1 00:00:03,160 --> 00:00:08,039 Speaker 1: Media. All right, Matt, I've read the YouTube comments and 2 00:00:08,039 --> 00:00:09,639 Speaker 1: this time I want it so you do not cut 3 00:00:09,720 --> 00:00:12,040 Speaker 1: me off with the music too fast. Okay, good right, 4 00:00:12,039 --> 00:00:15,080 Speaker 1: all right, let's go. This is this week's Better Offline monologue. 5 00:00:15,080 --> 00:00:28,320 Speaker 1: And I'm ed Zich. A lot of you have been 6 00:00:28,360 --> 00:00:30,320 Speaker 1: saying you want me to do something about Sora, and 7 00:00:30,400 --> 00:00:32,240 Speaker 1: if I'm honest, I haven't wanted to because I fund 8 00:00:32,280 --> 00:00:35,000 Speaker 1: the whole thing is so utterly pathetic. A few weeks ago, 9 00:00:35,080 --> 00:00:38,200 Speaker 1: open ai launched a half baked social networking app attached 10 00:00:38,200 --> 00:00:41,520 Speaker 1: to a compute intensive video and audio generator, and people 11 00:00:41,520 --> 00:00:44,600 Speaker 1: immediately began to do two things free count and generate 12 00:00:44,640 --> 00:00:47,879 Speaker 1: as many copyright violations as humanly possible, all because of 13 00:00:47,960 --> 00:00:50,880 Speaker 1: open AI's original plan was to ask copyright holders to 14 00:00:50,920 --> 00:00:53,760 Speaker 1: opt out of having their content presented in these videos. 15 00:00:54,400 --> 00:00:57,720 Speaker 1: Sora spent several days covered in Nazi spongebobs and pickagews 16 00:00:57,720 --> 00:01:00,600 Speaker 1: with guns before multiple Hollywood talent agents. He's, along with 17 00:01:00,640 --> 00:01:04,000 Speaker 1: the estate of Martin Luther King Junior, intervened the complained, 18 00:01:04,080 --> 00:01:07,000 Speaker 1: leading to open ai creating, to quote MPR an opt 19 00:01:07,040 --> 00:01:10,319 Speaker 1: in policy allowing all artists, performers, and individuals the right 20 00:01:10,360 --> 00:01:13,080 Speaker 1: to determine how and whether they can be simulated with 21 00:01:13,200 --> 00:01:15,840 Speaker 1: open AI, blocking the generation of well known characters on 22 00:01:15,840 --> 00:01:18,319 Speaker 1: its public feed and offering to take down material not 23 00:01:18,360 --> 00:01:21,200 Speaker 1: in compliance. It's unclear what happened with nintender, but I 24 00:01:21,200 --> 00:01:24,440 Speaker 1: imagine one of their seventy million lawyers attacked. And now 25 00:01:24,440 --> 00:01:26,440 Speaker 1: we've got that out of the way, let's talk about 26 00:01:26,480 --> 00:01:29,560 Speaker 1: SORA itself. I understand a lot of the people who 27 00:01:29,640 --> 00:01:32,080 Speaker 1: listen in film and TV they're kind of scared. And 28 00:01:32,120 --> 00:01:34,040 Speaker 1: I understand that you've seen a few clips that look 29 00:01:34,200 --> 00:01:36,840 Speaker 1: kind of sort of realistic, and that this, especially if 30 00:01:36,880 --> 00:01:39,480 Speaker 1: you're in the creative arts, is quite terrifying because your 31 00:01:39,520 --> 00:01:41,920 Speaker 1: mind naturally assumes that these clips can be strung together 32 00:01:41,959 --> 00:01:45,160 Speaker 1: into some sort of coherent whole. This isn't the case. 33 00:01:45,720 --> 00:01:48,360 Speaker 1: Every single good, and I use the term loosely, SARA 34 00:01:48,520 --> 00:01:51,760 Speaker 1: video is cherry picked for many, many, many terrible generations. 