WEBVTT - Monologue: The AI Industry Is Lying To You

0:00:03.200 --> 0:00:07.840
<v Speaker 1>Media. Hello, and welcome to this week's Better Offline Monologue.

0:00:07.920 --> 0:00:19.280
<v Speaker 1>I'm your host ed ze Trunk. Now. I'm turning forty

0:00:19.400 --> 0:00:21.800
<v Speaker 1>in a month or so, and at four years young,

0:00:21.920 --> 0:00:24.840
<v Speaker 1>I'm old enough to remember as far back as December eleventh,

0:00:24.840 --> 0:00:27.400
<v Speaker 1>twenty twenty five, when Disney and open Ai reached an

0:00:27.440 --> 0:00:31.040
<v Speaker 1>agreement to bring beloved characters from across Disney's brands to

0:00:31.160 --> 0:00:34.760
<v Speaker 1>sourer open AI's video platform. As part of the deal,

0:00:34.800 --> 0:00:37.479
<v Speaker 1>Disney would and i quote, become a major customer of

0:00:37.520 --> 0:00:40.560
<v Speaker 1>open Ai, use its API to build new products, tools

0:00:40.560 --> 0:00:43.560
<v Speaker 1>and experiences, as well as showing Sora videos in Disney

0:00:43.560 --> 0:00:46.680
<v Speaker 1>Plus and deploy chat GPT for its employees, as well

0:00:46.680 --> 0:00:50.040
<v Speaker 1>as making a one billion dollar equity investment in open Ai.

0:00:50.720 --> 0:00:53.720
<v Speaker 1>Except now, Sora is dead shot in the head by Clammy,

0:00:53.760 --> 0:00:56.680
<v Speaker 1>Sam Altman and open Ai because it cost fifteen million

0:00:56.720 --> 0:00:59.240
<v Speaker 1>dollars or more to run and had everybody making videos

0:00:59.240 --> 0:01:02.800
<v Speaker 1>of Goofy saying gorsh hold fuck anything that moves like

0:01:02.880 --> 0:01:06.040
<v Speaker 1>Frank from Blue Velvet. To be clear, open aye is

0:01:06.040 --> 0:01:08.760
<v Speaker 1>shutting down both Sora the app and access to the

0:01:08.840 --> 0:01:12.720
<v Speaker 1>SAA video model through its API, framing the changes, focusing

0:01:12.760 --> 0:01:15.400
<v Speaker 1>the company on business and enterprise customers, which is a

0:01:15.440 --> 0:01:18.480
<v Speaker 1>euphemism for a product that never had any real business

0:01:18.480 --> 0:01:21.560
<v Speaker 1>model or product market fit other than burning millions of

0:01:21.560 --> 0:01:26.200
<v Speaker 1>dollars of computerday and violating copyright. I think it's also

0:01:26.319 --> 0:01:29.600
<v Speaker 1>completely insane that nobody bothered to check whether Disney had

0:01:29.640 --> 0:01:33.160
<v Speaker 1>ever invested or even talked about investing that billion dollars.

0:01:33.400 --> 0:01:34.800
<v Speaker 1>I went and had a look, and there wasn't a

0:01:34.880 --> 0:01:37.640
<v Speaker 1>damn thing in any of the Disney's most recent quarterly

0:01:37.720 --> 0:01:40.720
<v Speaker 1>or annual reports. I should have checked this myself, but

0:01:40.760 --> 0:01:43.280
<v Speaker 1>it's wild that nobody else did either. There are business

0:01:43.319 --> 0:01:45.840
<v Speaker 1>and tech reporters, there are people that do this full

0:01:46.240 --> 0:01:49.680
<v Speaker 1>I but I also want to take this moment to

0:01:49.680 --> 0:01:54.640
<v Speaker 1>say something for the bottom my heart. I was right.

0:01:54.800 --> 0:01:56.880
<v Speaker 1>I was fucking right. I said it back in October.

