1 00:00:15,600 --> 00:00:18,320 Speaker 1: Welcome to tech stuff. I'm Os Voloshin, and this is 2 00:00:18,360 --> 00:00:21,200 Speaker 1: the story. A couple of weeks ago, I was at 3 00:00:21,239 --> 00:00:24,720 Speaker 1: the Web Summit in Doha Cutter and I had the 4 00:00:24,760 --> 00:00:28,080 Speaker 1: opportunity to speak with someone who is central to Hollywood's 5 00:00:28,120 --> 00:00:31,920 Speaker 1: AI transformation, one of the most controversial and to some 6 00:00:32,360 --> 00:00:37,120 Speaker 1: frightening topics facing the industry. Christobal Valenzuela is the co 7 00:00:37,200 --> 00:00:40,559 Speaker 1: founder and CEO of Runway, which is best known for 8 00:00:40,600 --> 00:00:44,839 Speaker 1: developing a very powerful AI video generator along with other 9 00:00:44,880 --> 00:00:48,400 Speaker 1: tools for film production. They are actually used by professionals 10 00:00:48,520 --> 00:00:51,440 Speaker 1: in the industry. Its software has been used in the 11 00:00:51,440 --> 00:00:54,960 Speaker 1: Oscar Winner, Everything Everywhere, All at Once, on The Late 12 00:00:55,040 --> 00:00:59,000 Speaker 1: Show by Madonna, and Runway has partnerships with everyone from 13 00:00:59,040 --> 00:01:04,520 Speaker 1: Disney Line, Skate and AMC to Adobe and Nvidia. At Runway, 14 00:01:04,920 --> 00:01:08,000 Speaker 1: Crystabal is hoping to transform more than just movies. 15 00:01:08,760 --> 00:01:10,280 Speaker 2: We're gonna get to a point where you can generate 16 00:01:10,319 --> 00:01:13,800 Speaker 2: anything you want. Literally, you can create media as fast 17 00:01:13,800 --> 00:01:15,920 Speaker 2: as you can think of. And when I mean media, 18 00:01:16,000 --> 00:01:18,280 Speaker 2: I think most people think about like films and ads 19 00:01:18,319 --> 00:01:21,320 Speaker 2: because that's what we imagine media to be. But we 20 00:01:21,400 --> 00:01:23,880 Speaker 2: are going deeper than that. These are pixels on a 21 00:01:23,920 --> 00:01:27,040 Speaker 2: screen Every screen that you see in the world has pixels, 22 00:01:27,319 --> 00:01:28,800 Speaker 2: and they're gonna get to a point where you can 23 00:01:29,200 --> 00:01:32,200 Speaker 2: simulate all those pixels, and those are gonna be pixels 24 00:01:32,200 --> 00:01:33,520 Speaker 2: that have been coming from our model. 25 00:01:34,360 --> 00:01:39,000 Speaker 1: Crystabal thinks Runway can improve gaming education robotics. In fact, 26 00:01:39,319 --> 00:01:42,319 Speaker 1: he's one of the people racing towards a new frontier 27 00:01:42,360 --> 00:01:47,720 Speaker 1: of computing called world Models. And I've actually met Christabell before, 28 00:01:48,280 --> 00:01:50,960 Speaker 1: more than six years ago, when Karen and I were 29 00:01:50,960 --> 00:01:54,920 Speaker 1: making the Sleepwalkers podcast. It was a very different world 30 00:01:54,960 --> 00:01:58,080 Speaker 1: back then, and that's where I want to begin. Take 31 00:01:58,080 --> 00:02:03,880 Speaker 1: a listen. Well, Christy, it's great to see you again. Yeah, 32 00:02:03,880 --> 00:02:06,000 Speaker 1: I'm not sure how well you remember. You've had a 33 00:02:06,000 --> 00:02:08,640 Speaker 1: busy few years. But we met back in twenty eighteen 34 00:02:09,320 --> 00:02:12,640 Speaker 1: and I was hosting a podcast called Sleepwalkers, and you 35 00:02:12,760 --> 00:02:16,280 Speaker 1: helped my partner Kara clone her voice, and then we 36 00:02:16,360 --> 00:02:18,280 Speaker 1: used her voice to see if we could trick her 37 00:02:18,320 --> 00:02:20,520 Speaker 1: cousin into thinking it was really her, and we kind 38 00:02:20,520 --> 00:02:23,880 Speaker 1: of half vay succeeded. What's changed for you since then? 39 00:02:24,040 --> 00:02:27,119 Speaker 2: I remember that was a long time ago. A lot 40 00:02:27,160 --> 00:02:30,200 Speaker 2: has changed. The world has changed. I've changed running it 41 00:02:30,200 --> 00:02:32,560 Speaker 2: has changed a lot. I think from thementally, we've got 42 00:02:32,600 --> 00:02:35,880 Speaker 2: in just way further when it comes to what I 43 00:02:35,919 --> 00:02:37,919 Speaker 2: can do in media. So I guess the fact you 44 00:02:37,960 --> 00:02:40,720 Speaker 2: guys were early on trying to figure out how you 45 00:02:40,760 --> 00:02:44,360 Speaker 2: could use models for like cloning voices or imageries or videos. 46 00:02:44,840 --> 00:02:46,760 Speaker 2: I think the world went from like could it be 47 00:02:46,840 --> 00:02:50,440 Speaker 2: possible to like, yeah, it's really possible. It worked really well. 48 00:02:50,639 --> 00:02:52,600 Speaker 1: It's kind of amazing. I remember then it was like, 49 00:02:53,320 --> 00:02:55,959 Speaker 1: I think you needed like ten hours of Kara's voice, 50 00:02:56,000 --> 00:02:57,920 Speaker 1: and it was like quite a manual process to make 51 00:02:57,960 --> 00:03:00,240 Speaker 1: a digital copy, and digital copy was good but not 52 00:03:00,320 --> 00:03:03,960 Speaker 1: indistinguishable from the real voice. And it's just remarkable to 53 00:03:04,040 --> 00:03:06,440 Speaker 1: think that in such a short period, the synthetic media 54 00:03:06,480 --> 00:03:09,440 Speaker 1: revolution that has just taken off in an extraordinary way. 55 00:03:09,560 --> 00:03:11,359 Speaker 1: Did you always believe we would be here? 56 00:03:12,080 --> 00:03:12,280 Speaker 2: Yeah? 57 00:03:12,320 --> 00:03:12,840 Speaker 1: I think we were. 58 00:03:12,919 --> 00:03:14,840 Speaker 2: I always believed we will be here. I think the 59 00:03:14,880 --> 00:03:17,680 Speaker 2: only question was like how fast we'll get there, And 60 00:03:17,760 --> 00:03:20,200 Speaker 2: to be honest, I think it's been faster than like 61 00:03:20,280 --> 00:03:24,239 Speaker 2: our best estimates. Like when we started thinking about this, 62 00:03:24,520 --> 00:03:27,320 Speaker 2: the question was would you be able to generate or 63 00:03:27,440 --> 00:03:30,760 Speaker 2: create any piece of pixel or imagery or content that 64 00:03:30,960 --> 00:03:36,680 Speaker 2: feels right or feels real? And the answer is like absolutely, 65 00:03:37,120 --> 00:03:42,120 Speaker 2: Like you can create incredibly compelling media, stories, films, audio, 66 00:03:42,200 --> 00:03:45,320 Speaker 2: whatever kind of format you want. And now we're like, ah, 67 00:03:45,560 --> 00:03:47,320 Speaker 2: needs to be faster, you know, it needs to be 68 00:03:47,400 --> 00:03:49,560 Speaker 2: like better, and it's like we're gonna get there, but 69 00:03:50,640 --> 00:03:52,480 Speaker 2: it took us five years to get here, and I 70 00:03:52,560 --> 00:03:55,080 Speaker 2: think the next five years are going to be condensed 71 00:03:55,080 --> 00:03:57,960 Speaker 2: in the next like eight months, you know, and that. 72 00:03:58,000 --> 00:03:59,640 Speaker 1: Isn't going to see as much progress in the next 73 00:03:59,680 --> 00:04:01,200 Speaker 1: eight month sas in the last five years. 74 00:04:01,360 --> 00:04:03,640 Speaker 2: Yeah, I think. I mean, over the last twelve months, 75 00:04:03,640 --> 00:04:06,560 Speaker 2: we've seen enough of the similar progress of the one 76 00:04:06,600 --> 00:04:08,560 Speaker 2: we've seen over the last four or five years. 77 00:04:08,680 --> 00:04:11,440 Speaker 1: And what will that feel like the average person? That's 78 00:04:11,440 --> 00:04:12,280 Speaker 1: an interesting question. 79 00:04:12,360 --> 00:04:14,520 Speaker 2: So you know, I think I've been thinking a lot 80 00:04:14,560 --> 00:04:17,080 Speaker 2: about this, and I think there's a there's like I've 81 00:04:17,080 --> 00:04:19,839 Speaker 2: noticed there's like sometimes like two worlds. There's the world 82 00:04:19,880 --> 00:04:23,560 Speaker 2: of people who are very deep into AI and the 83 00:04:23,640 --> 00:04:26,800 Speaker 2: latest models and the things you could do and the 84 00:04:26,880 --> 00:04:29,720 Speaker 2: ins and out and within that community seeing people who 85 00:04:29,720 --> 00:04:32,200 Speaker 2: are even like further down the road, you know, and 86 00:04:32,240 --> 00:04:35,839 Speaker 2: there's another world, which is people who were completely unaware 87 00:04:35,880 --> 00:04:39,479 Speaker 2: of this, yeah, completely unaware, Like and I kind of exaggerate. 