WEBVTT - Week in Tech: Can a Woolly Mouse Make a Mammoth? 0:00:00.280 --> 0:00:03.240 Welcome to Tech Stuff, a production of iHeart Podcasts and 0:00:03.279 --> 0:00:06.840 Kaleidoscope IMA's Veloscian, and today will bring you the headlines 0:00:06.840 --> 0:00:11.200 of the week, including a genetically edited rodent, the Wally Mouse. Then, 0:00:11.280 --> 0:00:13.760 on today's Tech Supports segment, we'll talk to four of 0:00:13.840 --> 0:00:16.960 Form Media's Jason Kebler about what the future of AI 0:00:17.079 --> 0:00:22.360 movies could look like. All of that on the weekend 0:00:22.400 --> 0:00:30.520 Tech Is Friday. It's March seventh. I'm excited to be 0:00:30.560 --> 0:00:33.240 back in the studio this week with our producer Eliza Dennis. 0:00:33.280 --> 0:00:34.640 We're glad to have you Stateside. 0:00:34.880 --> 0:00:36.480 Yes, it's felt like I was away for a long time. 0:00:36.600 --> 0:00:38.440 I'm wondering if that had something to do with this 0:00:38.560 --> 0:00:39.440 news cycle though. 0:00:39.720 --> 0:00:43.120 Yeah, there's a lot, lot lot to cover, so should 0:00:43.120 --> 0:00:43.639 we jump in. 0:00:43.800 --> 0:00:45.000 Yeah? Absolutely. So. 0:00:45.120 --> 0:00:46.879 It was a bit of a confusing week when it 0:00:46.920 --> 0:00:50.440 comes to chips and semiconductors, and I'll come back to 0:00:50.479 --> 0:00:54.120 why it was confusing. But Monday saw President Trump hold 0:00:54.160 --> 0:01:01.120 a press conference with the Taiwan Semiconductor Manufacturing Company aka TSMC. 0:01:01.640 --> 0:01:05.800 The clues in the name the company manufacture semiconductors and 0:01:06.120 --> 0:01:10.959 they produce ninety percent of the world's super advanced semiconductor chips. 0:01:11.000 --> 0:01:14.039 These are the chips that power AI training models, but 0:01:14.120 --> 0:01:17.679 also devices and basically are the backbone of the new 0:01:17.800 --> 0:01:22.000 global economy. However, the vast majority of the manufacturing takes 0:01:22.040 --> 0:01:26.240 place in Taiwan, and so many in Washington and beyond 0:01:26.280 --> 0:01:29.559 have worn that TSMC's dominance in the chip industry could 0:01:29.560 --> 0:01:33.720 create a national security risk, given that Taiwan is squarely 0:01:33.880 --> 0:01:38.120 in the bullseye of China's territorial ambitions. But this week, 0:01:38.319 --> 0:01:42.240 the Taiwanese company pledged to invest one hundred billion dollars 0:01:42.280 --> 0:01:44.360 in manufacturing chips on US soil. 0:01:44.760 --> 0:01:46.600 You know, this is so interesting to me because it 0:01:46.640 --> 0:01:50.160 comes after multiple announcements over the last couple of months 0:01:50.200 --> 0:01:54.240 about investments in things like data centers and AI infrastructure. 0:01:54.440 --> 0:01:57.720 And that was with Stargate, and then Apple actually recently 0:01:57.720 --> 0:02:01.040 made a pledge to make more products domestically with themestic contractors. 0:02:01.240 --> 0:02:03.520 Yes, I think they talked about five hundred billion dollars. 0:02:03.640 --> 0:02:06.559 But what was really interesting was that as soon as Tuesday, 0:02:06.720 --> 0:02:10.919 when Trump addressed Congress, he talked about his aggressive desire 0:02:11.000 --> 0:02:14.880 to dismantle the Act that actually TSMC is using in 0:02:14.960 --> 0:02:19.560 part to fund its semiconductor manufacturing in the US. The 0:02:19.680 --> 0:02:23.880 Chips Act was biden error legislation that basically created a 0:02:23.919 --> 0:02:29.320 platform for manufacturing semiconductor chips in the US. I don't know, 0:02:29.400 --> 0:02:31.080 I don't know how to square those two things, but 0:02:31.160 --> 0:02:33.560 that actually brings us to our next headline, which is 0:02:33.760 --> 0:02:39.160 a breakthrough indirectly interpreting and reading brain waves and converting 0:02:39.200 --> 0:02:39.680 them to text. 0:02:40.000 --> 0:02:41.480 The superpower I want. 0:02:41.639 --> 0:02:43.880 Yes, exactly what. You may be able to buy it 0:02:44.080 --> 0:02:46.040 if met had anything