0:01:57.000 --> 0:01:59.000
<v Speaker 1>I've said it. I said it back in twenty twenty four.

0:01:59.200 --> 0:02:03.720
<v Speaker 1>SAA was never business. SAURA was never replacing anything. SAURA

0:02:03.920 --> 0:02:08.400
<v Speaker 1>was always unreliable because the hallucination prone, large language model

0:02:08.400 --> 0:02:11.919
<v Speaker 1>bullshit it's built upon. It was never gonna work. Everyone

0:02:11.919 --> 0:02:14.960
<v Speaker 1>who said otherwise is a goddamn rube, and so many

0:02:15.000 --> 0:02:18.480
<v Speaker 1>people were wrong. The app that CNBC said was challenging

0:02:18.480 --> 0:02:21.280
<v Speaker 1>Hollywood and freaking out the movie industry, and The Hollywood

0:02:21.280 --> 0:02:24.560
<v Speaker 1>Reporter would suggest could somehow challenge Pixar and with sam

0:02:24.800 --> 0:02:28.280
<v Speaker 1>when successfully playing Hollywood, and the fucking anklo who should

0:02:28.280 --> 0:02:31.200
<v Speaker 1>be ashamed of themselves said, was Open Ai going to

0:02:31.240 --> 0:02:34.360
<v Speaker 1>war with Hollywood as it shook the industry, and Deadline

0:02:34.600 --> 0:02:37.600
<v Speaker 1>said made Hollywood sow and Boardrooms said was in a

0:02:37.639 --> 0:02:40.720
<v Speaker 1>standoff with Hollywood, and The La Times said was deepening

0:02:40.760 --> 0:02:43.400
<v Speaker 1>a battle between Hollywood and Open AI and igniting a

0:02:43.440 --> 0:02:47.480
<v Speaker 1>firestorm in Hollywood, and Puck said had Hollywood panicking, and

0:02:47.600 --> 0:02:50.200
<v Speaker 1>tech No Lama said was the end of copyright as

0:02:50.240 --> 0:02:52.560
<v Speaker 1>we know it, and Slat said was a case of

0:02:52.600 --> 0:02:56.320
<v Speaker 1>AI crushing Hollywood as it as we've known it, and

0:02:56.320 --> 0:02:57.880
<v Speaker 1>that I swear to God I wrote that right in

0:02:57.919 --> 0:03:00.600
<v Speaker 1>the script. But I'm keeping going anyway. This thing's completely

0:03:00.639 --> 0:03:03.080
<v Speaker 1>dead a little more than five months after everybody claimed

0:03:03.080 --> 0:03:06.680
<v Speaker 1>it was changing everything. You all were wrong, every single

0:03:06.720 --> 0:03:10.240
<v Speaker 1>goddamn wrong, wrong, wrong, wrong, wrong, wrong, on the wrong,

0:03:10.280 --> 0:03:12.639
<v Speaker 1>a sandwich of wrong that you're shoving up your asshole.

0:03:12.680 --> 0:03:16.840
<v Speaker 1>I'm so fucking tired of this. I cannot express how

0:03:16.880 --> 0:03:21.120
<v Speaker 1>angry I am, because it's almost as if everybody making

0:03:21.160 --> 0:03:25.200
<v Speaker 1>these proclamations, was instinctively printing whatever marketing copy had been

0:03:25.240 --> 0:03:29.040
<v Speaker 1>imagined by the AI labs to promote compute intensive vaporware.

0:03:29.200 --> 0:03:31.960
<v Speaker 1>And absolutely nobody is going to apologize to the people

0:03:32.000 --> 0:03:34.560
<v Speaker 1>working in the entertainment industry for scaring the fuck out

0:03:34.600 --> 0:03:37.680
<v Speaker 1>of them with ghost stories for no apparent reason. Every

0:03:37.920 --> 0:03:41.000
<v Speaker 1>single person who blindly repeated that SAURA existed and was

0:03:41.080 --> 0:03:44.400
<v Speaker 1>changing everything should be forced to apologize to their readers.