35 00:01:52,200 --> 00:01:54,560 Speaker 1: Every time you use SORA is random. It doesn't matter 36 00:01:54,560 --> 00:01:56,880 Speaker 1: how specific your prompt is or however many times you've 37 00:01:56,960 --> 00:01:59,960 Speaker 1: used it. SAA is effectively a giant video and audio slot. 38 00:02:00,960 --> 00:02:04,560 Speaker 1: You can never ever guarantee that Sorrow will generate something useful, 39 00:02:04,760 --> 00:02:07,160 Speaker 1: and as a result, can never really budget for using it. 40 00:02:07,600 --> 00:02:11,040 Speaker 1: The human eye is remarkably demanding and little visual inconsistencies 41 00:02:11,080 --> 00:02:14,760 Speaker 1: between scenes will make people feel weird and uncomfortable. Imagine 42 00:02:14,760 --> 00:02:17,079 Speaker 1: that extrapolated to ten or fifteen seconds at a time, 43 00:02:17,080 --> 00:02:19,200 Speaker 1: and how difficult it will be to get something that 44 00:02:19,240 --> 00:02:21,519 Speaker 1: makes visual sense before you have to think about things 45 00:02:21,560 --> 00:02:23,480 Speaker 1: like does this connect to the rest of the footage 46 00:02:23,520 --> 00:02:28,200 Speaker 1: I'm using? Okay, So the majority of actual professionals who 47 00:02:28,240 --> 00:02:30,440 Speaker 1: would use Sura would not be using the app. They'll 48 00:02:30,480 --> 00:02:33,280 Speaker 1: be connecting directly to the model on open ais API. 49 00:02:33,639 --> 00:02:37,960 Speaker 1: It's just it's not done via a classical app interface. Now, 50 00:02:38,360 --> 00:02:40,600 Speaker 1: then there's the problem of cost. This is where you 51 00:02:40,639 --> 00:02:43,600 Speaker 1: really need to start worrying if you're building things with Sourer. 52 00:02:44,200 --> 00:02:47,840 Speaker 1: So let's start off with the first problem. Cost. So 53 00:02:48,240 --> 00:02:51,920 Speaker 1: open ai offers two different Saur models. Sorra two, which 54 00:02:51,919 --> 00:02:54,480 Speaker 1: they say is designed for speed and flexibility and is 55 00:02:54,560 --> 00:02:57,799 Speaker 1: ideal for the exploration phase, and that costs ten cents 56 00:02:57,840 --> 00:02:59,959 Speaker 1: per second, and then there's Sora two pro, which is 57 00:03:00,160 --> 00:03:03,160 Speaker 1: either thirty cents or fifty cents a second depending on resolution, 58 00:03:03,639 --> 00:03:05,239 Speaker 1: and I quote it's the thing you go to for 59 00:03:05,480 --> 00:03:10,160 Speaker 1: production quality outputs. So you're either spending one, three, or 60 00:03:10,200 --> 00:03:12,680 Speaker 1: five dollars for every ten seconds of footage. And like 61 00:03:12,760 --> 00:03:16,000 Speaker 1: every generative model, the longer you generate, the higher the 62 00:03:16,040 --> 00:03:18,560 Speaker 1: likelihood of hallucinations, which in the case of Soro, means 63 00:03:18,560 --> 00:03:22,480 Speaker 1: bizarre animations, inconsistent details, or just flat out useless crap. 64 00:03:23,120 --> 00:03:26,720 Speaker 1: Then there's the problem of time. Open AI's own documentation 65 00:03:26,840 --> 00:03:29,800 Speaker 1: says that a single render may takes several minutes. At 66 00:03:29,800 --> 00:03:31,920 Speaker 1: the end of those several minutes, out pops a video 67 00:03:31,960 --> 00:03:34,440 Speaker 1: that may or may not be of any use. Open 68 00:03:34,480 --> 00:03:37,280 Speaker 1: Ai allows you to remix using more prompts, which allows 69 00:03:37,400 --> 00:03:41,080 Speaker 1: some iterative development, but these remixes also cost money and 70 00:03:41,160 --> 00:03:44,320 Speaker 1: also take several minutes. So let me walk you through 71 00:03:44,360 --> 00:03:47,280 Speaker 1: a scenario. You're making a short film. Let's just say 72 00:03:47,280 --> 00:03:50,240 Speaker 1: it's fifteen minutes long, which is nine hundred seconds. You 73 00:03:50,280 --> 00:03:52,360 Speaker 1: ask Zora to generate a man putting on a hat. 74 00:03:52,480 --> 00:03:55,560 Speaker 1: Your first eight generations each taking four minutes and five 75 00:03:55,600 --> 00:03:58,800 Speaker 1: dollars apiece, which takes about thirty two minutes and forty dollars. 76 00:04:00,000 --> 00:04:01,800 Speaker 1: I don't really do the job, so you do two more, 77 00:04:01,880 --> 00:04:04,920 Speaker 1: taking another four minutes apiece and ten more dollars. You 78 00:04:05,000 --> 00:04:08,160 Speaker 1: finally on the next try get something kind of useful, 79 00:04:08,320 --> 00:04:11,040 Speaker 1: which cost you another five dollars, and then you realize 80 00:04:11,080 --> 00:04:13,280 Speaker 1: you wanted him to wear a specific kind of hat. 81 00:04:13,440 --> 00:04:16,200 Speaker 1: This happens all the time when directing stuff. There are 82 00:04:16,360 --> 00:04:18,800 Speaker 1: minor changes you make that you realize when you're finally 83 00:04:18,839 --> 00:04:22,800 Speaker 1: in the moment, would look or sound or be better. So, yeah, 84 00:04:22,920 --> 00:04:26,680 Speaker 1: that doesn't go so well with probabilistic models. So shit, fuck, 85 00:04:26,720 --> 00:04:29,719 Speaker 1: you gotta do something, so you remix in another four minutes, 86 00:04:29,760 --> 00:04:33,400 Speaker 1: another five dollars. Fuck. Wrong hat, four minutes five dollars. Right, 87 00:04:33,440 --> 00:04:36,479 Speaker 1: hat is hand blends through it for some reason. Okay, 88 00:04:36,600 --> 00:04:39,440 Speaker 1: four minutes, five dollars. The hat's right, but when he 89 00:04:39,480 --> 00:04:41,640 Speaker 1: puts it on, his eye blinks. One of his eyes 90 00:04:41,760 --> 00:04:44,400 Speaker 1: just blinks three times for some reason, so you can't 91 00:04:44,400 --> 00:04:47,599 Speaker 1: really use that. Okay, four minutes five dollars. Looks kind 92 00:04:47,600 --> 00:04:51,560 Speaker 1: of good. Different hat again, four minutes, five dollars. Hmmm, 93 00:04:52,400 --> 00:04:55,160 Speaker 1: you've now spent eighty dollars in over an hour generating 94 00:04:55,160 --> 00:04:56,800 Speaker 1: a man trying to put on a hat. You're not 95 00:04:56,839 --> 00:05:00,640 Speaker 1: really much closer to having useful footage. And because as 96 00:05:00,640 --> 00:05:03,800 Speaker 1: you remix it again and again, keeps making these little errors, 97 00:05:03,800 --> 00:05:06,360 Speaker 1: because that's how these models go, it's impossible to tell 98 00:05:06,400 --> 00:05:08,480 Speaker 1: whether the next generation will be the one that works 99 00:05:08,520 --> 00:05:10,960 Speaker 1: or whether sorrow will spit out some new little fuck up. 100 00:05:12,320 --> 00:05:15,159 Speaker 1: So the more intricate something is, the more expensive it gets. 101 00:05:15,440 --> 00:05:18,200 Speaker 1: But you know what, you can find money places you 102 00:05:18,279 --> 00:05:21,200 Speaker 1: can't find more goddamn time. I guess you could have 103 00:05:21,240 --> 00:05:24,520 Speaker 1: a separate computer running more, but that's still gonna cost 104 00:05:24,520 --> 00:05:26,960 Speaker 1: a bunch of money. How many of these slot machines 105 00:05:27,000 --> 00:05:29,720 Speaker 1: are you gonna run at once? How many times are 106 00:05:29,760 --> 00:05:31,440 Speaker 1: you going to allow them to edit? How can you 107 00:05:31,480 --> 00:05:34,400 Speaker 1: have a coherent vision when you've got multiple people generating things? 