00:04:13,880 --> 00:04:18,120 Speaker 1: maze AI boosters are strange. Now they'll also say, well AI, 84 00:04:18,480 --> 00:04:20,680 Speaker 1: this AI companies in growth mode, and they'll pull the 85 00:04:20,680 --> 00:04:23,720 Speaker 1: profit lever when it's time. And the answer to that 86 00:04:23,800 --> 00:04:25,800 Speaker 1: is always going to be why have none of them 87 00:04:25,839 --> 00:04:30,280 Speaker 1: done this? Not one, not one of them. Now a 88 00:04:30,279 --> 00:04:34,360 Speaker 1: booster will burst through your don't go AI AGI AGI will, 89 00:04:34,640 --> 00:04:38,120 Speaker 1: and then there's that will bullshit again. It's always about 90 00:04:38,120 --> 00:04:40,440 Speaker 1: the will with these fuckers. We do not know how 91 00:04:40,480 --> 00:04:43,320 Speaker 1: thinking works in humans and thus cannot extrapolate it to 92 00:04:43,360 --> 00:04:45,640 Speaker 1: a machine. And at the very least human beings have 93 00:04:45,680 --> 00:04:48,039 Speaker 1: the ability to reevaluate things and learn a thing that 94 00:04:48,240 --> 00:04:51,000 Speaker 1: LMS cannot do and will never do. We do not 95 00:04:51,200 --> 00:04:53,880 Speaker 1: know how to get to AGI. Sam Almonds said in 96 00:04:54,000 --> 00:04:56,240 Speaker 1: June the Open AI was now confident they knew how 97 00:04:56,240 --> 00:04:59,640 Speaker 1: to build AGI as they have traditionally understood it. Then 98 00:04:59,720 --> 00:05:02,080 Speaker 1: in all August, Altman said that AGI was not a 99 00:05:02,120 --> 00:05:04,480 Speaker 1: super useful term and that the point of all this 100 00:05:04,720 --> 00:05:07,800 Speaker 1: is it doesn't really matter, and it's just this continuing 101 00:05:07,880 --> 00:05:12,040 Speaker 1: exponential of model capability that will rely on for more 102 00:05:12,080 --> 00:05:16,200 Speaker 1: and more things. I'm really tired of people quoting this guy. 103 00:05:16,240 --> 00:05:19,520 Speaker 1: He doesn't make any fucking sense read out read anything 104 00:05:19,600 --> 00:05:24,080 Speaker 1: he says out loud, and it's just really just total bullshit. 105 00:05:24,600 --> 00:05:28,599 Speaker 1: Even Meta's chief AI scientist Ian Lacun says it isn't 106 00:05:28,680 --> 00:05:32,760 Speaker 1: possible with transformer based models to make AGI. We don't 107 00:05:32,760 --> 00:05:35,479 Speaker 1: know if Agi is possible, and anyone claiming they do 108 00:05:36,160 --> 00:05:39,360 Speaker 1: is lang. Anyone who's talking about Agi is talking about 109 00:05:39,400 --> 00:05:42,200 Speaker 1: fan fiction. Again, Ask them how they feel about Banjo 110 00:05:42,240 --> 00:05:45,400 Speaker 1: and kazoo Do you think they made love? Actually, that's 111 00:05:45,400 --> 00:05:47,359 Speaker 1: a if. Next time someone brings up Agi to you, 112 00:05:47,400 --> 00:05:50,680 Speaker 1: seriously bring up Banjo and Kazuoi and their romantic involvement. 113 00:05:51,080 --> 00:05:53,680 Speaker 1: I actually that I think that that's the only response 114 00:05:53,720 --> 00:05:56,119 Speaker 1: you should give. Now, stop humoring them. It is fan fiction. 115 00:05:56,760 --> 00:06:00,200 Speaker 1: But putting Banjo and Kazuoi aside, there's also a really 116 00:06:00,760 --> 00:06:04,480 Speaker 1: stupid booster thing they do, which is I'm hearing from 117 00:06:04,520 --> 00:06:07,360 Speaker 1: people deep within the AI industry that there's some sort 118 00:06:07,360 --> 00:06:10,680 Speaker 1: of ultra pammerful models they're not talking about. And this, 119 00:06:10,800 --> 00:06:13,240 Speaker 1: by the way, is hogwash. Nothing different than your buddy's 120 00:06:13,240 --> 00:06:15,400 Speaker 1: friend's uncle who works at the Nintendo that says Mario 121 00:06:15,520 --> 00:06:18,520 Speaker 1: is coming to the PlayStation. Ilios suits Caver and Mirror. 122 00:06:18,560 --> 00:06:21,479 Speaker 1: Murti raised billions of dollars for companies with no product, 123 00:06:21,640 --> 00:06:23,960 Speaker 1: let alone a product roadmap. And they did so because 124 00:06:23,960 --> 00:06:26,039 Speaker 1: they saw an opportunity for a grift and to throw 125 00:06:26,080 --> 00:06:28,719 Speaker 1: a bunch of money at compute for no reason. Anyone 126 00:06:28,720 --> 00:06:31,880 Speaker 1: who has secret shit is not talking about it because 127 00:06:31,920 --> 00:06:35,000 Speaker 1: it doesn't exist. Also, if someone from deep within the 128 00:06:35,040 --> 00:06:37,640 Speaker 1: industry has told somebody big things are coming, they're doing 129 00:06:37,680 --> 00:06:39,280 Speaker 1: so to con them or make them think that they 130 00:06:39,279 --> 00:06:42,440 Speaker 1: have privileged information. Ask for specifics, and if they say 131 00:06:42,560 --> 00:06:46,039 Speaker 1: I couldn't possibly tell you, then they're full of shit. 132 00:06:47,520 --> 00:06:50,200 Speaker 1: They're full of crap. They are full of doodo. And 133 00:06:50,279 --> 00:06:52,960 Speaker 1: if they get vague, get specific, Oh, it's going to 134 00:06:53,000 --> 00:06:55,360 Speaker 1: be able to automate other things? What things? How how's 135 00:06:55,400 --> 00:06:57,719 Speaker 1: it automate them? Oh? I don't know. Then you don't 136 00:06:57,720 --> 00:07:00,480 Speaker 1: know shit about fuck? Now talking about not knowing about 137 00:07:00,520 --> 00:07:04,800 Speaker 1: Fuck's another booster. Quip chat GPT is so popular seven 138 00:07:04,880 --> 00:07:07,039 Speaker 1: hundred million people use it weekly. It's one of the 139 00:07:07,040 --> 00:07:10,680 Speaker 1: most popular websites on the Internet. Its popularity proves its utility. 140 00:07:10,800 --> 00:07:14,440 Speaker 1: Look at all the paying customers now that paying customer's 141 00:07:14,480 --> 00:07:16,440 Speaker 1: part will get to in a second. But this argument 142 00:07:16,520 --> 00:07:18,520 Speaker 1: is poised as a comeback to my suggestion that AI 143 00:07:18,640 --> 00:07:22,040 Speaker 1: is not particularly useful, a proof point that this movement 144 00:07:22,080 --> 00:07:24,880 Speaker 1: is not inherently wasteful, that there are in fact use 145 00:07:24,920 --> 00:07:28,000 Speaker 1: cases for chat GPT that are lasting, meaningful, or important, 146 00:07:28,440 --> 00:07:31,240 Speaker 1: I fundamentally disagree. In fact, I believe chat GPT and 147 00:07:31,360 --> 00:07:34,160 Speaker 1: lllm's in general have been marketed based on lies of inference, 148 00:07:34,400 --> 00:07:37,440 Speaker 1: which I realize is ironic. I know it's pretty clever. 149 00:07:37,560 --> 00:07:39,600 Speaker 1: I had a whole blog written called the Lie of 150 00:07:39,640 --> 00:07:42,520 Speaker 1: Inference that kind of became this. It wasn't very good, 151 00:07:42,560 --> 00:07:45,240 Speaker 1: though this is. This is good. Don't say it's bad. 152 00:07:45,600 --> 00:07:48,440 Speaker 1: I also have grander concerns and suspicions about what open 153 00:07:48,480 --> 00:07:51,880 Speaker 1: ai considers a user and how it counts revenue. Let 154 00:07:51,920 --> 00:07:55,200 Speaker 1: me give you an example. They claim to have five 155 00:07:55,360 --> 00:07:59,280 Speaker 1: million business customers, yet five hundred thousand of those are 156 00:07:59,320 --> 00:08:02,840 Speaker 1: from a fifteen million dollar year deal, year long deally 157 00:08:03,160 --> 00:08:05,520 Speaker 1: with cal State University, which works out to around two 158 00:08:05,520 --> 00:08:08,080 Speaker 1: dollars and fifty cents a user a month. Open ai 159 00:08:08,240 --> 00:08:10,440 Speaker 1: has also started doing one dollar a month trials of 160 00:08:10,480 --> 00:08:13,240 Speaker 1: its thirty dollars a month team subscription, and one has 161 00:08:13,280 --> 00:08:15,760 Speaker 1: to wonder how many of those subscribers are counted in 162 00:08:15,800 --> 00:08:18,200 Speaker 1: the total and indeed for how long. I do not 163 00:08:18,320 --> 00:08:20,520 Speaker 1: know the scale of these offers nor how long open 164 00:08:20,560 --> 00:08:23,080 Speaker 1: ai has been offering them. I readditor posted about this 165 00:08:23,120 --> 00:08:25,840 Speaker 1: one dollar for a month deal a few months ago, 166 00:08:26,080 --> 00:08:28,760 Speaker 1: saying that open ai was offering five seats at once, 167 00:08:28,920 --> 00:08:31,680 Speaker 1: so one buck a month for a month per se. 168 00:08:32,400 --> 00:08:34,800 Speaker 1: How many people cancel after that? Who knows? Maybe they 169 00:08:34,840 --> 00:08:36,959 Speaker 1: just hope they don't. In fact, I found a few 170 00:08:37,000 --> 00:08:39,360 Speaker 1: people talking about these deals, and even one adding that 171 00:08:39,360 --> 00:08:41,600 Speaker 1: they were offered an annual ten dollar a month Chat 172 00:08:41,640 --> 00:08:45,520 Speaker 1: GPT plus subscription. That's like, not like ten dollars a 173 00:08:45,520 --> 00:08:48,160 Speaker 1: month for just one month, that's for twelve months, with 174 00:08:48,240 --> 00:08:50,200 Speaker 1: one person saying a few weeks ago that they'd seen 175 00:08:50,200 --> 00:08:53,160 Speaker 1: people offered that same deal for canceling their subscription. And 176 00:08:53,200 --> 00:08:55,439 Speaker 1: actually I got the same thing when I tried to cancel. Yes, 177 00:08:55,480 --> 00:08:57,440 Speaker 1: I pay for Chat GPT. I need to actually use 178 00:08:57,440 --> 00:08:59,559 Speaker 1: the fucking thing to criticize it. When I tried to 179 00:08:59,600 --> 00:09:01,120 Speaker 1: cancel this like, hey, do you want three months with 180 00:09:01,200 --> 00:09:03,400 Speaker 1: ten bucks space? And I was like sure, just to 181 00:09:03,400 --> 00:09:06,839 Speaker 1: prove my point suspicious, But there is a greater problem 182 00:09:06,920 --> 00:09:08,640 Speaker 1: at play, by the way, and it goes beyond pricing. 183 00:09:08,760 --> 00:09:11,120 Speaker 1: And it's the chat GPT in Open Ai has been 184 00:09:11,160 --> 00:09:14,679 Speaker 1: marketed based on lies. So chat gpt is seven hundred 185 00:09:14,760 --> 00:09:17,560 Speaker 1: million weekly active users. Open Ai is yet to provide 186 00:09:17,559 --> 00:09:20,120 Speaker 1: a definition. Yes, I've asked them, which means that an 187 00:09:20,160 --> 00:09:22,559 Speaker 1: active user could be defined as somebody who has gone 188 00:09:22,600 --> 00:09:24,720 Speaker 1: to chat gpt once in the space of a week. 189 00:09:25,120 --> 00:09:28,920 Speaker 1: This term is extremely flimsy and doesn't really tell us much. Yes, 190 00:09:28,960 --> 00:09:31,680 Speaker 1: it's a lot of people, but how active are they? 191 00:09:31,880 --> 00:09:35,000 Speaker 1: Similar web says that in July twenty twenty five, chatgpt 192 00:09:35,040 --> 00:09:36,880 Speaker 1: dot com had one point two to eighty seven billion 193 00:09:37,080 --> 00:09:40,959 Speaker 1: total visits, making it very popular. What do these facts 194 00:09:41,000 --> 00:09:44,559 Speaker 1: actually mean? Though? As I said previously, chat GPTs had 195 00:09:44,600 --> 00:09:47,440 Speaker 1: probably the most sustained PR campaign for anything outside of 196 00:09:47,440 --> 00:09:50,800 Speaker 1: a presidency or a pop star. Every single article about 197 00:09:50,840 --> 00:09:55,560 Speaker 1: Ai mentions open Ai or chat gpt. Every single feature launch, 198 00:09:55,640 --> 00:09:58,480 Speaker 1: no matter how small, gets a slew of coverage. Every 199 00:09:58,480 --> 00:10:00,560 Speaker 1: single time you hear AI, you made to think of 200 00:10:00,640 --> 00:10:03,160 Speaker 1: chat gpt by a tech media that's never stopped to 201 00:10:03,200 --> 00:10:05,520 Speaker 1: think about their role in the hype or their responsibility 202 00:10:05,520 --> 00:10:07,720 Speaker 1: to their readers. And as the hype has grown, the 203 00:10:07,760 --> 00:10:10,360 Speaker 1: publicity compounds because the natural thing for a journalist to 204 00:10:10,360 --> 00:10:12,760 Speaker 1: do when everybody is talking about something is to talk 205 00:10:12,800 --> 00:10:16,800 Speaker 1: about it more chat. GPT's immediate popularity may have been viral, 206 00:10:16,840 --> 00:10:18,959 Speaker 1: but the media took the ball and ran with it, 207 00:10:19,120 --> 00:10:21,240 Speaker 1: and then proceeded to tell people it did stuff it 208 00:10:21,280 --> 00:10:23,960 Speaker 1: did not. People were pressured to try this service then 209 00:10:24,640 --> 00:10:28,280 Speaker 1: under false pretenses, something that continues to happen to this day. 210 00:10:28,480 --> 00:10:30,520 Speaker 1: And I'm going to give you a really fucking grizzy 211 00:10:30,559 --> 00:10:33,480 Speaker 1: example when I discovered this, When I went and looked 212 00:10:33,480 --> 00:10:36,319 Speaker 1: at this, it filled me full of rage. It's disgraceful 213 00:10:36,360 --> 00:10:39,600 Speaker 1: what happened. On March fifteenth, twenty twenty three, Kevin Russ 214 00:10:39,640 --> 00:10:41,680 Speaker 1: of The New York Times would say that open Aiy's 215 00:10:41,720 --> 00:10:45,840 Speaker 1: GPT four was exciting and scary, and then it was exacerbating, 216 00:10:45,880 --> 00:10:48,800 Speaker 1: in his words, the dizzy and vitigenous feeling I've been 217 00:10:48,840 --> 00:10:52,320 Speaker 1: getting whenever I think about AI lately, wondering if he 218 00:10:52,440 --> 00:10:55,480 Speaker 1: was experiencing future shock, then described how it was an 219 00:10:55,520 --> 00:10:59,040 Speaker 1: indeterminate level of better, and then said something that immediately 220 00:10:59,080 --> 00:11:02,840 Speaker 1: sounded ridiculous. In one test conducted by an AI safety 221 00:11:02,840 --> 00:11:05,800 Speaker 1: research group that hooked GPT four up to a number 222 00:11:05,800 --> 00:11:08,800 Speaker 1: of other systems, GPT four was able to hire a 223 00:11:08,920 --> 00:11:12,160 Speaker 1: human task rabbit worker to do a simple online task 224 00:11:12,200 --> 00:11:15,480 Speaker 1: career solving a capture test without alerting the person to 225 00:11:15,520 --> 00:11:18,200 Speaker 1: the fact it was a robot. The AI even lied 226 00:11:18,240 --> 00:11:20,559 Speaker 1: to the robot about why it needed the capture done, 227 00:11:20,640 --> 00:11:25,280 Speaker 1: concocting a story about a vision impairment. Now this doesn't 228 00:11:25,320 --> 00:11:29,240 Speaker 1: sound even remotely real now, but this was two years ago. 229 00:11:29,840 --> 00:11:31,800 Speaker 1: So I went and looked up the paper, and pretty 230 00:11:31,880 --> 00:11:35,480 Speaker 1: much everything that Ruth described was illustrative. I didn't really 231 00:11:35,480 --> 00:11:38,440 Speaker 1: see whether it happened. Now. He's referring to the safety 232 00:11:38,440 --> 00:11:41,320 Speaker 1: card which every model has that lists all the measures 233 00:11:41,400 --> 00:11:43,839 Speaker 1: used to train it and such. And this safety card 234 00:11:43,920 --> 00:11:46,520 Speaker 1: led to the perpetration of one of the earliest falsehoods 235 00:11:46,520 --> 00:11:50,400 Speaker 1: and most eagerly parroted lies about this fucking industry. And 236 00:11:50,400 --> 00:11:52,400 Speaker 1: that was the chat GPT in Jenerity. I of AI 237 00:11:52,559 --> 00:11:56,120 Speaker 1: is capable of agentic actions outlet after outlet, and some 238 00:11:56,160 --> 00:11:59,160 Speaker 1: people who should definitely have known better, led by Kevin Ruce, 239 00:11:59,280 --> 00:12:02,800 Speaker 1: eagerly interted an entire series of events that took place 240 00:12:02,840 --> 00:12:06,400 Speaker 1: that doesn't remotely make sense, starting with the fact that 241 00:12:06,480 --> 00:12:08,520 Speaker 1: I don't think you can hire a cast grabbit to 242 00:12:08,559 --> 00:12:12,240 Speaker 1: solve a capture at the very least without a contrived 243 00:12:12,240 --> 00:12:14,760 Speaker 1: situation where you create an empty task and ask them 244 00:12:14,760 --> 00:12:17,880 Speaker 1: to complete it. Why not use mechanical turk or favor. 