0:03:45.360 --> 0:03:48.280
<v Speaker 1>I cannot express the sheer amount of panic that spread

0:03:48.280 --> 0:03:51.040
<v Speaker 1>through every single part of the entertainment industry as a

0:03:51.040 --> 0:03:54.600
<v Speaker 1>result of these specious, poorly founded mythologies spread by people

0:03:54.600 --> 0:03:56.840
<v Speaker 1>that didn't give enough of a shit to understand what

0:03:57.040 --> 0:04:00.480
<v Speaker 1>was actually going on. Saura too, was always an act

0:04:00.480 --> 0:04:03.160
<v Speaker 1>of desperation, an attempt to create a marketing cycle to

0:04:03.200 --> 0:04:04.760
<v Speaker 1>prop up a tool that was burning as much as

0:04:04.800 --> 0:04:07.360
<v Speaker 1>fifty million dollars a day. The most of the mainstream

0:04:07.400 --> 0:04:10.720
<v Speaker 1>media bought into because they believe everything Open AI says

0:04:10.880 --> 0:04:13.320
<v Speaker 1>and are willing to extrapolate the destruction of an entire

0:04:13.360 --> 0:04:16.279
<v Speaker 1>industry from a fucking facade driven by a clammy man

0:04:16.320 --> 0:04:19.520
<v Speaker 1>that lies for a living. I am tired of everybody

0:04:19.560 --> 0:04:22.760
<v Speaker 1>scarfing down the slot from these companies. I'm tired of

0:04:22.800 --> 0:04:25.039
<v Speaker 1>the majority of reporters not doing the hard work to

0:04:25.040 --> 0:04:28.800
<v Speaker 1>actually understand what's going on. And good lord, do I

0:04:28.839 --> 0:04:31.960
<v Speaker 1>not have another example? Well? I do, Eh, you get

0:04:31.960 --> 0:04:35.000
<v Speaker 1>what I mean. I spent the last week or so

0:04:35.080 --> 0:04:37.800
<v Speaker 1>trying to work out how many actual data centers got

0:04:37.839 --> 0:04:40.919
<v Speaker 1>built last year and found that despite somewhere between one

0:04:41.000 --> 0:04:44.280
<v Speaker 1>hundred and ninety two hundred and forty gigawats supposed capacity

0:04:44.360 --> 0:04:47.480
<v Speaker 1>being announced, based on my analysis, only three gigawats of

0:04:47.520 --> 0:04:50.120
<v Speaker 1>IT load, which is the GPUs and the associated hardware

0:04:50.120 --> 0:04:53.000
<v Speaker 1>that's being turned on, got built in America in twenty

0:04:53.040 --> 0:04:56.680
<v Speaker 1>twenty five. To explain how bad this number, three gigawatts

0:04:56.720 --> 0:04:59.640
<v Speaker 1>of critical IT hardware is about ninety billion dollars per

0:05:00.240 --> 0:05:02.599
<v Speaker 1>Darling of TD Cohen and in Video sold about one

0:05:02.680 --> 0:05:04.800
<v Speaker 1>hundred and thirty five billion dollars worth of GPUs and

0:05:04.920 --> 0:05:08.960
<v Speaker 1>hardware to America in its last fiscal year. Based on

0:05:09.000 --> 0:05:11.240
<v Speaker 1>the last quarter of sales, which is in the sixty

0:05:11.279 --> 0:05:13.960
<v Speaker 1>two billion dollar range, sixty nine percent of that nice

0:05:14.080 --> 0:05:17.160
<v Speaker 1>is America. It now takes about six months to install

0:05:17.160 --> 0:05:21.520
<v Speaker 1>a single quarters sales worth of GPUs and associated hardware,