108 00:05:34,720 --> 00:05:37,680 Speaker 1: You can't. But you know what, perhaps perhaps the next 109 00:05:37,760 --> 00:05:40,880 Speaker 1: generation will be great, or perhaps it will be dogshit. 110 00:05:41,080 --> 00:05:43,280 Speaker 1: You have no way to know, because that's the magic 111 00:05:43,320 --> 00:05:47,080 Speaker 1: of generative AI. Yet these problems compound aggressively once you 112 00:05:47,120 --> 00:05:50,680 Speaker 1: need any kind of visual consistency. The man now has 113 00:05:50,720 --> 00:05:52,440 Speaker 1: to put the hat on and leave the house. How 114 00:05:52,480 --> 00:05:54,840 Speaker 1: does the house look? Is the hat the same? Does 115 00:05:54,880 --> 00:05:57,360 Speaker 1: he have wallpaper on his walls? Is there anyone else 116 00:05:57,400 --> 00:05:59,880 Speaker 1: in the house? What kind of table? Two chairs, one chair, 117 00:06:00,080 --> 00:06:02,560 Speaker 1: five chairs? How do you possibly keep all of these 118 00:06:02,600 --> 00:06:06,680 Speaker 1: things consistent? You don't, You can't. That's part of what 119 00:06:06,800 --> 00:06:10,760 Speaker 1: makes SAURA so goddamn awful. It's built specifically to make 120 00:06:10,839 --> 00:06:14,400 Speaker 1: you scared of them, to create superficially impressive clips, so 121 00:06:14,400 --> 00:06:17,280 Speaker 1: that brain dead Hollywood executives can claim it's the future. 122 00:06:17,360 --> 00:06:19,800 Speaker 1: Yet in a practical sense, it's impossible to budget, or 123 00:06:19,880 --> 00:06:22,919 Speaker 1: plan or guarantee anything about what SAURA might do. And 124 00:06:23,000 --> 00:06:26,120 Speaker 1: this is pretty much across the board for these generative 125 00:06:26,160 --> 00:06:30,080 Speaker 1: models making video and audio. Now, I've heard from a 126 00:06:30,080 --> 00:06:33,000 Speaker 1: few people that SAA is cheaper because it doesn't involve labor, 127 00:06:33,160 --> 00:06:35,200 Speaker 1: which is something you could say only if you believed 128 00:06:35,200 --> 00:06:38,520 Speaker 1: SAURA would give consistent outputs. And really, the only thing 129 00:06:38,560 --> 00:06:42,479 Speaker 1: that a probabilistic model like SAURA can do is guarantee inconsistency, 130 00:06:43,080 --> 00:06:46,240 Speaker 1: even by Hollywood accounting standards. A generative tool that will 131 00:06:46,279 --> 00:06:49,279 Speaker 1: cost hundreds or thousands of dollars to generate ten seconds 132 00:06:49,279 --> 00:06:52,640 Speaker 1: of shitty footage that is impossible to coherently connect to 133 00:06:52,640 --> 00:06:56,320 Speaker 1: more footage is a really terrible idea and also very 134 00:06:56,800 --> 00:07:00,320 Speaker 1: inconsistent in its costs too. And like I said earlier, 135 00:07:00,320 --> 00:07:03,919 Speaker 1: there's the issue of time. Every single entertainment product requires 136 00:07:03,920 --> 00:07:06,360 Speaker 1: some sort of time budgeting, and it's impossible to say 137 00:07:06,400 --> 00:07:09,120 Speaker 1: how long it will take SAURAW to generate something. Open 138 00:07:09,120 --> 00:07:12,120 Speaker 1: Eye doesn't even specify what several minutes