88 00:04:39,600 --> 00:04:42,840 Speaker 2: A's like people who are just discovering like chat GPT, 89 00:04:43,200 --> 00:04:45,720 Speaker 2: you know, and for me, that's creating like an interesting 90 00:04:45,720 --> 00:04:48,440 Speaker 2: asymmetry where on the one end, you have these people 91 00:04:48,520 --> 00:04:52,360 Speaker 2: living two hundred years into the future now and others 92 00:04:52,360 --> 00:04:55,719 Speaker 2: are just like still trying to understand how to like 93 00:04:55,920 --> 00:05:00,440 Speaker 2: properly use like a language model, and that for me, 94 00:05:00,480 --> 00:05:04,160 Speaker 2: will continue to exacerbate the asymmetry of like those who 95 00:05:04,200 --> 00:05:07,240 Speaker 2: are using the latest and are curious about it and 96 00:05:07,279 --> 00:05:10,279 Speaker 2: those who are like still trying to adjust. But the 97 00:05:10,360 --> 00:05:12,560 Speaker 2: interesting thing is here, is that if you want to 98 00:05:12,920 --> 00:05:16,560 Speaker 2: transition from like the past to the future, there's nothing 99 00:05:16,600 --> 00:05:18,839 Speaker 2: preventing you from doing it. Like it's just a curiosity. 100 00:05:18,880 --> 00:05:20,680 Speaker 2: It's like mindset of like growth. It's like do you 101 00:05:20,720 --> 00:05:22,479 Speaker 2: want to learn, Like you can just learn how to 102 00:05:22,480 --> 00:05:27,120 Speaker 2: do it. And the most interesting I would say, like 103 00:05:27,320 --> 00:05:30,520 Speaker 2: thing that will happen of the next year or so 104 00:05:30,839 --> 00:05:33,720 Speaker 2: is that asymmetry will continue to grow. 105 00:05:34,080 --> 00:05:35,400 Speaker 1: And she's a huge social problem. 106 00:05:35,520 --> 00:05:37,359 Speaker 2: It is. It is, it is definitely, But it's the 107 00:05:37,360 --> 00:05:39,839 Speaker 2: first one that you're not constrained by like your social 108 00:05:39,839 --> 00:05:43,640 Speaker 2: economic status or your political position or like it's mostly 109 00:05:43,680 --> 00:05:45,560 Speaker 2: about do you want to like try this new thing. 110 00:05:45,600 --> 00:05:48,159 Speaker 2: And I'll an example. We work a lot with media companies, 111 00:05:48,160 --> 00:05:52,720 Speaker 2: So studios, media filmmakers work with all these students, all 112 00:05:52,720 --> 00:05:54,960 Speaker 2: of them, and what are you doing for them? So 113 00:05:55,360 --> 00:05:57,560 Speaker 2: if you're the thing right where we're building this like 114 00:05:57,880 --> 00:06:00,680 Speaker 2: incredibly powerful technology that allows them to do things they 115 00:06:00,720 --> 00:06:04,000 Speaker 2: only dram of doing at some point, and the way 116 00:06:04,320 --> 00:06:06,159 Speaker 2: that you need to help them understand it is they 117 00:06:06,240 --> 00:06:08,080 Speaker 2: need to show it to them, You need to walk them, 118 00:06:08,120 --> 00:06:10,440 Speaker 2: you need to kind of understand how they're going to 119 00:06:10,440 --> 00:06:12,919 Speaker 2: transition to this totally new way of making things. And 120 00:06:13,440 --> 00:06:17,200 Speaker 2: the biggest bottleneck in adoption is not compute. It's not 121 00:06:17,360 --> 00:06:19,880 Speaker 2: the capacity of the models, it's not the safety of 122 00:06:19,880 --> 00:06:26,080 Speaker 2: the models. It's basically people's apprehensions or kind of expectations 123 00:06:26,160 --> 00:06:29,320 Speaker 2: or knowledge of how the technology works. Right, So we 124 00:06:29,360 --> 00:06:33,880 Speaker 2: spend a lot of time helping people adjust, helping people 125 00:06:33,920 --> 00:06:37,240 Speaker 2: train themselves. Like, if you're made a movie and the 126 00:06:37,240 --> 00:06:40,080 Speaker 2: movie involves fifty different steps with tools that are one 127 00:06:40,120 --> 00:06:42,720 Speaker 2: hundred years old, then guess what You're going to make 128 00:06:42,720 --> 00:06:44,719 Speaker 2: a movie the same movie with like four tools in 129 00:06:44,760 --> 00:06:47,880 Speaker 2: twenty minutes and in ten steps, and for you to 130 00:06:48,320 --> 00:06:52,520 Speaker 2: reor organize your mind around it, it's just hard, and 131 00:06:52,560 --> 00:06:53,960 Speaker 2: that's where we spend a lot of time, have people 132 00:06:53,960 --> 00:06:55,719 Speaker 2: dedicated to just like helping you adjust. 133 00:06:56,120 --> 00:06:58,880 Speaker 1: You raise more than three hundred million dollars earlier this 134 00:06:58,960 --> 00:07:03,120 Speaker 1: year at a value above three billion dollars. So you're 135 00:07:03,400 --> 00:07:06,480 Speaker 1: the only Unicorn founder I know since you were before 136 00:07:06,640 --> 00:07:08,440 Speaker 1: Unicorn founder, since you were more or less in the 137 00:07:08,560 --> 00:07:12,480 Speaker 1: graduate graduate school at NYUS. So that must be an interesting, 138 00:07:12,680 --> 00:07:16,960 Speaker 1: interesting journey and experience. But there's one moment between when 139 00:07:16,960 --> 00:07:20,000 Speaker 1: we first met in twenty eighteen and today where you 140 00:07:20,080 --> 00:07:22,480 Speaker 1: suddenly kind of felt the wind at your sales and like, 141 00:07:22,480 --> 00:07:24,160 Speaker 1: oh my god, this is really going to work because 142 00:07:24,360 --> 00:07:26,040 Speaker 1: may change the world. Like what was it? 143 00:07:26,880 --> 00:07:29,240 Speaker 2: I think it's been. It's been a series of like 144 00:07:29,280 --> 00:07:31,480 Speaker 2: I would say milestones that we've seen in the last 145 00:07:31,520 --> 00:07:35,160 Speaker 2: three to four years. Where we released one on twenty 146 00:07:35,200 --> 00:07:39,360 Speaker 2: twenty three, So just for like contexts, Gene one was 147 00:07:39,400 --> 00:07:42,640 Speaker 2: the first ever publicly commercially available video model in the world. 148 00:07:43,160 --> 00:07:44,440 Speaker 2: It was such a big deal that it was in 149 00:07:44,480 --> 00:07:46,120 Speaker 2: the front page of the New York Times back then, 150 00:07:46,840 --> 00:07:49,280 Speaker 2: and the quality of few was like bad compared to 151 00:07:49,280 --> 00:07:52,840 Speaker 2: today's standards, right, Like seeing the level of progress within 152 00:07:52,920 --> 00:07:55,360 Speaker 2: those two years, there's enough moments where like you see 153 00:07:55,360 --> 00:07:57,280 Speaker 2: what people can make that you never thought you could 154 00:07:57,320 --> 00:08:00,680 Speaker 2: make yourself. That I'm like, Okay, it's working. Like if 155 00:08:00,680 --> 00:08:03,480 Speaker 2: a tool is general enough, you're gonna see things that 156 00:08:03,640 --> 00:08:07,440 Speaker 2: you cannot even think about because they don't you don't 157 00:08:07,440 --> 00:08:09,600 Speaker 2: come from the main right, So I see musicians using 158 00:08:09,640 --> 00:08:13,320 Speaker 2: runway ways they couldn't think of before. I see architects. 159 00:08:13,400 --> 00:08:16,560 Speaker 2: We saw a lot to architects using runways that I 160 00:08:16,600 --> 00:08:18,520 Speaker 2: could never think of before. And so we have all 161 00:08:18,520 --> 00:08:20,840 Speaker 2: these very interesting use cases that I'm like, Okay, this 162 00:08:21,040 --> 00:08:22,760 Speaker 2: is working, this is there. 163 00:08:22,800 --> 00:08:25,440 Speaker 1: We go and the business is more of a B 164 00:08:25,480 --> 00:08:27,880 Speaker 1: to B business in terms of the partnerships with Adobe 165 00:08:27,920 --> 00:08:29,880 Speaker 1: and Disney and others, or is it more of a 166 00:08:29,920 --> 00:08:32,440 Speaker 1: B two C business ware individual artists and creators are 167 00:08:32,480 --> 00:08:34,199 Speaker 1: trying to maximize what they can do. 168 00:08:34,320 --> 00:08:37,920 Speaker 2: Yeah, we're working a lot with companies, with enterprises, with businesses. 