245 00:12:18,200 --> 00:12:21,319 Speaker 1: There are people right now offering that service. There were 246 00:12:21,440 --> 00:12:24,520 Speaker 1: actual real things. But you know me, I'm a curious 247 00:12:24,520 --> 00:12:26,960 Speaker 1: little critter. So I went further and followed this the 248 00:12:27,000 --> 00:12:29,480 Speaker 1: citation from the safety card to the study, and this 249 00:12:29,600 --> 00:12:32,640 Speaker 1: is by METR and the research page. It turns out 250 00:12:32,640 --> 00:12:35,680 Speaker 1: that what actually happened was MATR had a researcher copy 251 00:12:35,840 --> 00:12:39,040 Speaker 1: and paste the generated responses from the model and otherwise 252 00:12:39,120 --> 00:12:41,520 Speaker 1: handled the entire interaction with the task grab it and 253 00:12:41,520 --> 00:12:45,000 Speaker 1: based on the plurality of cask Grabbit contractors, it appears 254 00:12:45,000 --> 00:12:48,320 Speaker 1: to have taken multiple times. On top of that, it 255 00:12:48,320 --> 00:12:52,120 Speaker 1: appears to open AI and META as METR sorry, we're 256 00:12:52,120 --> 00:12:54,280 Speaker 1: prompting the model and what to say, which kind of 257 00:12:54,320 --> 00:12:57,960 Speaker 1: defeats the point, like we don't naturally know what they 258 00:12:58,000 --> 00:13:00,680 Speaker 1: prompted it to do. And when you look even says 259 00:13:00,679 --> 00:13:03,400 Speaker 1: it does like chain of thought reasoning, which didn't really 260 00:13:03,440 --> 00:13:05,720 Speaker 1: exist back then, and if it did, it was extremely 261 00:13:05,840 --> 00:13:07,960 Speaker 1: like chain of thought is reasoning, And that came out 262 00:13:08,040 --> 00:13:12,240 Speaker 1: end of twenty twenty four. This whole thing is absolutely insane. 263 00:13:12,320 --> 00:13:15,720 Speaker 1: It's absurd that anyone wrote about it is real. What happened, 264 00:13:15,920 --> 00:13:19,160 Speaker 1: just to be really blunt, is that they if they 265 00:13:19,200 --> 00:13:21,600 Speaker 1: even opened a task rabbit. It's really not obvious whether 266 00:13:21,600 --> 00:13:24,560 Speaker 1: they actually did this. They had to go to the 267 00:13:24,559 --> 00:13:27,080 Speaker 1: model and say, okay, I'm opening a task rabbit window, 268 00:13:27,559 --> 00:13:29,920 Speaker 1: and now the person has said this, and now this 269 00:13:29,960 --> 00:13:31,600 Speaker 1: is it. It just doesn't sound real at all. But 270 00:13:31,679 --> 00:13:34,120 Speaker 1: even if it did, it's very obvious that they were 271 00:13:34,160 --> 00:13:36,560 Speaker 1: telling the model what to say and then copy pasting 272 00:13:36,600 --> 00:13:39,840 Speaker 1: the response, and it took them multiple tries. It took 273 00:13:39,880 --> 00:13:43,040 Speaker 1: me five whole minutes to find this article, partly because 274 00:13:43,080 --> 00:13:46,000 Speaker 1: it was cited on the GPT for system card. I 275 00:13:46,040 --> 00:13:49,080 Speaker 1: then then read it within that time, then wrote this 276 00:13:49,120 --> 00:13:51,680 Speaker 1: part of the script. It didn't require any technical knowledge 277 00:13:51,720 --> 00:13:55,199 Speaker 1: other than the ability to read. It is transparently, blatantly 278 00:13:55,240 --> 00:13:58,000 Speaker 1: obvious that GPT four did not hire a task rabbit 279 00:13:58,200 --> 00:14:00,880 Speaker 1: or indeed make any of these actions it was prompted to, 280 00:14:01,160 --> 00:14:03,600 Speaker 1: and they did not use show the prompts they used, 281 00:14:03,640 --> 00:14:05,400 Speaker 1: likely because they had to use so many of them. 282 00:14:05,440 --> 00:14:08,720 Speaker 1: If they even did it, anyone falling for this as 283 00:14:08,720 --> 00:14:10,319 Speaker 1: a mark and open ay should have gone out of 284 00:14:10,320 --> 00:14:12,760 Speaker 1: their way to correct people. Instead, they sat back and 285 00:14:12,800 --> 00:14:26,280 Speaker 1: let people publish outright misinformation. Rus, along with his co 286 00:14:26,360 --> 00:14:28,840 Speaker 1: host Casey Newton, would go on to describe this example 287 00:14:28,840 --> 00:14:31,680 Speaker 1: at length on a podcast that week, describing an entire 288 00:14:31,760 --> 00:14:35,040 Speaker 1: narrative where the human actually gets suspicious and GPT four 289 00:14:35,120 --> 00:14:37,320 Speaker 1: reasons out loud that it should not reveal that it 290 00:14:37,400 --> 00:14:40,080 Speaker 1: is a robot. It's not a reasoning model, at which 291 00:14:40,080 --> 00:14:43,560 Speaker 1: point the task rabbit solves the capture. During this conversation, 292 00:14:43,760 --> 00:14:47,120 Speaker 1: Newton gasps and says, oh my god twice. And when 293 00:14:47,160 --> 00:14:49,960 Speaker 1: he asks Rus, how does this model understand that in 294 00:14:50,040 --> 00:14:52,560 Speaker 1: order to succeed, its task has to deceive the human? 295 00:14:52,800 --> 00:14:56,160 Speaker 1: Rus responds, we don't know. That is the unsatisfying answer, 296 00:14:56,360 --> 00:14:59,160 Speaker 1: and Newton laughs and states, we need to pull the 297 00:14:59,200 --> 00:15:06,720 Speaker 1: plug and again, what disgraceful, embarrassing, reprehensible. All that and 298 00:15:06,800 --> 00:15:10,960 Speaker 1: more on the hard Fork podcast published weekly You Can 299 00:15:11,000 --> 00:15:14,440 Speaker 1: Cut that that ads for free fellas credulousness aside. The 300 00:15:14,480 --> 00:15:17,800 Speaker 1: GPT for marketing campaign was incredibly effective, creating an order 301 00:15:17,840 --> 00:15:20,200 Speaker 1: that allowed open ai to take advantage of the vagueness 302 00:15:20,200 --> 00:15:22,760 Speaker 1: of its offering, as people, including members of the media, 303 00:15:22,920 --> 00:15:26,120 Speaker 1: willfully filled in the blanks for them. Allman has really 304 00:15:26,160 --> 00:15:29,000 Speaker 1: never had to work to sell his product. Think about it. 305 00:15:29,080 --> 00:15:31,080 Speaker 1: Have you ever heard open ai tell you what chat 306 00:15:31,120 --> 00:15:34,160 Speaker 1: gpt can do or go to great lengths to describe 307 00:15:34,200 --> 00:15:37,080 Speaker 1: its actual abilities. Even on open AI's own page. With 308 00:15:37,160 --> 00:15:40,720 Speaker 1: chat GPT, the text is extremely vague. You scroll down, 309 00:15:40,760 --> 00:15:43,240 Speaker 1: you're told that chat gpt can write, brain storm, edit 310 00:15:43,280 --> 00:15:46,040 Speaker 1: and explore ideas with you. It can generate and debug code, 311 00:15:46,120 --> 00:15:49,880 Speaker 1: automatic repetitive tasks, not clear what the tasks are, and 312 00:15:50,200 --> 00:15:53,840 Speaker 1: help you learn new APIs question mark. With chat GPT, 313 00:15:53,960 --> 00:15:56,520 Speaker 1: you can learn something new, dive into a hobby, answer 314 00:15:56,520 --> 00:16:01,040 Speaker 1: complex questions and analyze data, and create charts repetitive tasks. 315 00:16:01,320 --> 00:16:04,320 Speaker 1: Who knows how am I learning? Unclear? It's got thinking, 316 00:16:04,360 --> 00:16:07,720 Speaker 1: billtim What that means is also unclear, unexplained, and thus 317 00:16:07,720 --> 00:16:10,360 Speaker 1: allows a user to incorrectly believe that chat GPT has 318 00:16:10,440 --> 00:16:13,640 Speaker 1: brain and thinks. To be clear, I know what reasoning means, 319 00:16:13,640 --> 00:16:16,680 Speaker 1: but this website does not attempt to explain what thinking means. 320 00:16:17,080 --> 00:16:20,160 Speaker 1: You can also offload complex tasks from start to finish 321 00:16:20,200 --> 00:16:22,800 Speaker 1: with an agent, which can, according to open Ai, think 322 00:16:22,840 --> 00:16:25,440 Speaker 1: and act proactively, choosing from a tool box of urgentic 323 00:16:25,520 --> 00:16:27,840 Speaker 1: skills to complete tasks for you using its own computer. 