means, meaning you 139 00:07:12,120 --> 00:07:15,560 Speaker 1: can't really plan a production using it. SARA isn't cheaper, 140 00:07:15,640 --> 00:07:19,960 Speaker 1: SAWRA isn't easier, and SARA certainly isn't more efficient. But 141 00:07:20,720 --> 00:07:23,120 Speaker 1: you need to remember also that generative video models have 142 00:07:23,200 --> 00:07:25,920 Speaker 1: been around for over a year and they're not really 143 00:07:26,000 --> 00:07:29,800 Speaker 1: seeing mass use now. If this thing were capable of 144 00:07:29,800 --> 00:07:32,920 Speaker 1: making anything truly useful, you'd see it everywhere right now. 145 00:07:33,040 --> 00:07:34,520 Speaker 1: But you are seeing a little bit of it. And 146 00:07:34,560 --> 00:07:37,280 Speaker 1: I do want to address that you probably saw cal 147 00:07:37,320 --> 00:07:39,680 Speaker 1: She's ad and heard that it costs two thousand dollars 148 00:07:39,720 --> 00:07:41,280 Speaker 1: to make and took only a few days, But I 149 00:07:41,320 --> 00:07:43,760 Speaker 1: really encourage you to look at the actual commercial itself. 150 00:07:44,080 --> 00:07:48,080 Speaker 1: It's completely incoherent nonsense. Each shot completely disconnected with weird 151 00:07:48,120 --> 00:07:51,400 Speaker 1: glitches and animations in the crowds, and one point towards 152 00:07:51,400 --> 00:07:53,480 Speaker 1: the end, a woman is meant to say okay, see, 153 00:07:53,480 --> 00:07:56,000 Speaker 1: but the sea part does not map to her mouth. 154 00:07:56,400 --> 00:07:58,680 Speaker 1: It looks really bad and the only way you could 155 00:07:58,720 --> 00:08:00,960 Speaker 1: get away with something like this is having these quick 156 00:08:01,040 --> 00:08:04,280 Speaker 1: hit shots. And also please go and view the comments 157 00:08:04,280 --> 00:08:06,440 Speaker 1: about this that people just rip the fuck out of 158 00:08:06,440 --> 00:08:09,320 Speaker 1: this thing. But nevertheless, it was made using VO three, 159 00:08:09,360 --> 00:08:12,520 Speaker 1: Google's generative video model, and it apparently took three hundred 160 00:08:12,520 --> 00:08:15,600 Speaker 1: to four hundred clips to get fifteen usable shots stitched 161 00:08:15,600 --> 00:08:19,320 Speaker 1: together using traditional editing tools. Now, the reason this costs 162 00:08:19,320 --> 00:08:21,080 Speaker 1: two grand is that it sucked. And the reason you're 163 00:08:21,120 --> 00:08:23,800 Speaker 1: not seeing more advertisers do this is because it's impossible 164 00:08:23,800 --> 00:08:26,640 Speaker 1: to make a coherent video out of this footage. I 165 00:08:26,720 --> 00:08:29,880 Speaker 1: realize most commercials you see on TV may feel chaotic 166 00:08:30,000 --> 00:08:32,680 Speaker 1: or kind of bland, but they're remarkably precise, and the 167 00:08:32,720 --> 00:08:35,800 Speaker 1: generative shots used for the Cawshi commercial are chaotic and 168 00:08:35,840 --> 00:08:38,960 Speaker 1: failed to convey any real meaning beyond a person yelling 169 00:08:39,000 --> 00:08:42,600 Speaker 1: Indiana or OKC. The only reason it cost so little 170 00:08:42,720 --> 00:08:45,199 Speaker 1: was one guy put several days of prompting it to 171 00:08:45,360 --> 00:08:48,520 Speaker 1: it and the end result was shitting cow. She didn't 172 00:08:48,559 --> 00:08:51,400 Speaker 1: mind because this was a publicity move. Cow. She put 173 00:08:51,440 --> 00:08:54,040 Speaker 1: out the commercials specifically so the media would write it up, 174 00:08:54,040 --> 00:08:56,360 Speaker 1: and they succeeded because the media loves to feed on 175 00:08:56,440 --> 00:08:59,480 Speaker 1: scary stories like AI is going to replace human actors. 