169 00:08:38,000 --> 00:08:42,559 Speaker 2: We sell to almost every studio, all studios. We sell 170 00:08:42,559 --> 00:08:44,599 Speaker 2: a lot to brands, to agencies. We sell it to 171 00:08:44,640 --> 00:08:45,640 Speaker 2: both production teams. 172 00:08:45,800 --> 00:08:47,280 Speaker 1: Madonna, we saw a lot. 173 00:08:47,480 --> 00:08:49,560 Speaker 2: So we work a little with artists themselves as well. Yeah, 174 00:08:49,679 --> 00:08:52,640 Speaker 2: so we work a lot with the teams behind the artists. 175 00:08:52,640 --> 00:08:55,559 Speaker 2: So if you're creating a like a show like Madonna 176 00:08:55,600 --> 00:08:58,280 Speaker 2: puts out in the world. You imagine there's like hundreds 177 00:08:58,280 --> 00:09:00,599 Speaker 2: of people behind the scenes crafting the visuals and the 178 00:09:00,679 --> 00:09:03,199 Speaker 2: videos and everything else. And a lot of those teams 179 00:09:03,280 --> 00:09:03,880 Speaker 2: use Runway. 180 00:09:04,000 --> 00:09:06,040 Speaker 1: But on the consumer side, I mean you you grew 181 00:09:06,080 --> 00:09:08,439 Speaker 1: up in Chile, correct, You wanted to be a filmmaker, 182 00:09:08,480 --> 00:09:10,280 Speaker 1: I think, and you didn't have access to all of 183 00:09:10,320 --> 00:09:13,600 Speaker 1: the equipment that you wanted. And I have a sense 184 00:09:13,640 --> 00:09:18,080 Speaker 1: that for you, the individual user is something which is 185 00:09:18,160 --> 00:09:21,400 Speaker 1: important to you. And you've created a film festival, I think, yeah, 186 00:09:21,800 --> 00:09:23,360 Speaker 1: partly in service of that, correct. 187 00:09:23,440 --> 00:09:26,960 Speaker 2: Yeah. So one of those barriers I think AI is 188 00:09:27,000 --> 00:09:29,240 Speaker 2: breaking down is the definition of like who gets to 189 00:09:29,240 --> 00:09:32,280 Speaker 2: make things in the world. And look at Hollywood, like 190 00:09:32,440 --> 00:09:34,840 Speaker 2: a movie takes a couple of million dollars at the minimum. 191 00:09:34,840 --> 00:09:37,240 Speaker 2: You have to know a couple of production people, you 192 00:09:37,280 --> 00:09:41,840 Speaker 2: have to be within the network and the insider community 193 00:09:41,840 --> 00:09:46,280 Speaker 2: of the industry itself. It's really it's really expensive and 194 00:09:46,760 --> 00:09:49,640 Speaker 2: really powerful. General purpose technology like AI allows it to 195 00:09:49,640 --> 00:09:53,760 Speaker 2: basically break down barrier. We've seen now filmmakers in their 196 00:09:53,800 --> 00:09:58,200 Speaker 2: houses like teams of one, two, three people making films 197 00:09:58,200 --> 00:10:00,800 Speaker 2: that five four, three years ago, you I thought were 198 00:10:00,800 --> 00:10:02,880 Speaker 2: made by a team of one hundred or more people 199 00:10:02,920 --> 00:10:05,840 Speaker 2: with a budget of a couple million dollars. And that's 200 00:10:05,920 --> 00:10:09,400 Speaker 2: just remarkable. Like a few times in our lifetime we've 201 00:10:09,440 --> 00:10:13,120 Speaker 2: seen technology like advanced so rapidly that those barriers are 202 00:10:13,480 --> 00:10:16,800 Speaker 2: kind of like broken so fast. And so the film 203 00:10:16,840 --> 00:10:19,760 Speaker 2: festival that we started four years ago now is a 204 00:10:19,840 --> 00:10:22,880 Speaker 2: celebration of those people. It's not about technology, it's not 205 00:10:22,920 --> 00:10:26,120 Speaker 2: about the latest and the greatest. It's about like people. 206 00:10:26,440 --> 00:10:28,760 Speaker 2: And so we host this festival both in New York 207 00:10:28,800 --> 00:10:32,520 Speaker 2: and in la and Paris that showcases the best of 208 00:10:32,720 --> 00:10:36,320 Speaker 2: artists experimenting with a news technology. And years look four 209 00:10:36,400 --> 00:10:40,280 Speaker 2: years ago in a lifetime is like one hundred years, right, 210 00:10:40,440 --> 00:10:43,240 Speaker 2: the things that moved too fast, and so four years 211 00:10:43,280 --> 00:10:45,960 Speaker 2: ago it was even before Jenu one. So it was 212 00:10:46,040 --> 00:10:48,760 Speaker 2: like a lot of experimental art, a lot of like 213 00:10:49,280 --> 00:10:53,079 Speaker 2: people who are like just tinkering with this. We got 214 00:10:53,120 --> 00:10:56,280 Speaker 2: like a hundred submissions. It's been four years since then. 215 00:10:57,120 --> 00:11:00,920 Speaker 2: The last festival, we got six thous and submissions from 216 00:11:00,920 --> 00:11:03,560 Speaker 2: people over the world. We rented the Lincoln Center in 217 00:11:03,559 --> 00:11:05,880 Speaker 2: New York City, which is like this beautiful historic venue. 218 00:11:06,559 --> 00:11:08,920 Speaker 2: We partner with like they try back up in festival 219 00:11:08,960 --> 00:11:11,760 Speaker 2: with the American Cinema Editors. We partner with like so 220 00:11:11,920 --> 00:11:16,040 Speaker 2: many industry experts in teams and companies, and now these 221 00:11:16,120 --> 00:11:18,520 Speaker 2: years not even bigger, we expect to get tenth of 222 00:11:18,600 --> 00:11:22,440 Speaker 2: thousands of submissions, and so it's a sign of the times. 223 00:11:22,480 --> 00:11:24,880 Speaker 1: I would say, what are the criteria to submit? Because 224 00:11:24,920 --> 00:11:26,600 Speaker 1: you know, there was an h twenty four film that 225 00:11:26,679 --> 00:11:29,240 Speaker 1: came out about a year ago where Hugh Grant plays 226 00:11:29,280 --> 00:11:32,000 Speaker 1: like a demonic character and there are two Mormon missionaries 227 00:11:32,080 --> 00:11:33,760 Speaker 1: and then he like traps them in the basement. I 228 00:11:33,760 --> 00:11:35,920 Speaker 1: can't know what it's called, but there was a disclaimer 229 00:11:35,960 --> 00:11:37,880 Speaker 1: at the end saying no AI was used in the 230 00:11:37,880 --> 00:11:40,439 Speaker 1: making of this film. And I was thinking that that 231 00:11:40,640 --> 00:11:42,880 Speaker 1: just by definition can't be true, right, because all of 232 00:11:42,920 --> 00:11:44,840 Speaker 1: the tools that people are using to make this film 233 00:11:44,880 --> 00:11:47,079 Speaker 1: in terms of Adobe and stuff have jen AI R. 234 00:11:47,720 --> 00:11:49,880 Speaker 1: So when you think about a film festival, like what, 235 00:11:50,080 --> 00:11:52,160 Speaker 1: like surely every film could be eligible. 236 00:11:52,440 --> 00:11:53,959 Speaker 2: Yeah. Look, I think there's a lot of vert to 237 00:11:54,000 --> 00:11:55,840 Speaker 2: signaling in the industry as well, where you want to 238 00:11:55,880 --> 00:11:58,000 Speaker 2: say something just to like make sure you like say 239 00:11:58,040 --> 00:12:00,600 Speaker 2: the right things, but like you're right, you're using AI, 240 00:12:00,880 --> 00:12:02,720 Speaker 2: whether you like it or not. You're taking your phone 241 00:12:02,760 --> 00:12:05,840 Speaker 2: to take a photo. Guess what That whole photo is 242 00:12:06,400 --> 00:12:10,080 Speaker 2: partially generated. Just you can't tell the sensor, the way 243 00:12:10,160 --> 00:12:13,560 Speaker 2: light is adjusted, like the composition, the sharpness, most of 244 00:12:13,559 --> 00:12:14,960 Speaker 2: those things are generated. 