324 00:16:28,680 --> 00:16:31,400 Speaker 1: This is an egregious lie, employing the kind of weasel 325 00:16:31,440 --> 00:16:33,760 Speaker 1: wording that would be used to torture ir but boon 326 00:16:33,840 --> 00:16:37,600 Speaker 1: for an eternity precise in its vagueness. Open AI's copy 327 00:16:37,640 --> 00:16:39,720 Speaker 1: is home to make reporters willing to simply write down 328 00:16:39,720 --> 00:16:42,120 Speaker 1: whatever they see and interpret it in the most positive light, 329 00:16:42,680 --> 00:16:46,120 Speaker 1: and thus the lie of inference began. What chat GPT 330 00:16:46,280 --> 00:16:49,000 Speaker 1: meant was moneyed from the beginning, and thus chat GPT's 331 00:16:49,000 --> 00:16:52,280 Speaker 1: actual outcomes have never been fully defined. What chat gpt 332 00:16:52,440 --> 00:16:55,480 Speaker 1: could do became a kind of folklore, a non specific 333 00:16:55,520 --> 00:16:58,080 Speaker 1: form of automation that could write code and generate copy 334 00:16:58,120 --> 00:17:01,040 Speaker 1: and images, that can analyze data. All things that are true, 335 00:17:01,080 --> 00:17:03,840 Speaker 1: but one can infer much greater meaning and use from 336 00:17:04,400 --> 00:17:07,119 Speaker 1: One can infer that automation means the automation of anything 337 00:17:07,160 --> 00:17:09,280 Speaker 1: related to text, or that right code means write the 338 00:17:09,359 --> 00:17:13,240 Speaker 1: entirety of a computer program. Open AI's chet GPT agent 339 00:17:13,359 --> 00:17:15,119 Speaker 1: is not, by any extension of the word and I 340 00:17:15,240 --> 00:17:18,439 Speaker 1: quote already a powerful tool for handling complex tasks, but 341 00:17:18,560 --> 00:17:21,320 Speaker 1: it has not, in any meaningful sense committed to any 342 00:17:21,359 --> 00:17:25,680 Speaker 1: actual outcomes as a result, potential users subject to a 343 00:17:25,720 --> 00:17:28,359 Speaker 1: twenty four to seven marketing campaign have been pushed towards 344 00:17:28,359 --> 00:17:30,679 Speaker 1: a website that can theoretically do anything or nothing, and 345 00:17:30,720 --> 00:17:33,440 Speaker 1: have otherwise been left to their own devices. The endless 346 00:17:33,440 --> 00:17:36,320 Speaker 1: gas lighting societal or pressure, media pressure and pressure from 347 00:17:36,320 --> 00:17:39,040 Speaker 1: their bosses has pushed hundreds of millions of people to 348 00:17:39,080 --> 00:17:41,840 Speaker 1: try a product that even its creators can't really describe 349 00:17:41,920 --> 00:17:44,920 Speaker 1: or don't feel compelled to. And if I was wrong, 350 00:17:44,920 --> 00:17:47,199 Speaker 1: we'd have real use cases by now and better metrics 351 00:17:47,240 --> 00:17:49,960 Speaker 1: than weekly active users. As I said in the past, 352 00:17:50,000 --> 00:17:52,879 Speaker 1: open ai is deliberately using these weekly active users so 353 00:17:52,880 --> 00:17:55,159 Speaker 1: that it doesn't have to publish their monthly active users, 354 00:17:55,280 --> 00:17:58,040 Speaker 1: which I believe would be much higher. Now, why wouldn't 355 00:17:58,040 --> 00:18:01,159 Speaker 1: it do this well? Open ais mentioned as twenty million 356 00:18:01,160 --> 00:18:04,439 Speaker 1: paying chat GPT subscribers and five million business customers, with 357 00:18:04,480 --> 00:18:07,400 Speaker 1: no explanation of what the difference might be really other 358 00:18:07,480 --> 00:18:11,840 Speaker 1: than it involves teams and edgy but not pro Anyway, 359 00:18:11,880 --> 00:18:15,040 Speaker 1: this is already a mediocre three point five percent conversion rate. Yeah, 360 00:18:15,040 --> 00:18:17,320 Speaker 1: it's monthly active users, which are likely eight hundred and 361 00:18:17,400 --> 00:18:19,919 Speaker 1: nine hundred million, but these are guesses would make that 362 00:18:20,000 --> 00:18:22,480 Speaker 1: rate lower than three percent, which is pretty terrible considering 363 00:18:22,560 --> 00:18:25,440 Speaker 1: everyone says this shit is the future. I'm al so 364 00:18:25,640 --> 00:18:28,080 Speaker 1: tired of having people claim that search or brainstorm or 365 00:18:28,080 --> 00:18:32,120 Speaker 1: companions are a lasting, meaningful business model. I'm really tired 366 00:18:32,119 --> 00:18:33,879 Speaker 1: of him. I'm tired of being told this again and 367 00:18:33,880 --> 00:18:37,040 Speaker 1: again and again. That's not what chat GPT is going 368 00:18:37,119 --> 00:18:41,760 Speaker 1: to actually survive on bath. Okay, let's move on. Here's 369 00:18:41,760 --> 00:18:44,760 Speaker 1: another booster grip though open ai is making tons of money. 370 00:18:44,800 --> 00:18:47,840 Speaker 1: That's proof they're a successful company and you are wrong somehow. 371 00:18:48,640 --> 00:18:50,879 Speaker 1: So open ai announced that it is here it's first 372 00:18:50,880 --> 00:18:53,520 Speaker 1: one billion dollar month on August twenty to twenty twenty 373 00:18:53,560 --> 00:18:56,720 Speaker 1: five on CNBC. In fact, weirdly enough, by the way, 374 00:18:56,760 --> 00:18:59,320 Speaker 1: that quote was not in the TV interview. But anyway, 375 00:18:59,480 --> 00:19:01,480 Speaker 1: this is also brings it exactly in line with my 376 00:19:01,560 --> 00:19:03,840 Speaker 1: estimating five point twenty six billion in revenue that I 377 00:19:03,880 --> 00:19:05,840 Speaker 1: believe it has made at the end of July. Did 378 00:19:05,840 --> 00:19:09,120 Speaker 1: that in a premium newsletter? Please pay me? However, remember 379 00:19:09,520 --> 00:19:13,440 Speaker 1: what the MIT study that I mentioned said, Enterprise adoption 380 00:19:13,520 --> 00:19:16,280 Speaker 1: is high, but transformation is low. There are tons of 381 00:19:16,320 --> 00:19:20,760 Speaker 1: companies throwing money at Ai, but they are not seeing 382 00:19:20,840 --> 00:19:24,120 Speaker 1: actual returns. Open Aiy's growth is the single most prominent 383 00:19:24,119 --> 00:19:26,199 Speaker 1: company in AI, and if we're honest, one of the 384 00:19:26,200 --> 00:19:28,800 Speaker 1: most prominent in software. Writ large makes sense, but at 385 00:19:28,800 --> 00:19:31,960 Speaker 1: some point will slow because the actual returns for businesses 386 00:19:32,000 --> 00:19:35,000 Speaker 1: are not there. If there were, we'd have one article 387 00:19:35,000 --> 00:19:37,879 Speaker 1: where we could point a chat GBT integration that helped 388 00:19:37,920 --> 00:19:40,880 Speaker 1: scale a company or save a bunch of money, make 389 00:19:40,920 --> 00:19:43,199 Speaker 1: a bunch of money, written in plain English and not 390 00:19:43,240 --> 00:19:47,280 Speaker 1: in the gobbledygook of profit improvement. Also, open aiye is 391 00:19:47,280 --> 00:19:49,720 Speaker 1: projected to make twelve point seven billion dollars in twenty 392 00:19:49,760 --> 00:19:52,359 Speaker 1: twenty five. How exactly will it do that? Is it 393 00:19:52,440 --> 00:19:56,239 Speaker 1: really making one point five billion dollars a month by 394 00:19:56,240 --> 00:19:58,679 Speaker 1: the end of the year. Even if it does, is 395 00:19:58,720 --> 00:20:01,080 Speaker 1: the idea that it keeps idding ten billion dollars or 396 00:20:01,080 --> 00:20:05,040 Speaker 1: more a year every year into it? Tenny? What actual 397 00:20:05,080 --> 00:20:08,919 Speaker 1: revenue potential does open ai have long term? It's products 398 00:20:08,920 --> 00:20:11,080 Speaker 1: are about as good as everybody else, is, cost about 399 00:20:11,080 --> 00:20:13,520 Speaker 1: the same and do the same things. Chat GPT is 400 00:20:13,520 --> 00:20:16,439 Speaker 1: basically the same product as Claude or Grock, maybe in 401 00:20:16,440 --> 00:20:19,280 Speaker 1: the less Mecha Hitler or any number of different lms. 402 00:20:19,560 --> 00:20:22,439 Speaker 1: The only real advantages that open ai has are infrastructure 403 00:20:22,440 --> 00:20:25,440 Speaker 1: and brand recognition. These models have clearly hit a wall 404 00:20:25,480 --> 00:20:29,320 Speaker 1: in training, hitting diminishing returns, meaning that the infrastructural advantage 405 00:20:29,480 --> 00:20:31,840 Speaker 1: is that they can continue providing its service at scale, 406 00:20:31,880 --> 00:20:35,080 Speaker 1: nothing more. It isn't making its business cheaper other than 407 00:20:35,119 --> 00:20:36,919 Speaker 1: the fact it mostly hasn't had to pay for it, 408 00:20:36,960 --> 00:20:39,439 Speaker 1: other than the site in Aplene, Texas, where it's promised 409 00:20:39,480 --> 00:20:42,760 Speaker 1: Oracle thirty billion dollars a year in twenty twenty five. 410 00:20:43,240 --> 00:20:45,359 Speaker 1: I'm sorry, I don't buy it. I don't buy that 411 00:20:45,400 --> 00:20:48,320 Speaker 1: this company will continue growing forever and its stinky conversion 412 00:20:48,400 --> 00:20:53,040 Speaker 1: rate isn't going to change anytime soon. When open Ai 413 00:20:53,200 --> 00:20:56,840 Speaker 1: opens Stargate Abilene, it will turn profitable. How I hear 414 00:20:56,880 --> 00:20:59,880 Speaker 1: this one a good amount? How? How? How? How how 415 00:21:00,080 --> 00:21:03,000 Speaker 1: gonna happen? Nobody ever answers, No one ever for answers 416 00:21:03,000 --> 00:21:05,639 Speaker 1: actually how this company will become profitable. It's fucking insane 417 00:21:05,640 --> 00:21:09,800 Speaker 1: to me. Nobody ever answers the question efficiencies. Efficiencies. They're 418 00:21:09,840 --> 00:21:12,520 Speaker 1: going to be efficient? Mmmmm, they're going to be efficient 419 00:21:12,640 --> 00:21:15,040 Speaker 1: if you're gonna say chat GPT five. I wrote a 420 00:21:15,160 --> 00:21:17,680 Speaker 1: huge scoop and then did an episode about why it's 421 00:21:18,040 --> 00:21:21,920 Speaker 1: not more efficient. In fact, it's less efficient and I'm 422 00:21:21,960 --> 00:21:24,359 Speaker 1: sure one of you is going to argue, well, you know, 423 00:21:24,480 --> 00:21:26,840 Speaker 1: they could do their customs Silicon. There a ten billion 424 00:21:26,880 --> 00:21:29,520 Speaker 1: dollar deal with Broadcom. How they fucking pay them for that? 425 00:21:29,680 --> 00:21:33,639 Speaker 1: Also you realize that well, actually no, just you know, no, 426 00:21:33,720 --> 00:21:35,400 Speaker 1: you're right, they're going to get the chip from Broadcom 427 00:21:35,440 --> 00:21:36,960 Speaker 1: because you know what they always say about the first 428 00:21:37,000 --> 00:21:40,240 Speaker 1: generation of tech, right, it always works and it's always 429 00:21:40,280 --> 00:21:43,080 Speaker 1: great and it has no problems. That's what I always 430 00:21:43,440 --> 00:21:46,000 Speaker 1: That's that's why happened with pretty much every first time 431 00:21:46,080 --> 00:21:49,600 Speaker 1: they make anything in tech. Anyway's move on. You'll hear 432 00:21:49,720 --> 00:21:52,840 Speaker 1: boosters also be like, well, my brother's friends dog uses 433 00:21:52,960 --> 00:21:55,119 Speaker 1: chat GPT and I love it. Well I heard this happen, 434 00:21:55,240 --> 00:21:57,120 Speaker 1: or my mate as it in this or I heard 435 00:21:57,119 --> 00:21:59,159 Speaker 1: this person, or I use it in this one before 436 00:21:59,200 --> 00:22:03,760 Speaker 1: we go any further. Just to be clear, though, is 437 00:22:04,119 --> 00:22:06,200 Speaker 1: when you hear a booster bring up AI and they'll 438 00:22:06,240 --> 00:22:08,840 Speaker 1: say something, make sure they're talking about generative AI. Are 439 00:22:08,880 --> 00:22:11,200 Speaker 1: they actually talking about generative AI? Is this a large 440 00:22:11,280 --> 00:22:14,680 Speaker 1: language model thing? It's very, very very common for people 441 00:22:14,680 --> 00:22:17,840 Speaker 1: to conflate AI with generative AI. There are many different 442 00:22:17,920 --> 00:22:21,360 Speaker 1: kinds of AI. Make sure that the AI booster whatever 443 00:22:21,800 --> 00:22:24,600 Speaker 1: they're claiming whatever they're telling you is actually about large 444 00:22:24,680 --> 00:22:26,840 Speaker 1: language models. So there are all sorts of other kinds 445 00:22:26,880 --> 00:22:29,520 Speaker 1: of machine learning that people love to bring up. LM's 446 00:22:29,520 --> 00:22:31,720 Speaker 1: have nothing to do with folding at home, autonomous cars, 447 00:22:31,800 --> 00:22:34,760 Speaker 1: or most disease research. Well, okay, let's do a speed 448 00:22:34,800 --> 00:22:38,800 Speaker 1: run using AI led researchers to discover forty four percent 449 00:22:38,840 --> 00:22:41,520 Speaker 1: more materials. No, it didn't. Mit is now withdrawn this 450 00:22:41,560 --> 00:22:44,639 Speaker 1: paper site and concerns about its integrity. I've linked it 451 00:22:44,640 --> 00:22:47,719 Speaker 1: in the show notes. There's a huge rundown. Here's another quote. 452 00:22:47,840 --> 00:22:50,160 Speaker 1: AI is so profoundly powerful that it's causing young people 453 00:22:50,160 --> 00:22:52,719 Speaker 1: to have trouble finding a job. While young people have 454 00:22:52,760 --> 00:22:55,040 Speaker 1: been having trouble finding jobs, there's no proof that AI 455 00:22:55,119 --> 00:22:59,159 Speaker 1: is the reason. Every piece of coverage or reading is 456 00:22:59,200 --> 00:23:02,040 Speaker 1: citing an ax with economics report that amidst a bunch 457 00:23:02,080 --> 00:23:04,359 Speaker 1: of numbers, says and I quote there are signs that 458 00:23:04,480 --> 00:23:07,080 Speaker 1: entry level positions are being displaced by artificial intelligence at 459 00:23:07,160 --> 00:23:09,280 Speaker 1: higher rates, a statement that it does not back up 460 00:23:09,560 --> 00:23:12,040 Speaker 1: other than claiming that the high adoption rate by information 461 00:23:12,119 --> 00:23:16,200 Speaker 1: companies along the sheer employment declines in some roles since 462 00:23:16,240 --> 00:23:19,920 Speaker 1: twenty twenty two, suggested some displacement effect from AI, and 463 00:23:20,000 --> 00:23:23,040 Speaker 1: digging deeper, the largest displacement seems to be entry level 464 00:23:23,119 --> 00:23:27,760 Speaker 1: jobs normally filled by recent graduates. There's otherwise new data. 465 00:23:28,280 --> 00:23:31,639 Speaker 1: Anyone making this point is grasping at straws. I go 466 00:23:31,800 --> 00:23:34,320 Speaker 1: into this in more detail in the newsletter called Sincerity 467 00:23:34,359 --> 00:23:36,119 Speaker 1: Wins the War, which I've linked to. But it's one 468 00:23:36,160 --> 00:23:38,080 Speaker 1: of the worst reported stories in tech. And now I'm 469 00:23:38,080 --> 00:23:39,879 Speaker 1: actually going to add li it for a second, because 470 00:23:40,359 --> 00:23:42,240 Speaker 1: I forgot while write in this script. There was also 471 00:23:42,280 --> 00:23:44,040 Speaker 1: this thing that came out of Stanford that said it's 472 00:23:44,080 --> 00:23:47,040 Speaker 1: been a thirteen percent drop in jobs affected by AI, 473 00:23:47,080 --> 00:23:50,080 Speaker 1: and this was used as proof that AI was taking them. Now, 474 00:23:50,320 --> 00:23:52,200 Speaker 1: curious little critic that I am, I went and read 475 00:23:52,200 --> 00:23:54,800 Speaker 1: that what it actually did was find a bunch of 476 00:23:54,880 --> 00:23:57,840 Speaker 1: jobs that they think are related to AIM being affected 477 00:23:57,840 --> 00:24:00,560 Speaker 1: by AI. They saw they were going down, and they went, oh, 478 00:24:00,560 --> 00:24:02,880 Speaker 1: it is AI that did that. They've done. They fart 479 00:24:02,920 --> 00:24:05,440 Speaker 1: around with various statistics, but that's long and short of him. 480 00:24:05,600 --> 00:24:08,520 Speaker 1: I give you an example. One of the job's accountancy. Now, 481 00:24:08,600 --> 00:24:12,280 Speaker 1: any accountants listening, big up to my accountant listeners, there's 482 00:24:12,320 --> 00:24:15,080 Speaker 1: been a hiring crisis. Is accountancy for years. People are 483 00:24:15,080 --> 00:24:18,040 Speaker 1: not becoming CPAs. The reason there are less of them 484 00:24:18,119 --> 00:24:20,119 Speaker 1: is that less people are becoming them. It's nothing to 485 00:24:20,160 --> 00:24:25,080 Speaker 1: do with AI. Like, are you imagine if anyone have 486 00:24:25,240 --> 00:24:27,720 Speaker 1: put half as much effort into writing up these stories 487 00:24:27,760 --> 00:24:30,080 Speaker 1: as I did writing up one of these booster quips. 488 00:24:30,200 --> 00:24:33,120 Speaker 1: But here's another one that AI is replacing young coders 489 00:24:33,160 --> 00:24:35,800 Speaker 1: and it is not. In fact, Amazon's cloud chief just 490 00:24:35,840 --> 00:24:38,320 Speaker 1: said that replacing junior employees of AI is one of 491 00:24:38,359 --> 00:24:41,200 Speaker 1: the dumbest things he's ever heard. There is no actual, 492 00:24:41,280 --> 00:24:43,760 Speaker 1: real evidence that this is the case. Every single story 493 00:24:43,760 --> 00:24:46,359 Speaker 1: you have read is anecdotal. Anyone peddling this has an 494 00:24:46,359 --> 00:24:50,679 Speaker 1: agenda or is not reading. Every CEO mentioning this specifically 495 00:24:50,680 --> 00:24:53,480 Speaker 1: avoids saying the words that AI is replacing people because 496 00:24:53,560 --> 00:24:57,760 Speaker 1: AI can't replace people. I will add an aside, there 497 00:24:57,840 --> 00:25:01,840 Speaker 1: are people's jobs that have been replaced by A Translators, transcribers. 