to do with it, because they 0:02:46.120 --> 0:02:49.800 announced that in partnership with the Basque Center on Cognition, 0:02:50.000 --> 0:02:53.960 Brain and Language in Spain, researchers have been able to 0:02:54.000 --> 0:02:59.960 decode unspoken language, often reconstructing full sentences directly from brainwaves 0:03:00.480 --> 0:03:03.600 and not even requiring any surgical intervention. This is all 0:03:03.639 --> 0:03:05.600 stuff which can be measured outside the head. 0:03:05.800 --> 0:03:08.840 Yeah, and that's really the breakthrough here, right, because other 0:03:08.919 --> 0:03:14.400 research from companies like Neurolink have been extremely invasive, you know, 0:03:14.520 --> 0:03:17.840 electrodes being implanted into the brain. Invasive. 0:03:18.320 --> 0:03:20.639 Yeah, that's right. And this research is all about kind 0:03:20.680 --> 0:03:24.000 of putting monitors on the skull or around the head 0:03:24.280 --> 0:03:26.520 to be able to read brain waves without having to 0:03:26.800 --> 0:03:30.040 directly hook into the brain, which is obviously much less scary, 0:03:30.360 --> 0:03:33.760 and there's an amazing promise for people with cognitive impairments or 0:03:33.760 --> 0:03:37.560 brain injuries to be able to convert their thoughts into 0:03:37.800 --> 0:03:41.280 text and therefore speech. But there are also some concerns. 0:03:41.360 --> 0:03:45.560 Right the Vox headline was Meta's brain to text tech 0:03:45.760 --> 0:03:49.040 is here. We are not remotely ready, And of course 0:03:49.040 --> 0:03:52.800 the big concern here is privacy if private companies can 0:03:52.840 --> 0:03:55.520 actually read our thoughts. But there's actually a long way 0:03:55.520 --> 0:03:58.600 to go before this research leaves the lab. Nonetheless, the 0:03:58.680 --> 0:04:02.320 experiment was kind of a maze. So thirty five volunteers 0:04:02.400 --> 0:04:06.640 sat under magnetic brain imaging scanners and typed on a keyboard. 0:04:07.040 --> 0:04:10.000 Based on prior training, an AI model was able to 0:04:10.040 --> 0:04:13.640 predict what they were writing, and meture research is accurately 0:04:13.680 --> 0:04:17.080 decoded between seventy and eighty percent of what people typed. 0:04:17.120 --> 0:04:20.479 In other words, with seventy to eighty percent certainty, it 0:04:20.560 --> 0:04:23.280 could know before I clicked a T that I was 0:04:23.279 --> 0:04:25.960 about to click the T. And so the real promise 0:04:26.040 --> 0:04:29.880 here is actually a data from this research is beginning 0:04:29.920 --> 0:04:34.799 to give neuroscientists a path to understanding how abstract thoughts 0:04:35.120 --> 0:04:37.559 are converted into language by the human brain. 0:04:38.080 --> 0:04:39.840 Then I think the other part of this is that 0:04:39.880 --> 0:04:42.680 we're getting closer and closer to this idea that we 0:04:42.720 --> 0:04:45.120 can have wearables that do this kind. 0:04:45.000 --> 0:04:48.000 Of tech totally. But of course, a wearable headset that 0:04:48.080 --> 0:04:50.640 can can actually read your thoughts and translate them into 0:04:50.720 --> 0:04:54.960 language is something that you know, conceivably could change a 0:04:55.000 --> 0:04:58.279 lot of people's lives. In another kind of science fiction 0:04:58.480 --> 0:05:02.320 becomes science fact story, it's about the wooly mammoths. The 0:05:02.360 --> 0:05:07.039 headline from MPR was just irresistible hoping to revive mammoths, 0:05:07.360 --> 0:05:12.560 scientists create wooly mice. Yeah, and I think one of 0:05:12.560 --> 0:05:14.200 the scientists that we knew we could do it, but 0:05:14.400 --> 0:05:17.720 we didn't know they would be this cute and they're 0:05:17.720 --> 0:05:20.040 worth a look. But the story is about a company 0:05:20.080 --> 0:05:23.880 called Colossal Biosciences, and they are, by their own account, 0:05:24.120 --> 0:05:26.880 the first and only de extinction company. 