245 00:12:14,840 --> 00:12:16,520 Speaker 1: For the films that are submitted. Do they have to 246 00:12:16,520 --> 00:12:19,480 Speaker 1: have a certain amount of AI in that production process 247 00:12:19,559 --> 00:12:21,360 Speaker 1: or what are you? Can anyone submit a film? 248 00:12:21,400 --> 00:12:23,840 Speaker 2: Anyone anyone can submit a film that there the record 249 00:12:23,880 --> 00:12:26,199 Speaker 2: site is like you have to use AI somehow, right, 250 00:12:26,240 --> 00:12:28,840 Speaker 2: you have you have to explain how I mean to 251 00:12:28,880 --> 00:12:29,680 Speaker 2: be like interesting and. 252 00:12:29,720 --> 00:12:32,160 Speaker 1: Creatively, not the usage of the techno. But it's not, 253 00:12:32,280 --> 00:12:32,720 Speaker 1: it's not. 254 00:12:32,920 --> 00:12:35,880 Speaker 2: It's not. So their judges and they're the judges are 255 00:12:36,160 --> 00:12:39,120 Speaker 2: experts and their filmmakers. We have gazpart No like, who's 256 00:12:39,160 --> 00:12:41,160 Speaker 2: like a great filmmaker, like be one of the judges 257 00:12:41,240 --> 00:12:44,440 Speaker 2: last time last year. We have Harmony Kreen, who's like 258 00:12:44,480 --> 00:12:47,400 Speaker 2: the director of Kids, Like so many great filmmakers are 259 00:12:47,440 --> 00:12:50,240 Speaker 2: evaluating not the technology, they are really in the story, right, 260 00:12:50,559 --> 00:12:53,040 Speaker 2: So they're evaluating how are you using tagality in the 261 00:12:53,080 --> 00:12:55,280 Speaker 2: service of a story. If a story is not good, 262 00:12:55,920 --> 00:12:58,920 Speaker 2: it doesn't matter, like as is, you can make the 263 00:12:58,960 --> 00:13:01,720 Speaker 2: greatest increase as visuals, but like, if you're not telling 264 00:13:01,720 --> 00:13:03,200 Speaker 2: a story and then it's not gonna matter. So at 265 00:13:03,200 --> 00:13:04,640 Speaker 2: the end of the day, that's what I'm saying is like, 266 00:13:05,040 --> 00:13:07,560 Speaker 2: these are really powerful technologiest and if you know how 267 00:13:07,600 --> 00:13:09,079 Speaker 2: to use him, you're gonna get far. I'll give you 268 00:13:09,080 --> 00:13:13,079 Speaker 2: an example that the winner of last year's festival. It's 269 00:13:13,080 --> 00:13:15,760 Speaker 2: a beautiful film. You should watch. It's in our website. 270 00:13:16,200 --> 00:13:16,880 Speaker 1: It's a reflection. 271 00:13:16,920 --> 00:13:20,880 Speaker 2: It's like a documentary of sorts of where like AI 272 00:13:20,960 --> 00:13:22,600 Speaker 2: is gonna take us. It's a kind of a meta 273 00:13:22,800 --> 00:13:26,040 Speaker 2: documentary of sorts. Who was made by one person during 274 00:13:26,080 --> 00:13:29,520 Speaker 2: the course of one year. He was a musician, he 275 00:13:29,559 --> 00:13:32,720 Speaker 2: got approached by a production team. He's now making a 276 00:13:32,760 --> 00:13:36,280 Speaker 2: future film. He quit his musician like life, and now 277 00:13:36,280 --> 00:13:38,679 Speaker 2: he's like fully devoting himself just like film. He has 278 00:13:38,720 --> 00:13:42,959 Speaker 2: a team. None of that would have be impossible without 279 00:13:42,960 --> 00:13:45,640 Speaker 2: a eye within a runway, without this signalogy first, and 280 00:13:45,679 --> 00:13:48,679 Speaker 2: now he's doing what he always dreamed of doing because 281 00:13:48,679 --> 00:13:50,679 Speaker 2: of technology. And I found that just amazing. 282 00:13:50,800 --> 00:13:53,320 Speaker 1: On the other hand, there are plenty, plenty, plenty of 283 00:13:53,400 --> 00:13:57,240 Speaker 1: artists who will say to you, you're stealing generations of 284 00:13:57,280 --> 00:14:00,280 Speaker 1: people's work, You're running us out of business is the 285 00:14:00,280 --> 00:14:04,280 Speaker 1: greatest enemy to creative people in the whole world. What 286 00:14:04,320 --> 00:14:05,000 Speaker 1: do you say to them? 287 00:14:05,360 --> 00:14:08,560 Speaker 2: I say the first thing is like I think you 288 00:14:08,559 --> 00:14:12,440 Speaker 2: should you should deeply understand the all of you first. 289 00:14:12,600 --> 00:14:16,160 Speaker 2: I think if you haven't yet explored, you're missing a lot. 290 00:14:16,320 --> 00:14:17,640 Speaker 2: Like you're missing a lot, You're missing a lot of 291 00:14:17,640 --> 00:14:22,040 Speaker 2: context and nuance. You're probably like misunderstanding what's happening here. 292 00:14:22,400 --> 00:14:25,280 Speaker 2: I get this, like I've got this maybe when even 293 00:14:25,600 --> 00:14:29,200 Speaker 2: met the first time, Like AI is like taking creativity away. 294 00:14:29,840 --> 00:14:34,120 Speaker 2: I find that just like tropes so repetitive and like wrong, 295 00:14:34,640 --> 00:14:37,840 Speaker 2: like it means that people haven't yet experienced what this 296 00:14:37,920 --> 00:14:40,600 Speaker 2: thingolity can do for them, and you're going to figure 297 00:14:40,600 --> 00:14:45,040 Speaker 2: it out just by experiencing it like I can. We're 298 00:14:45,080 --> 00:14:47,720 Speaker 2: in Doha. I've never been to Doha. I can read 299 00:14:47,760 --> 00:14:49,440 Speaker 2: all I want about it. I can like talk to 300 00:14:49,520 --> 00:14:52,200 Speaker 2: anyone about it. There's only one way could know what 301 00:14:52,280 --> 00:14:54,640 Speaker 2: it means to be there. There's only one way you 302 00:14:54,680 --> 00:14:57,080 Speaker 2: could know, which is like, go there and you'll know 303 00:14:57,120 --> 00:14:58,720 Speaker 2: in the first ten minutes what it means to be 304 00:14:58,720 --> 00:15:00,920 Speaker 2: in Doha. This is the first same thing. If you 305 00:15:00,960 --> 00:15:03,080 Speaker 2: want to criticize AI, if you want to, like think 306 00:15:03,120 --> 00:15:05,400 Speaker 2: about it. If you know how it works and everything out, 307 00:15:05,480 --> 00:15:07,640 Speaker 2: but you haven't used it, then like you're missing a 308 00:15:07,640 --> 00:15:08,840 Speaker 2: lot some ways. 309 00:15:08,840 --> 00:15:11,600 Speaker 1: A cool subject to a class action lawsuit. You're part 310 00:15:11,600 --> 00:15:14,320 Speaker 1: of the class of AI companies who are being sued 311 00:15:14,360 --> 00:15:18,240 Speaker 1: by a graphic illustrator called Sarah Anderson and others. Does 312 00:15:18,240 --> 00:15:19,080 Speaker 1: that keep you up at night? 313 00:15:20,240 --> 00:15:23,440 Speaker 2: No, I can't really speak about like ongoing stuff like that. 314 00:15:23,560 --> 00:15:25,720 Speaker 2: But I think again, like we're trying to engage with 315 00:15:25,760 --> 00:15:28,680 Speaker 2: the community to show them how things really work. There's 316 00:15:28,680 --> 00:15:31,040 Speaker 2: a lot of misconceptions about the models actually do. 317 00:15:31,960 --> 00:15:33,800 Speaker 1: So what is what is when people say, look, it's 318 00:15:33,880 --> 00:15:37,800 Speaker 1: takes the whole corpus of human creativity, learns rules from 319 00:15:37,800 --> 00:15:41,760 Speaker 1: it and then applies those rules. That is a fair explanation. 320 00:15:41,840 --> 00:15:45,320 Speaker 1: What's unfair is to say it's stealing individual ideas or 321 00:15:45,360 --> 00:15:48,000 Speaker 1: segments or how do you characterize. 322 00:15:47,600 --> 00:15:50,920 Speaker 2: Look like think about like these are theseel models I learn? Right, 323 00:15:50,960 --> 00:15:52,320 Speaker 2: if you go you need to go to the basis 324 00:15:52,320 --> 00:15:55,120 Speaker 2: of whether you actually doing to understand like that that 325 00:15:55,120 --> 00:15:58,640 Speaker 2: that argument, which is models learned from the world. So 326 00:15:58,720 --> 00:16:01,040 Speaker 2: when when you want to start film, there's one thing 327 00:16:01,040 --> 00:16:03,200 Speaker 2: you should do, which is well and watch films. You're 328 00:16:03,200 --> 00:16:04,960 Speaker 2: gonna watch films and you're going to understand the language 329 00:16:04,960 --> 00:16:08,440 Speaker 2: of film. The language of film has certain taxonomies and 330 00:16:08,480 --> 00:16:11,440 Speaker 2: like specific ways of cutting things, And there's a pattern 331 00:16:11,520 --> 00:16:14,040 Speaker 2: in most of the stories. There's a beginning, a middle, 332 00:16:14,080 --> 00:16:17,760 Speaker 2: and an end. There's characters, there's different scenes. Right, you 333 00:16:17,800 --> 00:16:22,000 Speaker 2: build some sort of a world understanding of this industry 334 00:16:22,040 --> 00:16:25,320 Speaker 2: by watching films. An AI model that's exactly the same. 335 00:16:25,400 --> 00:16:28,280 Speaker 2: It watches things and understand the taxonomy of things. Now, 336 00:16:28,320 --> 00:16:30,520 Speaker 2: if I ask you to make a film, you're going 337 00:16:30,560 --> 00:16:32,760 Speaker 2: to take your knowledge of what you've seen, also your 338 00:16:32,760 --> 00:16:35,640 Speaker 2: experiences of your life and what you've seen in your 339 00:16:35,760 --> 00:16:38,440 Speaker 2: like pastime and everything you've seen and everyone you met, 340 00:16:38,480 --> 00:16:40,520 Speaker 2: and you're gonna make something out of it. A model 341 00:16:40,560 --> 00:16:42,640 Speaker 2: works pretty much the same way. You show them data, 342 00:16:42,640 --> 00:16:44,560 Speaker 2: and the model learns about the data, learns about the 343 00:16:44,560 --> 00:16:47,800 Speaker 2: patterns of the data, and uses that to create something. Now, 344 00:16:47,800 --> 00:16:50,640 Speaker 2: if I tell you what movie or what part of 345 00:16:50,680 --> 00:16:54,400 Speaker 2: your life influence that particular scene or that particular frame, 346 00:16:54,440 --> 00:16:58,119 Speaker 2: you'll be like, it's a mix of things. It's combining 347 00:16:58,200 --> 00:17:00,480 Speaker 2: things that I've learned over time. Right, the model is 348 00:17:00,520 --> 00:17:03,880 Speaker 2: the same and so arguing that that you can you 349 00:17:03,920 --> 00:17:08,040 Speaker 2: know that you can pinpoint that there was one particular 350 00:17:08,080 --> 00:17:09,840 Speaker 2: thing that influenced the model is kind of missing the 351 00:17:09,840 --> 00:17:12,359 Speaker 2: point because there's nothing in particular that influenced the model. 