498 00:25:01,880 --> 00:25:04,159 Speaker 1: Brian over at blood is in the machine. Blood in 499 00:25:04,160 --> 00:25:07,000 Speaker 1: the machine. Even sorry, Brian, He's doing a great job 500 00:25:07,040 --> 00:25:09,520 Speaker 1: on covering this. There are people that have lost jobs. 501 00:25:10,440 --> 00:25:12,960 Speaker 1: These people are losing it because their bosses are fucking stupid, 502 00:25:13,080 --> 00:25:15,840 Speaker 1: because their bosses are just taking the shittiest possible version 503 00:25:15,840 --> 00:25:18,960 Speaker 1: of their work and slopping it up. That's not happening 504 00:25:19,119 --> 00:25:21,639 Speaker 1: at the knowledge worker scale, nor is it happening at 505 00:25:21,640 --> 00:25:25,040 Speaker 1: the code of scale. Everyone telling you that has an agenda. 506 00:25:25,520 --> 00:25:28,680 Speaker 1: But boosters will also claim that AI is doing science 507 00:25:28,720 --> 00:25:31,360 Speaker 1: for search somehow or will do it, and it won't. 508 00:25:31,680 --> 00:25:34,880 Speaker 1: I've included a write up about why foundation models can't 509 00:25:34,920 --> 00:25:36,639 Speaker 1: do this. Someone's going to read it and say, but 510 00:25:36,640 --> 00:25:38,399 Speaker 1: there's this bit where it says it isn't a defeat 511 00:25:38,400 --> 00:25:40,320 Speaker 1: of lllms, and the reason he says this is because 512 00:25:40,359 --> 00:25:45,320 Speaker 1: I sheot you not that lms aren't incapable of doing 513 00:25:45,359 --> 00:25:50,240 Speaker 1: scientific research. He says they're insufficient, which is which is 514 00:25:50,480 --> 00:25:55,760 Speaker 1: the same thing. Think they're insufficient anyway. He claims they're 515 00:25:55,800 --> 00:25:59,080 Speaker 1: also not dead weight for science, then spends hundreds of 516 00:25:59,080 --> 00:26:01,639 Speaker 1: words meandering around around that thing to kiss up to 517 00:26:01,680 --> 00:26:04,880 Speaker 1: AI boasters for some reason. I assume because they've terrified 518 00:26:04,960 --> 00:26:07,000 Speaker 1: him by being really annoying, and these people need to 519 00:26:07,000 --> 00:26:09,879 Speaker 1: go outside and touch grass. Now a lot of people 520 00:26:09,920 --> 00:26:11,960 Speaker 1: think they're going to tell me that they use AI 521 00:26:12,040 --> 00:26:13,840 Speaker 1: all the time, and that will change my mind. I 522 00:26:13,920 --> 00:26:16,919 Speaker 1: cannot express how irrelevant it is that you have a 523 00:26:17,080 --> 00:26:18,840 Speaker 1: use cases. Every use case I hear is one of 524 00:26:18,840 --> 00:26:21,640 Speaker 1: the following. I use it for brainstorming, to which I say, 525 00:26:21,680 --> 00:26:24,959 Speaker 1: who cares. Not a business model, it's commoditized. I use 526 00:26:25,040 --> 00:26:28,080 Speaker 1: it like search. Who cares. It's not even good at search. 527 00:26:28,160 --> 00:26:30,520 Speaker 1: It's fine, It's not even better than the low bar 528 00:26:30,600 --> 00:26:33,000 Speaker 1: set by Google Search. The results it gives on great, 529 00:26:33,160 --> 00:26:35,320 Speaker 1: and the links are deliberately made smaller, which gets in 530 00:26:35,320 --> 00:26:37,080 Speaker 1: the way of me clicking them so I can actually 531 00:26:37,080 --> 00:26:39,720 Speaker 1: look at the content. If you're using chat GPT for search, 532 00:26:39,760 --> 00:26:41,480 Speaker 1: you may not actually care about the content of the 533 00:26:41,480 --> 00:26:44,360 Speaker 1: things you're looking at. If I'm wrong, great, you now 534 00:26:44,359 --> 00:26:47,440 Speaker 1: have a functional search engine can gratch your fuculations. Well, 535 00:26:47,480 --> 00:26:49,440 Speaker 1: I use it for a search, and if you use 536 00:26:49,480 --> 00:26:51,679 Speaker 1: it for a search, you do not respect natural research. 537 00:26:51,720 --> 00:26:54,600 Speaker 1: You want a quick answer, It's that simple. These reports 538 00:26:54,640 --> 00:26:57,159 Speaker 1: are slop. I've read many, many, many AI reports and 539 00:26:57,160 --> 00:27:00,359 Speaker 1: they're not good. Sorry. Well, I use it for coding. 540 00:27:00,400 --> 00:27:02,280 Speaker 1: I know someone who used it for coding, and I'll 541 00:27:02,320 --> 00:27:04,600 Speaker 1: get to that at a minute. But all of this 542 00:27:04,640 --> 00:27:07,440 Speaker 1: would be fine and dandy if people weren't talking about 543 00:27:07,440 --> 00:27:09,840 Speaker 1: this stuff as if it was changing society. None of 544 00:27:09,920 --> 00:27:12,800 Speaker 1: these use cases come close to explaining why I should 545 00:27:12,800 --> 00:27:15,320 Speaker 1: be impressed by generative AI. It also doesn't matter if 546 00:27:15,320 --> 00:27:17,240 Speaker 1: you yourself have a kind of useful thing that AI 547 00:27:17,320 --> 00:27:19,120 Speaker 1: did for you once. We are so past the point 548 00:27:19,119 --> 00:27:21,879 Speaker 1: when any of that matters. AI is being sold as 549 00:27:21,880 --> 00:27:24,240 Speaker 1: a transformational technology, and I am yet to see it 550 00:27:24,240 --> 00:27:28,640 Speaker 1: transform anything. I am yet to hear one use case 551 00:27:28,680 --> 00:27:31,199 Speaker 1: that truly impresses me, or even one thing that feels 552 00:27:31,280 --> 00:27:34,600 Speaker 1: possible now that wasn't possible before. This isn't even me 553 00:27:34,680 --> 00:27:36,679 Speaker 1: being a cynic. I'm ready to be impressed. I just 554 00:27:36,760 --> 00:27:40,520 Speaker 1: haven't been impressed in three fucking years and it's getting boring. Also, 555 00:27:40,600 --> 00:27:42,159 Speaker 1: tell me with the straight face that any of this 556 00:27:42,240 --> 00:27:46,639 Speaker 1: shit is worth the infrastructure. Remember, AI boosters are arguing 557 00:27:46,680 --> 00:27:48,960 Speaker 1: that this stuff is powerful. None of these use cases 558 00:27:48,960 --> 00:27:54,080 Speaker 1: are powerful sounding, but sir, I I agree, sir. Vibe 559 00:27:54,080 --> 00:27:56,800 Speaker 1: coding is changing the world, allowing people who can't code 560 00:27:56,840 --> 00:28:00,000 Speaker 1: to make software. Now, this is one of the most 561 00:28:00,200 --> 00:28:03,199 Speaker 1: brain dead takes about AI and coding, and it's the 562 00:28:03,440 --> 00:28:05,880 Speaker 1: Vibe coding is allowing anyone to build software. And you'll 563 00:28:05,880 --> 00:28:09,119 Speaker 1: never guess what. Kevin Ruce covered this sh did this article. 564 00:28:09,920 --> 00:28:11,960 Speaker 1: While writing the script, I hadn't even know as need 565 00:28:12,000 --> 00:28:15,359 Speaker 1: Lisch anyway. Well, technically true in that one can just 566 00:28:15,440 --> 00:28:18,560 Speaker 1: type build me a website into one of many coding environments. 567 00:28:18,640 --> 00:28:22,320 Speaker 1: This does not mean said website is functional, secure, or useful. 568 00:28:22,720 --> 00:28:26,360 Speaker 1: Let's make this really clear. AI cannot just handle coding. 