0:05:27.360 --> 0:05:29.520 Okay, this was a concept I had never heard of 0:05:29.600 --> 0:05:30.400 until this week. 0:05:30.720 --> 0:05:33.280 Yeah, this one's been one that's I've been intrigued by 0:05:33.320 --> 0:05:34.600 for it for a long time, and I hope we'll 0:05:34.600 --> 0:05:36.000 be able to cover it on an episode of the 0:05:36.040 --> 0:05:39.560 story before too long. But Colossal's website points out that 0:05:39.680 --> 0:05:42.880 nine hundred and two species are extinct and more than 0:05:42.960 --> 0:05:47.520 nine two hundred are critically endangered, and their mission is 0:05:47.560 --> 0:05:52.240 to restore extinct species to preserve biodiversity. It's a little controversial. 0:05:52.440 --> 0:05:54.960 Some people think there are more efficient ways to do 0:05:55.040 --> 0:06:00.400 conservation than reviving extinct species, you know, But to that 0:06:00.480 --> 0:06:02.479 I would say, I mean, look at the wooly mouse. 0:06:02.520 --> 0:06:05.240 This is whether or not you think this is the 0:06:05.240 --> 0:06:09.880 most efficient investment. It is absolutely wild. So picture a 0:06:09.920 --> 0:06:13.599 mouse with fluffy, orange tan fur that looks like it 0:06:13.640 --> 0:06:15.440 got very wet and then got a blow dry at 0:06:15.440 --> 0:06:17.360 the salon. You've got the picture. 0:06:18.000 --> 0:06:19.800 They are extremely cute, and. 0:06:19.920 --> 0:06:23.160 The way Colossal made them was first studying the wooly 0:06:23.200 --> 0:06:28.440 mammoth genome and then genetically engineering mice by modifying seven 0:06:28.520 --> 0:06:32.039 key genes to make them more like wooly mammoths. You know, 0:06:32.120 --> 0:06:35.440 the wool obviously being the most visible element, but also 0:06:35.480 --> 0:06:38.359 some things that were invisible, like the way the mice 0:06:38.400 --> 0:06:40.920 store fat and their fat metabolism makes them much more 0:06:40.920 --> 0:06:44.080 able to survive in the cold. And according to Colossal, 0:06:44.560 --> 0:06:49.839 the plan is to implant wooly mammoth esque modified embryos 0:06:50.040 --> 0:06:54.040 to Asian elephants by twenty twenty eight. This week, was 0:06:54.040 --> 0:06:58.040 also the Oscars and we both saw the movie that 0:06:58.120 --> 0:06:59.760 won Best Live Action. 0:07:00.640 --> 0:07:02.840 Please tell people about it. It's wonderful. 0:07:03.320 --> 0:07:07.000 So it's a Belgian Dutch copro called I'm Not a Robot? 0:07:07.120 --> 0:07:07.880 What did you make of it? 0:07:08.320 --> 0:07:10.480 I was extremely tickled by this promise. 0:07:11.200 --> 0:07:13.040 So, for those who haven't seen it, the film was 0:07:13.040 --> 0:07:16.720 written and directed by Victoria Wamadam and it's about a 0:07:16.760 --> 0:07:21.480 music producer who fails a series of capture tests and 0:07:22.000 --> 0:07:25.400 in so doing us to question whether she's in fact human. 0:07:26.320 --> 0:07:28.120 I mean the minute I knew that we were having 0:07:28.160 --> 0:07:30.840 a capture test as part of the plot to this movie, 0:07:30.920 --> 0:07:33.080 I was all in. I don't know if you have 0:07:33.160 --> 0:07:37.040 this feeling, but I hate failing captured tests, especially when 0:07:37.040 --> 0:07:40.160 you have to click I'm not a robot and all 0:07:40.200 --> 0:07:42.840 you have to do is choose squares that show images 0:07:42.920 --> 0:07:46.320 of street lights or motorcycles or bikes. How can I 0:07:46.360 --> 0:07:46.960 get that wrong? 0:07:47.240 --> 0:07:49.800 Yeah? So she's failing the tests again and again, even 0:07:49.800 --> 0:07:51.920 though it looks like she's doing it right. And then 0:07:51.920 --> 0:07:55.120 she gets to pop up with another quiz and one 0:07:55.120 --> 0:07:58.120 of the questions is did your parents die before you 0:07:58.160 --> 0:08:02.480 met them? And she answers, she answers, yes, and I 0:08:02.480 --> 0:08:04.760 don't want to spoil the whole plot. It gets pretty eerie, 0:08:05.120 --> 0:08:07.040 but it's a fascinating film well worth a watch. You 0:08:07.080 --> 0:08:08.920 can check it out actually on the New Yorker website 0:08:08.920 --> 0:08:12.040 because they were involved in releasing the film