176 00:09:00,320 --> 00:09:03,360 Speaker 1: Since the calshe ads pjas who made it has made 177 00:09:03,360 --> 00:09:06,040 Speaker 1: a few others a Popeyes wrap one where again go 178 00:09:06,120 --> 00:09:07,560 Speaker 1: and look at the comments. I'm not linking to it, 179 00:09:07,600 --> 00:09:08,760 Speaker 1: by the way, I don't want to send them any 180 00:09:08,760 --> 00:09:10,880 Speaker 1: fucking traffic. But the Pope is one. People are just 181 00:09:10,880 --> 00:09:14,240 Speaker 1: responding saying, this looks like shit, what is this? It's incoherent, 182 00:09:14,240 --> 00:09:18,240 Speaker 1: it's inconsistent. But the funniest one I found was David 183 00:09:18,240 --> 00:09:21,760 Speaker 1: Beckham's iomate health supplement ad, which ends with a shot 184 00:09:21,800 --> 00:09:23,320 Speaker 1: of the bottle of the product with a bunch of 185 00:09:23,360 --> 00:09:27,000 Speaker 1: garbled generative texts. It does not appear that PJAS has 186 00:09:27,080 --> 00:09:29,720 Speaker 1: got a ton more work than this, probably because the 187 00:09:29,760 --> 00:09:32,320 Speaker 1: outputs kind of suck and brands really do not like 188 00:09:32,360 --> 00:09:37,600 Speaker 1: inconsistent things. And also a fucking health supplement from David Beckham. 189 00:09:37,679 --> 00:09:40,800 Speaker 1: Jesus Christ just say it's a private equity film anyway. 190 00:09:41,360 --> 00:09:43,400 Speaker 1: To conclude, I also want to be clear that the 191 00:09:43,480 --> 00:09:46,120 Speaker 1: rates for these videos are heavily subsidized by big tech, 192 00:09:46,280 --> 00:09:49,240 Speaker 1: just like every other generative AI product or saw a 193 00:09:49,360 --> 00:09:52,040 Speaker 1: might cost thirty or fifty cents a second right now. 194 00:09:52,360 --> 00:09:54,680 Speaker 1: Once the AI bubble burst, these prices who will either 195 00:09:54,720 --> 00:09:58,040 Speaker 1: skyrocket or these models will cease to exist for public consumption. 196 00:09:58,679 --> 00:10:00,680 Speaker 1: The biggest clue I can give you is Google only 197 00:10:00,720 --> 00:10:03,480 Speaker 1: allows you to generate four or five VO three videos 198 00:10:03,520 --> 00:10:05,319 Speaker 1: a day on their two hundred and fifty dollars a 199 00:10:05,400 --> 00:10:09,319 Speaker 1: month Gemini ultra plan. That suggests that Google's video costs 200 00:10:09,320 --> 00:10:11,320 Speaker 1: a brutal and the open aiye is burning money by 201 00:10:11,320 --> 00:10:13,000 Speaker 1: the bucket for to let you fuck around on the 202 00:10:13,040 --> 00:10:15,559 Speaker 1: sau app. I don't recommend you do that, but if 203 00:10:15,600 --> 00:10:17,560 Speaker 1: you have just no, you're burning a hole in Clammy 204 00:10:17,559 --> 00:10:20,320 Speaker 1: Sammy's pocket. I will add that you may worry about 205 00:10:20,320 --> 00:10:23,120 Speaker 1: these models getting better. While they might be more nuanced 206 00:10:23,160 --> 00:10:25,439 Speaker 1: than their ability to generate video in five or ten 207 00:10:25,520 --> 00:10:28,520 Speaker 1: second bursts, their ability to generate longer or consistent videos 208 00:10:28,720 --> 00:10:32,080 Speaker 1: is inherently impossible due to the probabilistic nature of transformer 209 00:10:32,120 --> 00:10:35,240 Speaker 1: based models. In simple terms, these things are rolling the 210 00:10:35,280 --> 00:10:38,199 Speaker 1: dice every time. The way you prompt them is what 211 00:10:38,280 --> 00:10:41,520 Speaker 1: makes them generate, and they don't have minds or thoughts. 