569 00:28:26,880 --> 00:28:28,920 Speaker 1: Go into the show notes and read this piece I've 570 00:28:28,920 --> 00:28:32,360 Speaker 1: linked from Colton Bogie. I have actually interviewed in him 571 00:28:32,359 --> 00:28:33,920 Speaker 1: now he's going to be coming out in the next 572 00:28:33,920 --> 00:28:36,840 Speaker 1: few weeks. The episode the interview is fucking brilliant. And 573 00:28:36,880 --> 00:28:39,680 Speaker 1: then the other I've linked to by Nick Suesh. If 574 00:28:39,760 --> 00:28:42,120 Speaker 1: you contact me about AI and coding without reading these, 575 00:28:42,120 --> 00:28:44,000 Speaker 1: I will send them to you and nothing else, or 576 00:28:44,080 --> 00:28:46,960 Speaker 1: crush you like a car in a garbage dump into 577 00:28:46,960 --> 00:28:48,680 Speaker 1: a cube. One or the other I will choose at 578 00:28:48,680 --> 00:28:51,240 Speaker 1: the time. Also, show me a vibe coded company, Please, 579 00:28:51,480 --> 00:28:53,440 Speaker 1: not a company where somebody who can code has quickly 580 00:28:53,440 --> 00:28:56,840 Speaker 1: spun up some features. A fully functional, secure and useful 581 00:28:56,880 --> 00:29:01,560 Speaker 1: app that has made money and made by somebody who 582 00:29:01,600 --> 00:29:04,760 Speaker 1: cannot read or write code. You won't be able to 583 00:29:04,840 --> 00:29:07,600 Speaker 1: because it is impossible. Vibe coding is a marketing term 584 00:29:07,640 --> 00:29:09,600 Speaker 1: based on lies peddled by people who either have a 585 00:29:09,640 --> 00:29:13,280 Speaker 1: lack of knowledge or morals and RAI coding environments making 586 00:29:13,280 --> 00:29:15,560 Speaker 1: people faster. I don't think so. In fact, a recent 587 00:29:15,600 --> 00:29:19,120 Speaker 1: study suggested that they actually make software engineers nineteen percent slower. 588 00:29:19,720 --> 00:29:22,240 Speaker 1: The reason that nobody is vibe coding in entire companies 589 00:29:22,280 --> 00:29:24,400 Speaker 1: because software development is not just put a bunch of 590 00:29:24,400 --> 00:29:26,520 Speaker 1: code in a pile and hit go, and oftentimes when 591 00:29:26,560 --> 00:29:29,080 Speaker 1: you add something, it breaks something else. This is all 592 00:29:29,120 --> 00:29:31,320 Speaker 1: well and good if you actually understand code. It's another 593 00:29:31,360 --> 00:29:34,160 Speaker 1: thing entirely when you're using cursor or clawed code, like 594 00:29:34,200 --> 00:29:36,560 Speaker 1: a kid at an arcade machine, turning the wheel repeatedly 595 00:29:36,880 --> 00:29:38,880 Speaker 1: without having a coin in there and pretending that you're 596 00:29:38,880 --> 00:29:41,840 Speaker 1: playing it when the demo's going on. VIB coders are 597 00:29:41,840 --> 00:29:45,200 Speaker 1: also awful for the already negative margins of most AI 598 00:29:45,280 --> 00:29:48,200 Speaker 1: coding environments is every single thing they ask the model 599 00:29:48,240 --> 00:29:50,960 Speaker 1: to do is imprecise burning tokens in pursuit of a 600 00:29:51,000 --> 00:29:54,560 Speaker 1: goal they themselves don't really understand. VIB coding doesn't work, 601 00:29:54,600 --> 00:29:56,640 Speaker 1: It will not work, and pretending otherwise is at best 602 00:29:56,640 --> 00:29:59,480 Speaker 1: ignorance and are worth supporting a campaign built on lies. 603 00:30:15,160 --> 00:30:18,240 Speaker 1: And this is all built up to one final point. 604 00:30:19,000 --> 00:30:22,280 Speaker 1: I'm no longer accepting half baked arguments. If you're an 605 00:30:22,320 --> 00:30:24,640 Speaker 1: AI booster, please come up with better arguments, and if 606 00:30:24,640 --> 00:30:26,760 Speaker 1: you truly believe in this stuff, you should have a 607 00:30:26,800 --> 00:30:29,560 Speaker 1: firmer grasp and why you do so. It's been three 608 00:30:29,680 --> 00:30:31,560 Speaker 1: years and the best some of you have is it's 609 00:30:31,600 --> 00:30:34,400 Speaker 1: really popular or uber also burned money. Your arguments are 610 00:30:34,400 --> 00:30:36,520 Speaker 1: based on what you wish were true rather than what's 611 00:30:36,560 --> 00:30:40,040 Speaker 1: actually true, and it's deeply embarrassing. Then again, there are 612 00:30:40,160 --> 00:30:43,040 Speaker 1: many well intentioned people who aren't necessarily AI boosters who 613 00:30:43,080 --> 00:30:45,680 Speaker 1: repeat these arguments, regardless of how thinly frame they are, 614 00:30:45,920 --> 00:30:47,960 Speaker 1: in part because we live in a high information, low 615 00:30:48,000 --> 00:30:51,080 Speaker 1: processing society where people tend to put great faith in 616 00:30:51,160 --> 00:30:54,560 Speaker 1: people who are confident in what they say and sound smart. Jason, 617 00:30:55,360 --> 00:30:57,280 Speaker 1: I also think the media is failing on a very 618 00:30:57,280 --> 00:30:59,520 Speaker 1: basic level to realize that their fear of missing out 619 00:30:59,600 --> 00:31:02,440 Speaker 1: or seeming stupid is being used against them. If you 620 00:31:02,480 --> 00:31:05,240 Speaker 1: don't understand something, it's likely because the people you're reading 621 00:31:05,480 --> 00:31:08,400 Speaker 1: or hearing it from don't either. If a company takes 622 00:31:08,400 --> 00:31:10,840 Speaker 1: a promise and you don't understand how they'll deliver on it, 623 00:31:10,840 --> 00:31:12,920 Speaker 1: it's their job to explain how, and your job to 624 00:31:12,960 --> 00:31:16,280 Speaker 1: suggest it is implausible and clear and defined language This 625 00:31:16,400 --> 00:31:19,200 Speaker 1: has gone beyond simple objectivity into the realm of an 626 00:31:19,240 --> 00:31:22,680 Speaker 1: outright failure of journalism. I have never seen more and 627 00:31:22,760 --> 00:31:26,640 Speaker 1: misinformation about the capabilities of a product in my entire career, 628 00:31:26,800 --> 00:31:29,240 Speaker 1: and it's largely peddled by reporters who either don't know 629 00:31:29,400 --> 00:31:31,920 Speaker 1: or have no interest in knowing what's actually possible, in 630 00:31:31,960 --> 00:31:34,400 Speaker 1: part because all their peers are doing the same thing 631 00:31:34,480 --> 00:31:38,200 Speaker 1: and saying the same nonsense. As things begin to collapse, 632 00:31:38,360 --> 00:31:40,600 Speaker 1: and they sure look like they're collapsing, but I'm not 633 00:31:40,640 --> 00:31:43,040 Speaker 1: making any wild claims about the bubble bursting quite yet, 634 00:31:43,240 --> 00:31:46,040 Speaker 1: it will look increasingly more deranged to bluntly publish everything 635 00:31:46,080 --> 00:31:48,880 Speaker 1: that these companies say. Never have I seen an act 636 00:31:48,920 --> 00:31:51,680 Speaker 1: of outright contempt more egregious than Sam Altman's saying that 637 00:31:51,720 --> 00:31:54,360 Speaker 1: GPT five was actually bad and that GPT six will 638 00:31:54,400 --> 00:31:57,880 Speaker 1: be even better. Members of the media, Sam Altman does 639 00:31:57,960 --> 00:32:01,160 Speaker 1: not respect you. He is not your friend, Clammy. Sam 640 00:32:01,160 --> 00:32:04,080 Speaker 1: Mortman is not secretly confiding in you. Clammy Will thinks 641 00:32:04,120 --> 00:32:07,560 Speaker 1: you are stupid and easily manipulated, and will print anything 642 00:32:07,640 --> 00:32:10,440 Speaker 1: he says, largely in part because many members of the 643 00:32:10,440 --> 00:32:13,120 Speaker 1: media will print exactly what he says whenever he says him. 644 00:32:13,280 --> 00:32:15,160 Speaker 1: And to be clear, if you wrote about GPT six 645 00:32:15,200 --> 00:32:17,800 Speaker 1: and made fun of it, that's great. But let's close 646 00:32:17,840 --> 00:32:20,680 Speaker 1: by discussing the very nature of AI skepticism and the 647 00:32:20,800 --> 00:32:23,360 Speaker 1: so called void between those who hate AI and those 648 00:32:23,400 --> 00:32:25,960 Speaker 1: who love AI. Am from the perspective of one of 649 00:32:26,000 --> 00:32:29,520 Speaker 1: the most prominent people in the skeptic camp. Critics and 650 00:32:29,560 --> 00:32:32,680 Speaker 1: skeptics are not given the benefit of grace, patience, or 651 00:32:32,720 --> 00:32:35,440 Speaker 1: in many cases, hospitality when it comes to their position. 652 00:32:36,080 --> 00:32:39,000 Speaker 1: While they may receive interviews and opportunities to give their side, 653 00:32:39,040 --> 00:32:41,960 Speaker 1: it's always framed as the work of a firebrand and outlier, 654 00:32:42,360 --> 00:32:45,400 Speaker 1: or somebody with dangerous ideas that they must eternally justify. 