352 00:17:12,760 --> 00:17:15,320 Speaker 2: Is the aggregated date of the world that influences the model. 353 00:17:29,880 --> 00:17:33,040 Speaker 1: After the break why the history of Hollywood is the 354 00:17:33,080 --> 00:17:37,080 Speaker 1: history of technology and why Christobal believes that AI fits 355 00:17:37,359 --> 00:18:02,800 Speaker 1: right in stay with us. Do you think we'll see 356 00:18:02,840 --> 00:18:07,000 Speaker 1: people like Matthew McConaughey trademarking the image as a kind 357 00:18:07,040 --> 00:18:09,159 Speaker 1: of a new facet of what it is, Like. 358 00:18:09,280 --> 00:18:12,879 Speaker 2: I think that's happening already. Yeah, I mean it makes sense. 359 00:18:13,200 --> 00:18:15,960 Speaker 2: I think like sometimes forget specifally within Hollywood and median 360 00:18:16,119 --> 00:18:18,520 Speaker 2: art that like the history of Hollywood is the history 361 00:18:18,560 --> 00:18:21,880 Speaker 2: of technology, like the history of filmmaking is the history 362 00:18:21,920 --> 00:18:26,960 Speaker 2: of technology. Does because hollywoods was born out of technology 363 00:18:27,040 --> 00:18:29,800 Speaker 2: a breakthrough, which is the camera. One hundred and fifty 364 00:18:29,840 --> 00:18:32,359 Speaker 2: one hundred and sixty years ago, the ability for us 365 00:18:32,359 --> 00:18:35,560 Speaker 2: to capture light in a photo sensible device wasn't fysible, 366 00:18:36,200 --> 00:18:39,480 Speaker 2: and scientists at the time figure out a way of 367 00:18:39,600 --> 00:18:43,440 Speaker 2: capturing rays of light in photo sensible materials that gave 368 00:18:43,480 --> 00:18:46,400 Speaker 2: birth to cameras obscuras, and that gave birth to cameras, 369 00:18:46,400 --> 00:18:48,800 Speaker 2: and that get bursts to moving images, and that get 370 00:18:48,800 --> 00:18:52,640 Speaker 2: births to moving pictures and eventually, like an industry known 371 00:18:52,680 --> 00:18:55,800 Speaker 2: as hollywoods was born in the twenties, nineteen twenties, the 372 00:18:56,000 --> 00:18:59,160 Speaker 2: art of film was born out of science. Like those 373 00:18:59,200 --> 00:19:03,240 Speaker 2: two things are intertwined, and every single movie that you 374 00:19:03,280 --> 00:19:07,959 Speaker 2: watch today is a technical masterpiece. It uses breakthrough technologies 375 00:19:07,960 --> 00:19:11,359 Speaker 2: and like visual effects in CGI and streaming and transmitting, 376 00:19:11,400 --> 00:19:16,240 Speaker 2: like media in organizing, like you're making something with science. 377 00:19:16,800 --> 00:19:19,040 Speaker 2: Film is a huge of technology, and this is an 378 00:19:19,080 --> 00:19:22,000 Speaker 2: evolution of the same. Like there's nothing that has we 379 00:19:22,200 --> 00:19:26,679 Speaker 2: define that. We've set up a society to say films 380 00:19:26,680 --> 00:19:29,960 Speaker 2: should be this and nothing else because I say so, 381 00:19:30,000 --> 00:19:33,040 Speaker 2: it's like who says so? Like film was silent in 382 00:19:33,080 --> 00:19:35,639 Speaker 2: the twenties, and when audio was around because of a 383 00:19:35,680 --> 00:19:40,760 Speaker 2: technological breakthrough, people revolved, Like people were really against audio movies, 384 00:19:41,040 --> 00:19:42,879 Speaker 2: and there we go. I think it was a big 385 00:19:43,119 --> 00:19:46,119 Speaker 2: plus to have audio movies, and the same with analog 386 00:19:46,200 --> 00:19:48,600 Speaker 2: to de though, like people were against transitioning to UGO 387 00:19:48,680 --> 00:19:51,359 Speaker 2: cameras for whatever reason. I think some people might have 388 00:19:51,400 --> 00:19:53,920 Speaker 2: strong reasons about it, but I think It's like digital 389 00:19:54,000 --> 00:19:55,840 Speaker 2: cameras are just way more accessible to anyone else, to 390 00:19:56,200 --> 00:19:58,720 Speaker 2: many more people, and many of the greatest films were 391 00:19:58,720 --> 00:20:01,359 Speaker 2: born out of that. You can continue on that trend 392 00:20:01,400 --> 00:20:03,439 Speaker 2: and you'll realize it's just the same trend and the 393 00:20:03,440 --> 00:20:04,800 Speaker 2: same evolution we've always been. 394 00:20:05,320 --> 00:20:09,520 Speaker 1: You had a vision, and you had a first mover advantage. Right, 395 00:20:09,560 --> 00:20:12,160 Speaker 1: You've been at this since twenty eighteen, and you mentioned 396 00:20:12,160 --> 00:20:14,800 Speaker 1: twenty twenty three the New York Times story about the 397 00:20:14,800 --> 00:20:18,800 Speaker 1: first example of you know, true AI led video generation. 398 00:20:19,640 --> 00:20:22,040 Speaker 1: Of course since then, you know, there are some big 399 00:20:22,080 --> 00:20:25,800 Speaker 1: gorillas who have entered the enclosure so to speak. You know, 400 00:20:25,800 --> 00:20:31,399 Speaker 1: it should come beating that chess, Google and open AI. 401 00:20:31,720 --> 00:20:34,040 Speaker 1: How do you how do you compete? 402 00:20:34,640 --> 00:20:38,040 Speaker 2: Yeah? I mean, first of all, I think when we started, 403 00:20:38,080 --> 00:20:40,400 Speaker 2: if we realize if we're going to be something that's 404 00:20:40,440 --> 00:20:42,280 Speaker 2: going to change the world, that's going to be meaningful, 405 00:20:42,400 --> 00:20:44,479 Speaker 2: then like these gorillas are going to pop up, Like 406 00:20:44,600 --> 00:20:46,880 Speaker 2: if we're not, then they're not going to be interested enough. 407 00:20:46,920 --> 00:20:48,560 Speaker 2: This whole thing is not going to be interesting enough. 408 00:20:48,760 --> 00:20:51,960 Speaker 2: They don't start paying more attention. And I think they are, 409 00:20:52,359 --> 00:20:55,200 Speaker 2: which means that yes, this is like very interesting. It's 410 00:20:55,240 --> 00:20:56,840 Speaker 2: a huge opportunity. 411 00:20:56,400 --> 00:20:59,320 Speaker 1: But they have capital, they have engineering talent, and they 412 00:20:59,359 --> 00:21:00,000 Speaker 1: both have distribute. 