and on YouTube, 0:08:12.280 --> 0:08:15.160 and as a tech nerd, I was rooting for them 0:08:15.200 --> 0:08:17.320 to win the Best Live Action Short and did. 0:08:18.040 --> 0:08:18.280 Yes. 0:08:18.680 --> 0:08:22.000 Congratulations team. I'm not a robot, So. 0:08:22.040 --> 0:08:24.920 Stick around as well after the break for a look 0:08:24.920 --> 0:08:28.720 at how AI was used in this year's OSCAR nominated 0:08:28.760 --> 0:08:32.320 feature films, including The Brutalist, and for a conversation with 0:08:32.400 --> 0:08:35.160 Jason Kebler about what it's like to attend an AI 0:08:35.320 --> 0:08:45.640 film festival. Stay with us, Welcome back. The Oscars were 0:08:45.640 --> 0:08:48.600 on Sunday, so we're going to stick with movies. Back 0:08:48.600 --> 0:08:51.839 in twenty twenty three, the Hollywood Writers' Strike was this 0:08:52.080 --> 0:08:57.000 fascinating early example of a very public negotiation over how 0:08:57.160 --> 0:09:02.760 AI might could, and even would disrupt and displace human labour. Ultimately, 0:09:02.760 --> 0:09:05.280 the Writers Guild of America signed an agreement with the 0:09:05.280 --> 0:09:09.439 Alliance of Motion Picture and Television Producers that Generative AI 0:09:09.559 --> 0:09:13.240 would not reduce or eliminate writers and their pay. But 0:09:13.320 --> 0:09:15.920 this was not a commitment by the industry not to 0:09:16.040 --> 0:09:19.920 use generative AI in filmmaking, far from it. In fact, 0:09:19.920 --> 0:09:23.600 this January, the editor of the Triple Oscar winning movie 0:09:23.679 --> 0:09:27.240 The Brutalist told an industry publication that he had used 0:09:27.240 --> 0:09:31.040 generative AI a few times in post production. Some of 0:09:31.040 --> 0:09:35.160 the actors in The Brutalist, namely Felicity Jones and Adrian Brody, 0:09:35.640 --> 0:09:39.240 performed their roles with a heavy Hungarian accent, and they 0:09:39.240 --> 0:09:42.920 even had some dialogue in Hungarian. To prepare for the roles, 0:09:43.000 --> 0:09:45.839 Brody and Jones spent months with a dialect coach to 0:09:45.880 --> 0:09:50.480 perfect their accents, but as The Brutalist editor David Joncho, 0:09:50.920 --> 0:09:54.720 a native Hungarian speaker, pointed out, English speakers can have 0:09:54.720 --> 0:09:58.880 a hard time pronouncing certain sounds. In post he tried 0:09:58.920 --> 0:10:01.760 to perfect the Hungarian in dialogue, and first the team 0:10:01.800 --> 0:10:03.600 had the actor's reader of the lines in the studio. 0:10:04.080 --> 0:10:06.840 Then they tried having other actors say the lines, but 0:10:07.080 --> 0:10:10.560 that also didn't sound right, so Yoncho turned to AI. 0:10:11.200 --> 0:10:15.079 He fared Brody and Jones's voices into the program respeecher 0:10:15.600 --> 0:10:19.199 and then using his own voice, Yoncho refined certain vowels 0:10:19.200 --> 0:10:22.000 and letters for accuracy, a process that could have been 0:10:22.000 --> 0:10:24.880 done without generative AI, like in an audio editors such 0:10:24.880 --> 0:10:28.120 as pro Tools, but Respeecher made the process much more 0:10:28.120 --> 0:10:31.720 efficient and of course Adrian Brody won the Oscar for 0:10:31.760 --> 0:10:35.800 Best Actor. As us Say Today reported, not all viewers 0:10:35.840 --> 0:10:39.040 would pleased with the news. Don't think it's too reactionary 0:10:39.080 --> 0:10:41.240 to say this movie should lose the Academy buzz. It 0:10:41.280 --> 0:10:44.880 was getting one person posted on eggs. But the manipulation 0:10:44.960 --> 0:10:48.559 of vocal tracks is not uncommon in movies. Deadline noted 0:10:48.559 --> 0:10:51.600 that combinations of vocal tracks will use in performances like 0:10:51.720 --> 0:10:55.800 Romy Mallock's Oscar winning portrayal of Freddie Mercury, and Respeecher 0:10:55.840 --> 0:10:57.880 may have been used in another film nominated for Best 0:10:57.880 --> 0:11:01.440 Picture this year, Amelia Perez. The rise of generative AI 0:11:01.559 --> 0:11:04.440 has been remarkably fast in creative industries. But one big 