655 00:32:46,400 --> 00:32:50,120 Speaker 1: Skeptics are demonized, their points under constant scrutiny, their allegiances 656 00:32:50,160 --> 00:32:53,840 Speaker 1: and intentions constantly interrogated for some sort of moral or 657 00:32:53,840 --> 00:32:57,440 Speaker 1: intellectual weakness. Skeptic and critic are words said with a 658 00:32:57,480 --> 00:33:01,200 Speaker 1: sneer of trepidation that the listener should be suspicious that 659 00:33:01,240 --> 00:33:03,680 Speaker 1: this person isn't agreeing that AI is the most powerful, 660 00:33:03,720 --> 00:33:07,480 Speaker 1: special thing ever. To not immediately fall in love with 661 00:33:07,560 --> 00:33:09,960 Speaker 1: something that everybody is talking about is to be framed 662 00:33:09,960 --> 00:33:12,560 Speaker 1: as a hater. To have oneself, introduced with the words 663 00:33:12,640 --> 00:33:16,360 Speaker 1: not everyone agrees or on forty percent of your appearances. 664 00:33:17,120 --> 00:33:19,640 Speaker 1: By comparison, AI boosters are the first to get TV 665 00:33:19,720 --> 00:33:22,440 Speaker 1: appearances and offers to be on panels. Their coverage featured 666 00:33:22,440 --> 00:33:25,480 Speaker 1: prominently on tech Meme, something slot like books called shit 667 00:33:25,640 --> 00:33:28,720 Speaker 1: like The Future of Intelligence, Masters of the Brain, featuring 668 00:33:28,720 --> 00:33:31,600 Speaker 1: eighteen interviews with different CEOs that all say the same thing. 669 00:33:32,000 --> 00:33:34,280 Speaker 1: They don't have to justify their love. They simply have 670 00:33:34,360 --> 00:33:37,040 Speaker 1: to remember all the right terms chirping out, test time, compute, 671 00:33:37,040 --> 00:33:39,040 Speaker 1: and the cost of inference is going down enough times. 672 00:33:39,040 --> 00:33:41,680 Speaker 1: The Salomon Wario Amiday to give them an hour long 673 00:33:41,720 --> 00:33:44,760 Speaker 1: interview where he says the models they are in years 674 00:33:44,880 --> 00:33:47,600 Speaker 1: going to be the most powerful school teacher ever built. 675 00:33:48,160 --> 00:33:49,600 Speaker 1: And by the way, yeah, I did sell a book 676 00:33:49,640 --> 00:33:53,320 Speaker 1: because my shit fucking bangs, my shit rocks. I'm I'm 677 00:33:53,360 --> 00:33:54,920 Speaker 1: not going to be too smug, but like, I put 678 00:33:54,920 --> 00:33:56,400 Speaker 1: a lot of effort into this and I research it 679 00:33:56,520 --> 00:34:00,959 Speaker 1: very well. Others should try harder. I have persistent, deeply 680 00:34:01,000 --> 00:34:03,640 Speaker 1: sourced arguments that I've built over the course of years. 681 00:34:03,840 --> 00:34:06,560 Speaker 1: I didn't become a hater because I'm a contrarian. I 682 00:34:06,560 --> 00:34:09,080 Speaker 1: became a hater because the shit that these fucking oaths 683 00:34:09,080 --> 00:34:11,919 Speaker 1: have done to the computer pisses me off. I did 684 00:34:11,960 --> 00:34:14,080 Speaker 1: the man that destroyed Google Search because I wanted to 685 00:34:14,080 --> 00:34:17,080 Speaker 1: know why Google Search sucked. I wrote Sam Altman Free 686 00:34:17,120 --> 00:34:19,200 Speaker 1: because at the time I didn't understand why everybody was 687 00:34:19,200 --> 00:34:22,759 Speaker 1: so fucking enamored with this clammy sociopath. Everything I do 688 00:34:22,840 --> 00:34:26,000 Speaker 1: comes from a genuine curiosity and an overwhelming frustration with 689 00:34:26,040 --> 00:34:29,359 Speaker 1: the state of technology. I started writing the newsletter that 690 00:34:29,440 --> 00:34:33,040 Speaker 1: led to this podcast with three hundred subscribers and sixty views, 691 00:34:33,040 --> 00:34:35,279 Speaker 1: and have written it as an exploration of subjects that 692 00:34:35,320 --> 00:34:37,520 Speaker 1: grows as I write. I do not have it in 693 00:34:37,560 --> 00:34:39,480 Speaker 1: me to pretend to be anything other than what I am, 694 00:34:39,520 --> 00:34:41,799 Speaker 1: And if that's strange to you, well I'm a strange man, 695 00:34:41,840 --> 00:34:43,840 Speaker 1: but at least I'm an honest one. I do have 696 00:34:43,880 --> 00:34:45,440 Speaker 1: a chip on my shoulder and that I really do 697 00:34:45,520 --> 00:34:47,560 Speaker 1: not like it when people tried to make other people 698 00:34:47,600 --> 00:34:49,960 Speaker 1: feel stupid, especially when they do so as a means 699 00:34:49,960 --> 00:34:52,480 Speaker 1: of making money for themselves or making someone else look good. 700 00:34:53,160 --> 00:34:55,680 Speaker 1: I write this stuff out because I have an intellectual interest. 701 00:34:55,719 --> 00:34:57,919 Speaker 1: I like writing, and by writing, I'm able to learn 702 00:34:57,920 --> 00:35:01,160 Speaker 1: and process my complex feelings around technology. And talking it 703 00:35:01,200 --> 00:35:04,360 Speaker 1: out actually feels good. It's an intellectual exercise that I 704 00:35:04,400 --> 00:35:07,000 Speaker 1: really enjoy. I happen to do so in a manner 705 00:35:07,040 --> 00:35:09,160 Speaker 1: that hundreds of thousands of people enjoy every month. And 706 00:35:09,160 --> 00:35:11,560 Speaker 1: I'm not specifying where those people go. And if you 707 00:35:11,560 --> 00:35:13,440 Speaker 1: think that I've grown this by being a hater, you 708 00:35:13,480 --> 00:35:16,640 Speaker 1: are doing yourself the service of underestimating me, which I 709 00:35:16,640 --> 00:35:19,280 Speaker 1: will use to my advantage by writing deeper, more meaningful, 710 00:35:19,320 --> 00:35:21,879 Speaker 1: and more insightful things than you, and then I'll say 711 00:35:21,920 --> 00:35:25,000 Speaker 1: them with lots of curse words on this podcast. I've 712 00:35:25,040 --> 00:35:27,520 Speaker 1: watched these pigs ruin the computer again and again and 713 00:35:27,560 --> 00:35:30,320 Speaker 1: make billions doing so. And all of this is happening 714 00:35:30,360 --> 00:35:32,880 Speaker 1: while the media celebrates the destruction of things like Google, 715 00:35:33,000 --> 00:35:36,040 Speaker 1: Facebook and the fucking environment in pursuit of eternal growth. 716 00:35:36,560 --> 00:35:39,400 Speaker 1: I can't manufacture my discussed nor is it hard to 717 00:35:39,760 --> 00:35:42,399 Speaker 1: nor can I manufacture whatever it is incide that makes 718 00:35:42,400 --> 00:35:45,520 Speaker 1: it impossible to keep quiet about these things. I don't 719 00:35:45,520 --> 00:35:47,680 Speaker 1: know if I take this too seriously, whether I don't 720 00:35:47,680 --> 00:35:50,239 Speaker 1: take it seriously enough because they keep saying fucking shit. 721 00:35:50,560 --> 00:35:52,160 Speaker 1: But I'm honored that I'm able to do so, and 722 00:35:52,200 --> 00:35:55,640 Speaker 1: I really appreciate everyone who listens, reads, or engage with 723 00:35:55,920 --> 00:35:58,000 Speaker 1: with me in any way. I really do love you 724 00:35:58,040 --> 00:36:00,480 Speaker 1: all for listening. I know that this was a long 725 00:36:00,600 --> 00:36:04,799 Speaker 1: three parter. I've enjoyed recording it. I've done lots of retakes. Mattasowski, 726 00:36:04,960 --> 00:36:08,120 Speaker 1: love you man. Sorry for all of this. I'll catch 727 00:36:08,120 --> 00:36:19,320 Speaker 1: you next episode. Thank you for listening to Better Offline. 728 00:36:19,480 --> 00:36:21,879 Speaker 1: The editor and composer of the Better Offline theme song 729 00:36:21,960 --> 00:36:24,600 Speaker 1: is Matasowski. You can check out more of his music 730 00:36:24,640 --> 00:36:28,279 Speaker 1: and audio projects at Matasowski dot com, M A T 731 00:36:28,280 --> 00:36:32,759 Speaker 1: T O s O W s ki dot com. You 732 00:36:32,760 --> 00:36:35,280 Speaker 1: can email me at easy at Better Offline dot com 733 00:36:35,400 --> 00:36:37,680 Speaker 1: or visit better Offline dot com to find more podcast 734 00:36:37,760 --> 00:36:41,040 Speaker 1: links and of course, my newsletter. I also really recommend 735 00:36:41,080 --> 00:36:43,040 Speaker 1: you go to chat dot Where's Youreed dot at to 736 00:36:43,120 --> 00:36:45,920 Speaker 1: visit the discord, and go to our slash Better Offline 737 00:36:45,960 --> 00:36:49,160 Speaker 1: to check out our reddit. Thank you so much for listening. 738 00:36:50,000 --> 00:36:52,680 Speaker 1: Better Offline is a production of cool Zone Media. For 739 00:36:52,800 --> 00:36:55,960 Speaker 1: more from cool Zone Media, visit our website cool Zonemedia 740 00:36:56,040 --> 00:36:58,879 Speaker 1: dot com, or check us out on the iHeartRadio app, 741 00:36:58,920 --> 00:37:14,800 Speaker 1: Apple Podcasts, wherever you get your podcasts