212 00:10:41,920 --> 00:10:45,480 Speaker 1: They're just rolling the dice every time on whatever you 213 00:10:45,520 --> 00:10:48,679 Speaker 1: say and trying to interpret what you mean. Human beings, 214 00:10:48,720 --> 00:10:51,920 Speaker 1: by the way, are extremely magical. I think you really 215 00:10:52,040 --> 00:10:55,760 Speaker 1: underestimate how amazing people are. When we direct someone on 216 00:10:55,800 --> 00:10:59,280 Speaker 1: a film set, even like an assistant director. That person 217 00:10:59,360 --> 00:11:01,719 Speaker 1: keeps the product moving and make sure everyone gets what 218 00:11:01,760 --> 00:11:03,760 Speaker 1: they need and pushes back in a director when something 219 00:11:03,800 --> 00:11:06,480 Speaker 1: might be impractical. A director is a visionary, but also 220 00:11:06,679 --> 00:11:10,520 Speaker 1: an actor is someone that takes interpretation and then is 221 00:11:10,600 --> 00:11:13,320 Speaker 1: directed to do different things. But that direction is not 222 00:11:13,360 --> 00:11:16,960 Speaker 1: a fucking prompt move your elbow, look look at this way, 223 00:11:17,040 --> 00:11:19,840 Speaker 1: look that way. The things that operate on a film 224 00:11:19,920 --> 00:11:23,280 Speaker 1: or TV set are inherently different to just plugging words 225 00:11:23,320 --> 00:11:27,320 Speaker 1: into a fucking model, and I get them. I get 226 00:11:27,360 --> 00:11:30,000 Speaker 1: everyone in Hollywood who's scared right now. I get everyone 227 00:11:30,040 --> 00:11:33,440 Speaker 1: in creatives, in creative arts even who is scared right now. 228 00:11:33,760 --> 00:11:37,560 Speaker 1: I feel for you. These people are losing. These people 229 00:11:37,720 --> 00:11:42,480 Speaker 1: are losing. This stuff does not work, it's inconsistent, it's 230 00:11:42,520 --> 00:11:46,200 Speaker 1: incredibly expensive on subsidized rates, and in the end, I 231 00:11:46,320 --> 00:11:49,280 Speaker 1: really really believe that once the bubble pops, these things 232 00:11:49,320 --> 00:11:52,800 Speaker 1: are going away. Thank you so much for listening. Reach 233 00:11:52,840 --> 00:11:54,719 Speaker 1: out if you have any thoughts. I always love to 234 00:11:54,760 --> 00:11:58,360 Speaker 1: hear from people. E Z at better offline dot com. 235 00:11:58,600 --> 00:12:01,280 Speaker 1: I love getting your emails. I love getting your your 236 00:12:01,280 --> 00:12:04,960 Speaker 1: weird little missives on Reddit. I really am I'm truly blessed, 237 00:12:05,720 --> 00:12:07,800 Speaker 1: and I love you all. I love how many of 238 00:12:07,800 --> 00:12:10,600 Speaker 1: you listener. I love how communicative you are. It's been 239 00:12:10,600 --> 00:12:14,319 Speaker 1: a big week with the Anthropic exclusive, and yeah, I'm 240 00:12:14,320 --> 00:12:17,040 Speaker 1: gonna have already a better offline next week as well. Crap, 241 00:12:17,040 --> 00:12:21,040 Speaker 1: I've got a good do an episode. Shit damn. Oh well, 242 00:12:21,080 --> 00:12:22,719 Speaker 1: I have the best job in the world anyway, Thank 243 00:12:22,760 --> 00:12:23,360 Speaker 1: you for listening.