0:11:04.520 --> 0:11:07.640 question I have is how far could this go and 0:11:07.720 --> 0:11:10.880 how soon? And to answer that, we want to turn 0:11:10.960 --> 0:11:13.679 to our friend Jason Kebler at four or four Media, 0:11:13.760 --> 0:11:16.959 who not too long ago attended a film festival of 0:11:17.080 --> 0:11:20.040 AI generated movies. Jason, welcome back to the show. 0:11:20.120 --> 0:11:21.120 Hey, thanks for having me. 0:11:21.360 --> 0:11:23.320 Before we get into that film festival you went to 0:11:23.600 --> 0:11:26.839 could you just explain how respeech it works and how 0:11:26.880 --> 0:11:29.439 it was used in the editing process for the Brutalist. 0:11:29.760 --> 0:11:34.720 Yeah. So, respeacher is an AI voice synthesizer, and so 0:11:35.160 --> 0:11:39.600 it takes training data of an actor's voice and runs 0:11:39.600 --> 0:11:42.240 it against a large language model. So in this case, 0:11:42.240 --> 0:11:45.800 it would probably be examples of the Hungarian language, et cetera. 0:11:46.280 --> 0:11:49.520 And it would take Adrian Brodie's voice and make it 0:11:49.760 --> 0:11:56.440 more closely match other examples of Hungarian language. And it's 0:11:56.520 --> 0:11:59.120 very interesting because this technology is sort of one of 0:11:59.160 --> 0:12:04.360 the first native AI technologies that was widely used commercially, 0:12:04.920 --> 0:12:08.800 not just Respeecher, but another company called eleven Labs has 0:12:09.640 --> 0:12:13.520 become really famous for like Eric Adams, the mayor of 0:12:13.600 --> 0:12:18.520 New York City, did a calling campaign to various communities 0:12:18.520 --> 0:12:21.240 in New York City where he spoke English, but then 0:12:21.360 --> 0:12:24.920 eleven Labs translated his voice into like fifteen different languages. 0:12:25.320 --> 0:12:28.920 And it's not just like a robot voice reading it 0:12:29.080 --> 0:12:33.239 sounds like Eric Adams speaking Mandarin or Eric Adams speaking Hungarian. 0:12:33.760 --> 0:12:37.600 And so increasingly this is being used in movies, not 0:12:37.640 --> 0:12:40.600 just Respeecher, but also eleven Labs and other tools like it, 0:12:41.160 --> 0:12:44.080 and it really is like one of the first big 0:12:44.120 --> 0:12:47.160 commercial uses of generative AI in movies. 0:12:47.720 --> 0:12:49.560 To me, it feels like it's not that far away 0:12:49.600 --> 0:12:52.520 from other post production tools that have been super charged 0:12:52.559 --> 0:12:55.679 by AI, like description, podcast editing, or other tools like that. 0:12:56.040 --> 0:12:58.880 Yeah, I mean it's really interesting because I think that 0:12:59.160 --> 0:13:02.280 music had this a long time ago, with things like autotune, 0:13:02.600 --> 0:13:06.880 and it's like many, many, many popular artists use autotune, 0:13:06.920 --> 0:13:10.439 and this is a very similar technology. I mean it's 0:13:10.679 --> 0:13:13.840 it's in the same family of technologies at least. So 0:13:13.920 --> 0:13:17.800 it just becomes a question of how much post can 0:13:17.880 --> 0:13:21.000 there be for the human performance to still be there. 0:13:21.040 --> 0:13:24.160 And I think it's a really open question at this point. 0:13:24.280 --> 0:13:27.480 I think if you asked me a while ago, I 0:13:27.480 --> 0:13:31.120 would say they're changing the performance in some fundamental way. 0:13:31.200 --> 0:13:35.080 But I think everything in a movie is so carefully edited, 0:13:35.160 --> 0:13:38.600 so carefully shot. They do hundreds of takes for certain 0:13:38.679 --> 0:13:42.000 scenes and then splice together different takes and cuts, and 0:13:42.040 --> 0:13:44.280 so I think it really is a spectrum of what 0:13:44.360 --> 0:13:48.240 you are willing to accept if you're in the Academy 0:13:48.240 --> 0:13:50.679 and need to decide whether someone is worthy of an 0:13:50.679 --> 0:13:54.040 award for this, I think audiences sort of have to 0:13:54.120 --> 0:13:56.240 accept it because it's being done, and it's been done 0:13:56.280 --> 0:13:57.960 for a long time. And I think that if you 0:13:58.040 --> 0:14:01.679 start like having purity tests about this sort of thing, 0:14:01.679 --> 0:14:03.800 I think it's going to be pretty difficult to know 0:14:03.880 --> 0:14:06.040 which movies to see and which are not to see, 0:14:06.080 --> 0:14:08.840 because ye, honestly, the only reason we know that this 0:14:09.040 --> 0:14:11.400 was used at all was because the editor talked about 0:14:11.400 --> 0:14:12.320 it to the media. 