0:05:22.400 --> 0:05:24.400
<v Speaker 1>even if you think AI is the biggest and most

0:05:24.440 --> 0:05:27.160
<v Speaker 1>hugest and most special. Boy, what's the fucking point of

0:05:27.200 --> 0:05:30.679
<v Speaker 1>buying these GPUs two to four years in advance, Because

0:05:30.720 --> 0:05:33.599
<v Speaker 1>even if it takes six it takes six months to

0:05:33.640 --> 0:05:36.719
<v Speaker 1>install a single quarters worth of sales. From what I found,

0:05:36.760 --> 0:05:39.680
<v Speaker 1>it's actually taking a lot longer. In some cases. Jensen

0:05:39.760 --> 0:05:42.279
<v Speaker 1>Hwang is announcing a new GPU every year. By the

0:05:42.320 --> 0:05:45.040
<v Speaker 1>time you install your Blackwells, they'll already have Vera Ruben.

0:05:45.240 --> 0:05:47.640
<v Speaker 1>By the time they have Vera Ruben, they'll have something else.

0:05:47.680 --> 0:05:49.840
<v Speaker 1>I don't know, Daphnie from Scooby Doo. They can have

0:05:50.200 --> 0:05:52.680
<v Speaker 1>no These things are the only cool thing that Nvidia

0:05:52.720 --> 0:05:55.880
<v Speaker 1>does is choosing the names of people anyway. And with

0:05:55.920 --> 0:05:59.359
<v Speaker 1>every new generation of these GPUs comes new power requirements

0:05:59.360 --> 0:06:02.320
<v Speaker 1>and rack side While the first generation of Vera Ruben,

0:06:02.360 --> 0:06:04.480
<v Speaker 1>which is the next gen coming out very soon, will

0:06:04.520 --> 0:06:07.559
<v Speaker 1>fit into the current Oberon in video racks and video

0:06:07.640 --> 0:06:11.480
<v Speaker 1>is already teasing it's Kiber racks, which will require even

0:06:11.480 --> 0:06:14.560
<v Speaker 1>more power, even more liquid cooling, which will mean those

0:06:14.880 --> 0:06:19.520
<v Speaker 1>ooh ew ew, those old, stinky, dirty blackwells, You're gonna

0:06:19.520 --> 0:06:21.680
<v Speaker 1>have to throw them right in the trash. No more

0:06:21.720 --> 0:06:25.640
<v Speaker 1>black well for me. Thanks, they've got Verra Rubin in town. Ew.

0:06:25.839 --> 0:06:30.600
<v Speaker 1>You're gonna make me use blackwell? How gross? Ew Jensen

0:06:30.640 --> 0:06:34.320
<v Speaker 1>put those stinky blackwells in the toilet. Ew, you couldn't

0:06:34.320 --> 0:06:37.160
<v Speaker 1>fit in the far too large I've tried yet. Things

0:06:37.240 --> 0:06:39.680
<v Speaker 1>only get worse when you look at the actual construction

0:06:39.800 --> 0:06:43.200
<v Speaker 1>in process per Siteline Climate. Despite over one hundred and

0:06:43.240 --> 0:06:46.839
<v Speaker 1>ninety gigawatts of global capacity that's supposedly in process, with

0:06:46.960 --> 0:06:50.480
<v Speaker 1>sixteen gigwatts that's meant to come online this year, only

0:06:50.640 --> 0:06:54.280
<v Speaker 1>five gigawatts off, it is actually under construction, and under

0:06:54.279 --> 0:06:57.520
<v Speaker 1>constructure can mean everything from a single steel beam to

0:06:57.560 --> 0:07:01.000
<v Speaker 1>a nearly finished buildout. Walk around London some time. You

0:07:01.040 --> 0:07:03.640
<v Speaker 1>should just just look at building projects in London or

0:07:03.680 --> 0:07:06.240
<v Speaker 1>New York. Look at how quickly or not quickly those