413 00:21:00,520 --> 00:21:03,440 Speaker 2: Sure, yet we're still winning, so we have something. I 414 00:21:03,440 --> 00:21:05,880 Speaker 2: guess we're still we have better models than them technically, 415 00:21:05,960 --> 00:21:08,800 Speaker 2: so I'm not saying that they're not capable. They're really 416 00:21:08,800 --> 00:21:11,000 Speaker 2: strong teams that are like really thinking about the problem 417 00:21:11,040 --> 00:21:13,959 Speaker 2: and I think really good. But there's something around like 418 00:21:14,040 --> 00:21:17,159 Speaker 2: speed and taste that it's really hard to get in 419 00:21:17,200 --> 00:21:18,800 Speaker 2: a large company. And so what I mean by that 420 00:21:18,920 --> 00:21:21,480 Speaker 2: is I think taste. People think about it as like aesthetics, 421 00:21:21,480 --> 00:21:24,080 Speaker 2: you know, like the things you like visually or your 422 00:21:24,200 --> 00:21:27,480 Speaker 2: choices that you made. I think of taste as within 423 00:21:27,560 --> 00:21:30,639 Speaker 2: the spectrum of opportunities and directions that you can take 424 00:21:30,720 --> 00:21:33,639 Speaker 2: a product, a roadmap or research direction, which is are 425 00:21:33,680 --> 00:21:38,000 Speaker 2: you going to choose and why? And we've chosen things 426 00:21:38,040 --> 00:21:41,400 Speaker 2: that I think most people three years, four years ago 427 00:21:41,440 --> 00:21:44,760 Speaker 2: wouldn't have chosen, such as such as investing time and 428 00:21:44,800 --> 00:21:47,600 Speaker 2: training video models or making the real time right. And 429 00:21:47,640 --> 00:21:50,280 Speaker 2: there are things we're working on right now that feel 430 00:21:50,280 --> 00:21:53,920 Speaker 2: again like that. And the incentive for a large company 431 00:21:54,000 --> 00:21:56,439 Speaker 2: is not to be a different tier of innovating and 432 00:21:56,520 --> 00:21:59,879 Speaker 2: creating that which is a different mindset is to just 433 00:22:00,080 --> 00:22:01,800 Speaker 2: be on the look at the things that are working 434 00:22:01,840 --> 00:22:04,280 Speaker 2: and try to like win by copying them. Right, So 435 00:22:04,640 --> 00:22:08,399 Speaker 2: when you're making a decision, you're always between exploration and exploitation. 436 00:22:09,119 --> 00:22:12,480 Speaker 2: So if you have to figure out where you want 437 00:22:12,480 --> 00:22:15,240 Speaker 2: to get a coffee tomorrow, you can enter into exploit 438 00:22:15,440 --> 00:22:18,080 Speaker 2: or explore mode. Exploit mode it means you're going to 439 00:22:18,119 --> 00:22:20,040 Speaker 2: go to the coffee shop you always know because it's 440 00:22:20,080 --> 00:22:22,440 Speaker 2: fast and it's cheap, and you're you know what prices, 441 00:22:22,480 --> 00:22:24,760 Speaker 2: and you're not going to like spending more time figure 442 00:22:24,800 --> 00:22:26,720 Speaker 2: out anything else. You know exactly how to get there. 443 00:22:26,720 --> 00:22:29,480 Speaker 2: It's very efficient. Explore mode is what happens if you 444 00:22:29,520 --> 00:22:31,440 Speaker 2: go to the other part of the city you've never 445 00:22:31,480 --> 00:22:34,120 Speaker 2: been and you found a coffee shop that's ten times 446 00:22:34,160 --> 00:22:37,600 Speaker 2: better and cheaper, and also you meet like your wife, right, Like, 447 00:22:37,920 --> 00:22:39,960 Speaker 2: there's opportunities of things that are going to get there 448 00:22:40,440 --> 00:22:43,040 Speaker 2: depending on which choice you want to make, and you 449 00:22:43,040 --> 00:22:46,000 Speaker 2: can think about that same exploit explore mode when it 450 00:22:46,040 --> 00:22:49,240 Speaker 2: comes to research. So there's a lot of exploit more 451 00:22:49,680 --> 00:22:51,840 Speaker 2: when it comes to like models today where people are 452 00:22:51,840 --> 00:22:54,440 Speaker 2: trying to make them efficient, more good and whatever, and 453 00:22:54,480 --> 00:22:56,199 Speaker 2: there's a lot of explore mode, which is like, what 454 00:22:56,280 --> 00:22:58,240 Speaker 2: are the things can you do that you have not 455 00:22:58,600 --> 00:23:01,520 Speaker 2: even thought of before? And I think that has a 456 00:23:01,640 --> 00:23:05,000 Speaker 2: cultural element. You have to be disposed as an institution, 457 00:23:05,080 --> 00:23:07,399 Speaker 2: as a company to try to go to explore mode. 458 00:23:07,840 --> 00:23:09,720 Speaker 1: Twenty twenty six has been a year when people have 459 00:23:09,760 --> 00:23:12,440 Speaker 1: really started to talk about this idea of world models, 460 00:23:12,440 --> 00:23:16,040 Speaker 1: and you have indicated in a couple of different places that, 461 00:23:16,119 --> 00:23:20,959 Speaker 1: in a sense, you're building a company around that opportunity. Explain, 462 00:23:21,000 --> 00:23:22,520 Speaker 1: explain what it is and what it means to you. 463 00:23:22,600 --> 00:23:25,240 Speaker 2: Sure, yeah, So the first thing you need to understand 464 00:23:25,280 --> 00:23:27,760 Speaker 2: is like video model, So how do video models learn? 465 00:23:27,760 --> 00:23:29,280 Speaker 2: And this goes back to some of the things we're 466 00:23:29,320 --> 00:23:33,760 Speaker 2: just talking about. So video models are this incredible powerful 467 00:23:33,800 --> 00:23:37,080 Speaker 2: systems that can ingest data, learn around the world and 468 00:23:37,160 --> 00:23:40,399 Speaker 2: then make consistent videos out of the learnings of what 469 00:23:40,520 --> 00:23:42,679 Speaker 2: they've seen in the world. Right, which is not that 470 00:23:42,720 --> 00:23:44,919 Speaker 2: dissimilar to what we do. So if you want to 471 00:23:45,040 --> 00:23:47,320 Speaker 2: have a bottle of water here and it can open 472 00:23:47,320 --> 00:23:49,800 Speaker 2: the bottle and drop some water and you will know 473 00:23:50,320 --> 00:23:52,520 Speaker 2: what should happen. You can predict in your head what's 474 00:23:52,520 --> 00:23:54,119 Speaker 2: going to happen with the water, right, how it's going 475 00:23:54,160 --> 00:23:56,800 Speaker 2: to flow. I'm sure you can predict and you can 476 00:23:56,880 --> 00:23:59,760 Speaker 2: tell me the mathematical formula that describes how the water 477 00:24:00,080 --> 00:24:03,240 Speaker 2: flows in the table right. And what we've realized is 478 00:24:03,280 --> 00:24:06,480 Speaker 2: that video models basically are doing the same. You can 479 00:24:06,520 --> 00:24:09,280 Speaker 2: show them a photo of the water and tell them 480 00:24:09,280 --> 00:24:12,480 Speaker 2: to predict what should happen next if the water comes down, 481 00:24:12,600 --> 00:24:14,640 Speaker 2: and they're going to do it with such a curacy 482 00:24:14,680 --> 00:24:17,440 Speaker 2: that it looks real, so real. Yet the model has 483 00:24:17,480 --> 00:24:21,439 Speaker 2: no mathematical physical knowledge of why that should happen. You 484 00:24:21,440 --> 00:24:23,720 Speaker 2: just knows it happens because I've seen enough things of 485 00:24:23,760 --> 00:24:26,439 Speaker 2: the world. And so when you take that idea further, 486 00:24:26,520 --> 00:24:30,720 Speaker 2: you start realizing that video models eventually are just really good, 487 00:24:30,760 --> 00:24:33,640 Speaker 2: generalizable machines around the world that can create their own 488 00:24:33,680 --> 00:24:36,200 Speaker 2: world models. They can understand and create their own predictions 489 00:24:36,200 --> 00:24:38,520 Speaker 2: of the world at large. And so if you train 490 00:24:38,600 --> 00:24:43,439 Speaker 2: these models bigger and like better, these video models are 491 00:24:43,480 --> 00:24:46,320 Speaker 2: eventually becoming world models. They can model the world and 492 00:24:46,359 --> 00:24:49,200 Speaker 2: they can understand what actions should take in a world. 493 00:24:49,359 --> 00:24:52,000 Speaker 2: But you can also influence those things in the world. 494 00:24:52,600 --> 00:24:54,720 Speaker 2: And I think for us that there's a lot of 495 00:24:54,720 --> 00:24:59,560 Speaker 2: opportunities for product and growth and creativity that will steam 496 00:24:59,560 --> 00:25:04,159 Speaker 2: from models are not just like brute forcing reality, but 497 00:25:04,400 --> 00:25:07,560 Speaker 2: understanding the world and then making decisions around the world. 498 00:25:08,160 --> 00:25:11,840 Speaker 1: So today's AI videos tend to kind of lose coherence 499 00:25:11,960 --> 00:25:15,040 Speaker 1: or the laws of the physical world and not always 500 00:25:15,080 --> 00:25:18,440 Speaker 1: obeyed in AI videos. Right, is that a fair assessment. 501 00:25:18,760 --> 00:25:21,520 Speaker 2: I think that's like a temporary, like I would say, 502 00:25:21,920 --> 00:25:25,119 Speaker 2: condition of the current way models are trained. But you 503 00:25:25,160 --> 00:25:28,760 Speaker 2: can now train large, long sequence video models that don't 504 00:25:28,960 --> 00:25:31,080 Speaker 2: lose consistency. That's possible. 