0:14:12.800 --> 0:14:15.280 Yeah. And also, I mean, to be fair to Adrian Brody, 0:14:15.360 --> 0:14:18.439 I doubt that many Academy members would have voted against 0:14:18.520 --> 0:14:21.600 him on the basis to his owncungarian accident wasn't quite perfect, 0:14:21.720 --> 0:14:23.960 So I'm not sure that this was like the key 0:14:24.000 --> 0:14:27.320 input to his victory. But what you said about like 0:14:27.880 --> 0:14:30.600 the role of post production and what that means visa 0:14:30.720 --> 0:14:33.600 v like the original product made me think about this 0:14:33.760 --> 0:14:36.960 AI generated film festival that you went to. So, first 0:14:37.000 --> 0:14:40.160 of all, what made this an AI generated film festival? 0:14:40.240 --> 0:14:42.760 How much of the films were AI generated? 0:14:43.440 --> 0:14:46.520 Yeah, so it varied for each movie, but I think 0:14:46.560 --> 0:14:49.320 that if you walked in off the street, you would say, oh, 0:14:49.400 --> 0:14:51.720 these films were made with AI And what I mean 0:14:51.760 --> 0:14:56.480 by that is each movie had visuals that were clearly 0:14:56.520 --> 0:14:59.680 AI generated, like a lot of the backgrounds were constantly 0:14:59.720 --> 0:15:02.920 changed in a way that if you were using a camera, 0:15:03.560 --> 0:15:07.800 they wouldn't happen. A lot of people had like faces 0:15:07.840 --> 0:15:10.600 that were morphing from scene to scene. One thing I 0:15:10.640 --> 0:15:14.280 will say though, is that TCL was very clear that 0:15:14.320 --> 0:15:17.040 all of the scripts were written by humans, and all 0:15:17.160 --> 0:15:19.320 the voices were done by humans, and all of the 0:15:19.440 --> 0:15:24.000 music was done by humans. The artificial intelligence was limited 0:15:24.040 --> 0:15:26.440 to the visuals in different movies. 0:15:27.080 --> 0:15:28.640 Can you just take me back to kind of how 0:15:28.640 --> 0:15:31.520 you got invited and what questions you had going in? 0:15:32.200 --> 0:15:35.920 Yeah, So I went to the Chinese Theater in Hollywood, 0:15:36.320 --> 0:15:39.520 which is ironically where the oscars are. It's like the 0:15:39.560 --> 0:15:44.040 same complex. And that theater is owned by TCL, which 0:15:44.160 --> 0:15:48.760 is a Chinese TV manufacturer, and like a lot of 0:15:48.800 --> 0:15:52.040 other TV manufacturers at this point, they have their own 0:15:52.200 --> 0:15:55.920 free streaming TV service if you buy a TCL TV, 0:15:56.360 --> 0:16:00.000 And TCL is the first company to put fully AI 0:16:00.160 --> 0:16:03.600 generated movies on its streaming service. And so this was 0:16:03.640 --> 0:16:08.920 a premiere of five films that were created using generative AI. 0:16:09.120 --> 0:16:12.200 And so I had been writing basically about this technology 0:16:12.200 --> 0:16:14.400 for a while and they invited me to come watch them. 0:16:14.720 --> 0:16:16.680 So, despite the fact that you'll perhaps more on the 0:16:16.680 --> 0:16:19.520 skeptical side, they welcome you into the film festival. 