0:07:06.279 --> 0:07:10.720
<v Speaker 1>get built. Oh, I don't know. Go look around Stargate

0:07:10.760 --> 0:07:13.400
<v Speaker 1>and all the data centers there. One of the infrastructure

0:07:13.400 --> 0:07:17.000
<v Speaker 1>people from open AI was boasting of the Wisconsin data

0:07:17.000 --> 0:07:19.360
<v Speaker 1>center the other day and he just had a single

0:07:19.400 --> 0:07:21.240
<v Speaker 1>steel beam. I'm not even fucking with you. It's just

0:07:21.280 --> 0:07:23.320
<v Speaker 1>a one beam. And they were like four guys. There's

0:07:23.360 --> 0:07:25.520
<v Speaker 1>three guys working on the beam and another guy walking

0:07:25.600 --> 0:07:29.360
<v Speaker 1>up to it. I love builders, but anyway, in my premium,

0:07:29.520 --> 0:07:32.240
<v Speaker 1>shit in my premium this week, just hit my microphone,

0:07:32.320 --> 0:07:35.480
<v Speaker 1>just going to keep going. I also dug into most

0:07:35.520 --> 0:07:37.720
<v Speaker 1>of the major data center projects that have been announced,

0:07:37.760 --> 0:07:40.040
<v Speaker 1>and then basically every case found that they had either

0:07:40.120 --> 0:07:43.240
<v Speaker 1>barely broken ground, hadn't actually picked the land yet, or

0:07:43.240 --> 0:07:46.160
<v Speaker 1>it stalled entirely, as was the case with fermi supposed

0:07:46.240 --> 0:07:49.200
<v Speaker 1>eleven giga what data center project in Amarillo, Texas, which

0:07:49.280 --> 0:07:53.200
<v Speaker 1>promptly dismissed workers in February after moving forward without permitting

0:07:53.280 --> 0:07:57.400
<v Speaker 1>or funding necessary to actually build the fucking thing. That's

0:07:57.400 --> 0:08:01.480
<v Speaker 1>a public company by the way, for Merick's gone, Oh god,

0:08:02.360 --> 0:08:07.280
<v Speaker 1>it slightly Avengers of assholes. Despite the many, many statements

0:08:07.520 --> 0:08:10.240
<v Speaker 1>the AI is inevitable or unstoppable, the reality is that

0:08:10.280 --> 0:08:12.800
<v Speaker 1>the things aren't getting built, billions of dollars aren't getting sent,

0:08:13.000 --> 0:08:16.360
<v Speaker 1>industries are left undisrupted, and the only people actually benefiting

0:08:16.360 --> 0:08:19.560
<v Speaker 1>are grifters, venture capitalists and banks providing debt to non

0:08:19.560 --> 0:08:23.120
<v Speaker 1>existent data center projects. Look, I believe by the end

0:08:23.160 --> 0:08:25.680
<v Speaker 1>of this era, to quote just Dario one him and

0:08:25.760 --> 0:08:28.680
<v Speaker 1>Kakashi the lone Gunman, these guys are two of the

0:08:28.720 --> 0:08:32.240
<v Speaker 1>best out there and their pseudonymous accounts. This era will

0:08:32.280 --> 0:08:34.920
<v Speaker 1>be remembered as the largest waste of capital in history.

0:08:35.400 --> 0:08:37.800
<v Speaker 1>Most data centers aren't getting built, and those that are

0:08:37.840 --> 0:08:40.320
<v Speaker 1>won't be ready before twenty to twenty seven at best.