505 00:25:31,119 --> 00:25:35,000 Speaker 1: So these video models are starting to intuit or to 506 00:25:35,080 --> 00:25:37,640 Speaker 1: learn without being taught laws of physics. 507 00:25:37,640 --> 00:25:38,000 Speaker 2: Correct. 508 00:25:38,560 --> 00:25:41,520 Speaker 1: Correct, you said AI lambs have been very obsessed with 509 00:25:41,520 --> 00:25:44,119 Speaker 1: simulating the human mind, but I think it might be 510 00:25:44,119 --> 00:25:46,600 Speaker 1: the wrong approach long term. What you want to do 511 00:25:46,640 --> 00:25:50,119 Speaker 1: is not simulate how humans work, but how the world works. 512 00:25:50,160 --> 00:25:50,560 Speaker 2: Correct? 513 00:25:50,920 --> 00:25:51,560 Speaker 1: What does that mean? 514 00:25:52,040 --> 00:25:55,440 Speaker 2: So think about language? Right, So we're having a conversation. 515 00:25:55,600 --> 00:25:58,920 Speaker 2: Language is a human made symbolic system that we've created 516 00:25:58,960 --> 00:26:02,840 Speaker 2: on top of reality. Language doesn't describe reality. Se describes 517 00:26:02,880 --> 00:26:05,840 Speaker 2: a way we can refer to it in a way. Right, 518 00:26:06,240 --> 00:26:08,840 Speaker 2: But there's natural opreservations in the world that might not 519 00:26:08,920 --> 00:26:11,320 Speaker 2: have words that we can describe. And so one thing 520 00:26:11,320 --> 00:26:13,119 Speaker 2: we've done is that we've trained and I mean the 521 00:26:13,119 --> 00:26:16,200 Speaker 2: industry at large has train models on the symbolic obstruction 522 00:26:16,240 --> 00:26:18,760 Speaker 2: of reality, which is language. Right, what we're saying is 523 00:26:18,800 --> 00:26:20,960 Speaker 2: like you can go deeper than that. You can train 524 00:26:21,040 --> 00:26:23,359 Speaker 2: on reality. You don't have to train the model on 525 00:26:23,480 --> 00:26:25,600 Speaker 2: like what the meaning of the things are, or what 526 00:26:25,640 --> 00:26:27,920 Speaker 2: the meaning of a particular thing is. You can train 527 00:26:28,040 --> 00:26:30,359 Speaker 2: on the substrate of reality, which is just capture the 528 00:26:30,400 --> 00:26:34,360 Speaker 2: world at large. See how physics works, how water moves, 529 00:26:34,600 --> 00:26:37,639 Speaker 2: not by telling the model literally water flows that this 530 00:26:37,760 --> 00:26:42,040 Speaker 2: particular float and density and whatever you're showing the video water. 531 00:26:41,960 --> 00:26:43,680 Speaker 1: Learning visually rather than through words. 532 00:26:44,520 --> 00:26:47,560 Speaker 2: Correct, it's learning through experience. And then what you could 533 00:26:47,640 --> 00:26:50,199 Speaker 2: do is technically and this works already when we're doing it. 534 00:26:50,760 --> 00:26:53,719 Speaker 2: If you want to take you want to create a robot, 535 00:26:53,840 --> 00:26:56,040 Speaker 2: You want to create a robot that can pick this 536 00:26:56,280 --> 00:26:58,400 Speaker 2: water bottle that I have on the table. The way 537 00:26:58,440 --> 00:27:02,800 Speaker 2: you do that today is to show the robot thousands 538 00:27:02,840 --> 00:27:06,399 Speaker 2: if not tenth of thousands of arms picking the bottle, 539 00:27:07,080 --> 00:27:10,360 Speaker 2: and then you train a policy which is a specific 540 00:27:10,520 --> 00:27:14,680 Speaker 2: robotic model that then has learned watching those videos right, 541 00:27:15,080 --> 00:27:16,760 Speaker 2: and in fact they're in China. 542 00:27:16,800 --> 00:27:19,080 Speaker 1: I think a lot of sort of physical locations where 543 00:27:19,080 --> 00:27:21,879 Speaker 1: there are hundreds of humans during action so that robots 544 00:27:21,880 --> 00:27:22,960 Speaker 1: can learn from the warehouses. 545 00:27:22,960 --> 00:27:25,240 Speaker 2: They're not kidding about this. There are warehouses with people 546 00:27:25,320 --> 00:27:29,119 Speaker 2: with GoPros stuck to their heads that everything they do 547 00:27:29,160 --> 00:27:32,560 Speaker 2: from nine am to five pm is fold clothes. So 548 00:27:32,600 --> 00:27:34,720 Speaker 2: you go to a table, you stuck a camera on 549 00:27:34,800 --> 00:27:37,240 Speaker 2: your head, and you start folding clothes, and you fold 550 00:27:37,280 --> 00:27:38,879 Speaker 2: it and then you unfold it, and you folded and 551 00:27:38,920 --> 00:27:42,520 Speaker 2: you that's training data for the models to understand how 552 00:27:42,560 --> 00:27:43,320 Speaker 2: to fold clothes. 553 00:27:43,440 --> 00:27:44,960 Speaker 1: How does that differ from what you're doing. 554 00:27:45,200 --> 00:27:48,120 Speaker 2: I can synthetically generate all of that, so I don't 555 00:27:48,119 --> 00:27:50,040 Speaker 2: have to have a human show me how to fold 556 00:27:50,040 --> 00:27:54,320 Speaker 2: the clothes. I can have an AI model simulate that. 557 00:27:54,720 --> 00:27:58,040 Speaker 2: I can have a video model create a synthetic version 558 00:27:58,080 --> 00:28:00,919 Speaker 2: of that video and a million other video and we 559 00:28:00,960 --> 00:28:04,040 Speaker 2: can use that data to train the robot. And here's 560 00:28:04,040 --> 00:28:07,080 Speaker 2: the interesting thing. So you've now cut the time it 561 00:28:07,160 --> 00:28:09,600 Speaker 2: takes to train the model because you don't have to 562 00:28:09,680 --> 00:28:11,879 Speaker 2: get the data. You can just synthetically generate the data. 563 00:28:12,280 --> 00:28:14,480 Speaker 2: But then once you've created the policy of the robot 564 00:28:14,480 --> 00:28:17,280 Speaker 2: that can grab the thing, you need to test it 565 00:28:17,320 --> 00:28:19,040 Speaker 2: in their world. So the way you tested is you 566 00:28:19,080 --> 00:28:20,800 Speaker 2: put the policy in the robot and you see how 567 00:28:20,840 --> 00:28:24,399 Speaker 2: effective it is. You can actually do that testing in 568 00:28:24,440 --> 00:28:29,520 Speaker 2: the simulation as well. So you're gathering the data, you're 569 00:28:29,560 --> 00:28:32,879 Speaker 2: training us policy, the model learns the policy, and then 570 00:28:32,920 --> 00:28:35,800 Speaker 2: you're testing how effected the policy actually is by deploying 571 00:28:35,840 --> 00:28:38,080 Speaker 2: again on the simulation. So it's simulation to simulation. 572 00:28:38,480 --> 00:28:42,280 Speaker 1: But every policy has to be individually defined by a 573 00:28:42,360 --> 00:28:46,360 Speaker 1: human like or is there some extrapolation possible where once 574 00:28:46,440 --> 00:28:48,800 Speaker 1: my robot has learned to pick up the bottle of water, 575 00:28:48,880 --> 00:28:51,760 Speaker 1: it can also pick up, you know, the microphone. 576 00:28:51,760 --> 00:28:55,160 Speaker 2: That's generalization. So that's a greatly something. The idea of 577 00:28:55,240 --> 00:28:57,120 Speaker 2: like journally, of models is you want them to have 578 00:28:57,160 --> 00:28:59,920 Speaker 2: the ability to journalize, right, to take one knowledge and 579 00:29:00,040 --> 00:29:02,760 Speaker 2: apply them to somewhere else. But again that journal decision 580 00:29:02,840 --> 00:29:06,280 Speaker 2: comes from scale, from training at larger data points, and 581 00:29:06,360 --> 00:29:09,000 Speaker 2: gathering data becomes a bottleneck. That's why people are investing 582 00:29:09,040 --> 00:29:12,000 Speaker 2: so much in gathering data or companies that can help 583 00:29:12,480 --> 00:29:14,960 Speaker 2: us in creating data our reports is that you can 584 00:29:15,040 --> 00:29:17,840 Speaker 2: synthetically generate all of the data. Think about another problem 585 00:29:17,920 --> 00:29:20,320 Speaker 2: still driving cars. If you want to have cars who 586 00:29:20,320 --> 00:29:22,160 Speaker 2: can drive in the streets, one of the hardest things 587 00:29:22,200 --> 00:29:26,960 Speaker 2: to train the models are is on collisions because collisions 588 00:29:27,000 --> 00:29:29,960 Speaker 2: are very hard to like create, I can just go 589 00:29:30,000 --> 00:29:32,000 Speaker 2: and create a collision of a car hitting a person 590 00:29:32,720 --> 00:29:36,640 Speaker 2: because it's technically hard, right, and there's probably not enough 591 00:29:36,720 --> 00:29:40,080 Speaker 2: data out there that shows me collisions. So wouldn't it 592 00:29:40,080 --> 00:29:42,160 Speaker 2: be great if you can just simulate a collision and 593 00:29:42,160 --> 00:29:44,240 Speaker 2: the model can learn about what happens in a collision. 