0:16:19.720 --> 0:16:22.320 I was pretty shocked that they invited me, because honestly, 0:16:22.360 --> 0:16:25.360 I had written about a trailer that they released for 0:16:25.440 --> 0:16:28.160 an AI generated film and I kind of dunked on it. 0:16:28.200 --> 0:16:31.640 I said, it was really terrible. It's called Last Train Paris, 0:16:31.680 --> 0:16:34.800 and it was like an AI generated rom com. And 0:16:35.320 --> 0:16:38.320 in the YouTube video, it's like the lip syncing of 0:16:38.360 --> 0:16:42.000 the audio and the lips is like really bad. The 0:16:42.120 --> 0:16:46.960 characters move incredibly robotically, and it has this very dreamlike 0:16:47.400 --> 0:16:52.200 quality to it that is very common with AI generated visuals, 0:16:52.280 --> 0:16:54.640 where it's not like a cool effect. It's like, wow, 0:16:54.680 --> 0:16:57.760 this is really distracting because the background is constantly swirling 0:16:57.800 --> 0:17:00.400 and changing and things are popping in and out. And 0:17:00.480 --> 0:17:03.680 after I wrote that article, they still decided to invite me, 0:17:03.760 --> 0:17:05.760 So I thought that was brave of them. 0:17:05.920 --> 0:17:07.159 But what did you think, I mean, what were you 0:17:07.200 --> 0:17:08.400 kind of expecting going into it? 0:17:08.880 --> 0:17:11.440 Going in? I thought that they would be pretty bad, 0:17:11.560 --> 0:17:14.600 to be totally honest with you, just because the state 0:17:14.640 --> 0:17:18.200 of the art at the time. This was back in December, 0:17:18.520 --> 0:17:22.840 which it was only three months ago, but at the time, 0:17:23.040 --> 0:17:28.920 AI video generators were pretty bad, and I didn't think 0:17:28.960 --> 0:17:32.600 that TCL had access to some proprietary system that we 0:17:32.680 --> 0:17:35.119 hadn't seen before. I figured that they would be using 0:17:35.560 --> 0:17:38.040 the state of the art that you can find on 0:17:38.119 --> 0:17:41.600 the internet, and I think that those tools are not 0:17:41.800 --> 0:17:43.680 very good, and so I thought that they would be bad, 0:17:43.800 --> 0:17:46.280 to be totally honest with you, and they were bad. 0:17:48.640 --> 0:17:51.120 Can you describe some of the highlights on the low Lights? 0:17:51.400 --> 0:17:54.760 Yeah? I thought that the films themselves were just they 0:17:54.760 --> 0:17:57.199 felt pretty rushed. So one of them was called The 0:17:57.240 --> 0:18:00.600 Slug and it's about a woman who turns into a slug. 0:18:00.760 --> 0:18:02.719 She has a disease that turns her into a slug 0:18:02.760 --> 0:18:05.600 and it feels like The Substance, which is another you know, 0:18:05.640 --> 0:18:10.160 Oscar nominated film. The visuals on it are wild. Things 0:18:10.200 --> 0:18:12.800 are just like constantly changing. Her face is changing, the 0:18:13.080 --> 0:18:15.600 you know, the food is changing. There's a lot of 0:18:15.640 --> 0:18:19.840 like weird screams that happen that are not super well 0:18:19.880 --> 0:18:23.439 timed with the dialogue. And then also there's like a 0:18:23.480 --> 0:18:26.240 scene where the woman takes a bath and there's like 0:18:26.240 --> 0:18:29.040 a close up on some bath salts and like the 0:18:29.160 --> 0:18:33.639 text on that label is like an alien language because 0:18:33.680 --> 0:18:37.800 AI has like a really bad time generating text, and 0:18:38.359 --> 0:18:39.800 I guess you can take it with a grain of 0:18:39.840 --> 0:18:42.560 salt or say like, hey, this is early technology. But 0:18:42.840 --> 0:18:45.200 when you're watching something as a viewer in a movie 0:18:45.280 --> 0:18:48.439 theater on this giant screen and the text is completely 0:18:48.800 --> 0:18:51.360 not even in English, it's like, wow, it really takes 0:18:51.359 --> 0:18:52.360 you out of the narrative. 0:18:52.520 --> 0:18:54.800 I would say, I mean it's a weird idea, right, 0:18:54.880 --> 0:18:57.040 because I mean you mentioned this is for TCL, the 0:18:57.119 --> 0:19:01.040 Chinese TV manufacturer, and the assumption be like, they don't 0:19:01.080 --> 0:19:02.639 want you to change the channel, right, they want you 0:19:02.680 --> 0:19:04.800 to have their own channel on kind of in the 0:19:04.840 --> 0:19:07.240 background so that you know your attention is with them 0:19:07.280 --> 0:19:08.800 and they can sell you ads whatever it may be. 0:19:09.320 --> 0:19:11.800 But that's very different to like putting hundreds of people 0:19:11.960 --> 0:19:14.359 in a movie theater and kind of fulcing them to 0:19:14.400 --> 0:19:16.160 watch with full attention, right, yeah. 0:19:16.280 --> 0:19:16.520 