0:08:40.920 --> 0:08:43.720
<v Speaker 1>Most AI startups make piddley amounts of revenue and burn

0:08:44.000 --> 0:08:46.480
<v Speaker 1>hundreds of millions of dollars to make single digit millions

0:08:46.480 --> 0:08:49.040
<v Speaker 1>of dollars of revenue, and constantly piss off their users

0:08:49.080 --> 0:08:52.000
<v Speaker 1>by changing rate limits because every single one is subsidized,

0:08:52.200 --> 0:08:54.559
<v Speaker 1>every single one is spending a couple of dollars or

0:08:54.600 --> 0:08:57.000
<v Speaker 1>more per dollar of revenue, and nothing is changing and

0:08:57.040 --> 0:09:00.160
<v Speaker 1>nothing is improving. Fuck off with your whole inferance this

0:09:00.240 --> 0:09:03.360
<v Speaker 1>profitable thing, by the fucking way, to paraphrase my good

0:09:03.400 --> 0:09:05.920
<v Speaker 1>friend Casey Kagawa, who made this point to me over

0:09:06.160 --> 0:09:11.200
<v Speaker 1>Signal fairly recently. Either inference isn't profitable or API cars

0:09:11.240 --> 0:09:14.560
<v Speaker 1>are just a big they're the biggest ripoff ever that

0:09:14.640 --> 0:09:17.600
<v Speaker 1>they could be charging even less than that. Indeed, if

0:09:17.600 --> 0:09:20.880
<v Speaker 1>inference was so profitable, why are they charging a couple

0:09:20.920 --> 0:09:24.400
<v Speaker 1>of bucks for a million dollars worth of input tokens?

0:09:24.400 --> 0:09:27.640
<v Speaker 1>They should be charging pennies. Why are they charging so

0:09:27.720 --> 0:09:31.320
<v Speaker 1>much of it's so profitable take that eighth of dits anyway,

0:09:31.360 --> 0:09:33.080
<v Speaker 1>And for the most part, the media has failed to

0:09:33.120 --> 0:09:35.840
<v Speaker 1>notice all this stuff. I really must be clear how

0:09:36.040 --> 0:09:39.680
<v Speaker 1>shocking this is. I really have read all over the shop.

0:09:39.720 --> 0:09:43.040
<v Speaker 1>I still see people making the Amazon Web Services lost

0:09:43.040 --> 0:09:45.440
<v Speaker 1>a lot of money. Argument, that's about thirty eight billion

0:09:45.520 --> 0:09:48.200
<v Speaker 1>dollars in over a decade that was spent on Amazon

0:09:48.200 --> 0:09:50.880
<v Speaker 1>Web Services. So walk that one bag. Well, yeah, I

0:09:50.880 --> 0:09:53.640
<v Speaker 1>don't know. I still see the same talking points pedaled

0:09:53.720 --> 0:09:56.640
<v Speaker 1>and it's just ridiculous. Big numbers get quoted without a

0:09:56.679 --> 0:10:00.800
<v Speaker 1>second thought. Unrealistic or fantastical timelines are public strad verbatim.

0:10:01.000 --> 0:10:04.800
<v Speaker 1>I'm talking about that. What ten twenty billion dollar nebuis

0:10:04.920 --> 0:10:08.320
<v Speaker 1>deal with meta that just data center doesn't exist, haven't

0:10:08.360 --> 0:10:11.560
<v Speaker 1>even raised the money. And scrutiny appears, by the way,

0:10:12.040 --> 0:10:15.080
<v Speaker 1>to only be reserved for haters or skeptics, of those

0:10:15.080 --> 0:10:18.720
<v Speaker 1>who say fuck too much on a podcast. Well, the

0:10:18.760 --> 0:10:22.360
<v Speaker 1>scrutiny machine over here is just getting started, and you know,

0:10:22.840 --> 0:10:27.719
<v Speaker 1>you just know that I will be absolutely insufferable when

0:10:27.760 --> 0:10:31.920
<v Speaker 1>I'm finally proven rain. Thank you all for listening. Should

0:10:31.960 --> 0:10:34.400
<v Speaker 1>have some fun stuff next week. I'm still working on

0:10:35.160 --> 0:10:42.800
<v Speaker 1>I love you all,