Yeah. 0:19:16.520 --> 0:19:21.399 And it's very interesting because before the movies played, two 0:19:21.440 --> 0:19:25.119 TCL executives addressed the audience, and it was very interesting 0:19:25.280 --> 0:19:27.520 the difference between what they were saying and what the 0:19:27.560 --> 0:19:31.600 filmmakers were saying, because the TCL executives were business people 0:19:32.000 --> 0:19:35.639 and they were saying our research shows that almost no 0:19:35.680 --> 0:19:38.800 one changes the channel once they're watching something like this, 0:19:38.960 --> 0:19:42.000 like they are watching it in the background usually, and 0:19:42.119 --> 0:19:44.520 so their hope is that you're just going to be 0:19:44.520 --> 0:19:46.199 too lazy to change the channel. 0:19:46.280 --> 0:19:48.800 So inspiring creative brief. 0:19:49.440 --> 0:19:51.920 Right, right, And then the other executives said, like, we're 0:19:51.920 --> 0:19:55.240 going to use this as part of our targeted advertising strategy, 0:19:56.440 --> 0:20:00.280 which was pretty dystopian. And then the actual filmmakers came 0:20:00.280 --> 0:20:02.040 on and said, you know, we put our heart and 0:20:02.080 --> 0:20:03.960 soul into this, and we think this is the future 0:20:03.960 --> 0:20:06.680 of the industry. So that was kind of like a 0:20:06.680 --> 0:20:09.200 whiplash situation for me in the audience. 0:20:12.240 --> 0:20:14.720 When we come back, more from Jason Kebler about the 0:20:14.800 --> 0:20:19.240 rapid advances in generative AI video technology and how the 0:20:19.280 --> 0:20:22.440 state of the art is evolving in real time, stay 0:20:22.440 --> 0:20:37.320 with us. Welcome back to our conversation with Jason Kebler 0:20:37.359 --> 0:20:40.639 from four or four Media, where we continue our conversation 0:20:40.920 --> 0:20:45.359 about a recent AI film festival he attended. There was 0:20:45.400 --> 0:20:47.240 one film though, which I think was like a kind 0:20:47.240 --> 0:20:51.600 of blended documentary and AI film that you thought was 0:20:51.640 --> 0:20:52.960 potentially a bit more interesting. 0:20:53.560 --> 0:20:56.560 Yeah, I thought it was pretty cool. I mean, it 0:20:56.640 --> 0:20:58.640 still had a lot of problems, but It was called 0:20:58.640 --> 0:21:02.040 The Best Day of My Life, and it was mountaineering 0:21:02.119 --> 0:21:07.840 documentary where a mountaineer who got trapped in an avalanche 0:21:08.520 --> 0:21:11.600 is talking directly to the camera, like the actual person 0:21:11.640 --> 0:21:14.119 is talking directly to the camera recounting his story, and 0:21:14.280 --> 0:21:18.680 as he is telling his story, they flashed to generative 0:21:18.720 --> 0:21:22.760 AI depictions of what he is saying, And so I 0:21:22.840 --> 0:21:24.920 thought that was kind of interesting because this is something 0:21:24.960 --> 0:21:27.040 that happened to the guy. He obviously didn't bring a 0:21:27.080 --> 0:21:30.639 camera with him at the time, and you were able 0:21:30.640 --> 0:21:34.600 to sort of like see what he was describing. 0:21:34.440 --> 0:21:37.960 In a way that was actually viscerally compelling, or in 0:21:37.960 --> 0:21:40.560 a way that's still felt a bit uncanny and jarring. 0:21:40.640 --> 0:21:42.800 In a way that made me think that maybe this 0:21:42.880 --> 0:21:45.920 has potential in the future, but this isn't quite there yet, 0:21:45.960 --> 0:21:50.719 because it similarly like the there's various scenes in the film, 0:21:51.359 --> 0:21:55.600 and the guy who's happening to changes in each scene. 0:21:55.640 --> 0:21:59.719 It's like his face looks different in different scenes. He 0:21:59.800 --> 0:22:02.800 was under snow because it was an avalanche, and then 0:22:02.840 --> 0:22:05.080 in the next scene all of the snow had turned 0:22:05.119 --> 0:22:07.639 to mud, and then it turned back to snow, and 0:22:08.160 --> 0:22:12.480 it was like, similarly took you out of the narrative, 0:22:12.680 --> 0:22:16.879 but I thought that the idea behind it was pretty 0:22:16.880 --> 0:22:20.239