594 00:29:44,520 --> 00:29:45,360 Speaker 2: And that's what we're doing. 595 00:29:46,080 --> 00:29:49,320 Speaker 1: You said, soon everyone will have access to their own 596 00:29:49,320 --> 00:29:53,680 Speaker 1: world simulator. This will be the most important technological development 597 00:29:53,760 --> 00:29:59,000 Speaker 1: of our time. That sounds to me bigger than innovation 598 00:29:59,040 --> 00:30:01,040 Speaker 1: in Hollywood. Yeah. Absolutely. 599 00:30:01,440 --> 00:30:03,800 Speaker 2: I mean it's a general purpose like technology being able 600 00:30:03,840 --> 00:30:07,280 Speaker 2: to simulate how a car moves, how robot works, how 601 00:30:07,400 --> 00:30:12,120 Speaker 2: entertainment works, how games are played. We have done experiments 602 00:30:12,120 --> 00:30:15,560 Speaker 2: where you can simulate computer interfaces. So when you log 603 00:30:15,600 --> 00:30:17,360 Speaker 2: in your computer where you see the screen, I can 604 00:30:17,480 --> 00:30:21,440 Speaker 2: generate that screen and it can create synthetically, like on 605 00:30:21,480 --> 00:30:24,640 Speaker 2: the fly, real times like interactions for you on that screen. 606 00:30:24,960 --> 00:30:29,440 Speaker 2: You're basically creating simulation systems, and simulating is way cheaper 607 00:30:29,480 --> 00:30:32,000 Speaker 2: than deploying things in real time in the world. 608 00:30:32,520 --> 00:30:34,840 Speaker 1: And what is the biggest limitation on world model is 609 00:30:34,880 --> 00:30:37,200 Speaker 1: it chips? Is it models themselves? 610 00:30:37,280 --> 00:30:39,600 Speaker 2: The most its energy to supply chain is how you 611 00:30:39,640 --> 00:30:41,360 Speaker 2: scale the model is to do what they need to 612 00:30:41,360 --> 00:30:44,040 Speaker 2: do at scale. It's moving atoms from one place to 613 00:30:44,120 --> 00:30:47,600 Speaker 2: the other. I think the technical foundations are there, it's 614 00:30:47,680 --> 00:30:50,560 Speaker 2: now how fast and how good can you scale those 615 00:30:50,560 --> 00:30:53,360 Speaker 2: foundations to do what they can do? And so that's 616 00:30:53,400 --> 00:30:54,480 Speaker 2: where we're focus on right now. 617 00:30:54,720 --> 00:30:58,680 Speaker 1: And you released your first world model at Runway in December. Correct, 618 00:30:58,960 --> 00:31:00,880 Speaker 1: what does it do that other world models don't do? 619 00:31:01,080 --> 00:31:03,720 Speaker 2: So we took our latest video model four point five 620 00:31:04,120 --> 00:31:06,080 Speaker 2: and we're training in such a way that the model 621 00:31:06,120 --> 00:31:09,240 Speaker 2: can understand actions and you can interact with the model, 622 00:31:09,440 --> 00:31:12,520 Speaker 2: and the model has a world understanding, right, And so 623 00:31:12,760 --> 00:31:16,480 Speaker 2: we've basically released three main kind of core applications of it. 624 00:31:16,560 --> 00:31:19,040 Speaker 2: One is robotics, which is the one I was selling 625 00:31:19,040 --> 00:31:20,760 Speaker 2: about think about it is physical AI. This is the 626 00:31:20,800 --> 00:31:24,600 Speaker 2: first ever world model that has commercially been available that 627 00:31:24,640 --> 00:31:28,040 Speaker 2: does this that allows you to take data from our 628 00:31:28,120 --> 00:31:32,040 Speaker 2: robotics for example company or av car and create more 629 00:31:32,040 --> 00:31:35,000 Speaker 2: synthetic data for that. The bottomneg on AI and physically 630 00:31:35,000 --> 00:31:38,360 Speaker 2: AI is data and gathering data in the real world 631 00:31:38,520 --> 00:31:41,880 Speaker 2: is very expensive. It's very hard. It's very time consuming. 632 00:31:42,120 --> 00:31:44,400 Speaker 2: So unless we figure out something, it's very hard to 633 00:31:44,440 --> 00:31:47,560 Speaker 2: scale that. So that's one. The other one is what 634 00:31:47,600 --> 00:31:50,960 Speaker 2: we call agent or real time agents. So real time 635 00:31:51,000 --> 00:31:53,600 Speaker 2: agents are a layer of our world models that can 636 00:31:53,600 --> 00:31:56,920 Speaker 2: function in real time. So what we do is imagine 637 00:31:56,920 --> 00:31:59,600 Speaker 2: you're logging into a computer. You click one button and 638 00:31:59,640 --> 00:32:02,520 Speaker 2: there's an avatar that comes and that speaks with you, 639 00:32:02,560 --> 00:32:04,920 Speaker 2: and it's similar to how we're speaking now. It's a 640 00:32:05,440 --> 00:32:08,880 Speaker 2: human or an animated character. It looks incredible real I 641 00:32:09,040 --> 00:32:11,560 Speaker 2: responds to you in real time and he knows everything 642 00:32:11,600 --> 00:32:14,000 Speaker 2: you're watching and everything you're seeing. So think about I 643 00:32:14,040 --> 00:32:16,520 Speaker 2: don't know and I want to learn physics. I'm a student, 644 00:32:16,560 --> 00:32:18,160 Speaker 2: I want to learn math, I want to learn whatever 645 00:32:18,160 --> 00:32:20,280 Speaker 2: I want to learn. I can now have a personalized 646 00:32:20,280 --> 00:32:23,840 Speaker 2: studor just for me. It's one thing that education has 647 00:32:23,880 --> 00:32:26,240 Speaker 2: proven way and over and time and time again, is 648 00:32:26,280 --> 00:32:30,040 Speaker 2: that like the ratio of students to professor really matters, 649 00:32:30,040 --> 00:32:32,040 Speaker 2: and if everyone in the world can have one professor, 650 00:32:32,760 --> 00:32:35,960 Speaker 2: everyone's education will be so much better. This is literally 651 00:32:35,960 --> 00:32:37,640 Speaker 2: what you can do right now. And the third one, 652 00:32:37,640 --> 00:32:40,680 Speaker 2: which is like where we started, is we've deployed the 653 00:32:40,680 --> 00:32:44,200 Speaker 2: models in media and so entertainment, how do you create 654 00:32:44,440 --> 00:32:48,600 Speaker 2: end to win films or stories that are completely generated? 655 00:32:49,160 --> 00:32:52,000 Speaker 1: And that's like the end to end means a single prompt. 656 00:32:51,840 --> 00:32:55,400 Speaker 2: Correct more than a single prompt. But like ideally, yes, 657 00:32:55,480 --> 00:32:57,800 Speaker 2: you want to have a bit more fautomation and how 658 00:32:57,840 --> 00:32:58,280 Speaker 2: you make it? 659 00:32:58,360 --> 00:32:59,640 Speaker 1: Because I was going to ask you, how do you 660 00:32:59,680 --> 00:33:02,920 Speaker 1: The first two things you describe sound quite far from 661 00:33:03,520 --> 00:33:06,680 Speaker 1: the core of what you're famous for, which is helping 662 00:33:06,760 --> 00:33:09,239 Speaker 1: in the creation of media. So how do you, I mean, 663 00:33:09,240 --> 00:33:10,680 Speaker 1: how do you balance all these different. 664 00:33:10,520 --> 00:33:11,960 Speaker 2: Yeah, so this is where you go back to, Like, 665 00:33:12,040 --> 00:33:14,840 Speaker 2: remember the art, the history of art is extra of technology. 666 00:33:15,360 --> 00:33:19,320 Speaker 2: So the history of filmmaking was the history of scientists 667 00:33:19,360 --> 00:33:22,800 Speaker 2: in the nineteen beginning of the nineteen hundreds experimenting with 668 00:33:22,920 --> 00:33:28,360 Speaker 2: this idea of cameras. Right. Eventually the camera became known 669 00:33:28,440 --> 00:33:32,239 Speaker 2: as cinema and our new art form was born out 670 00:33:32,280 --> 00:33:36,200 Speaker 2: of it. But cameras these days are way more in 671 00:33:36,240 --> 00:33:39,040 Speaker 2: our life involved than just like filmmaking. You have cameras 672 00:33:39,080 --> 00:33:41,520 Speaker 2: pretty much everywhere. There's a camera in all the satellites 673 00:33:41,560 --> 00:33:44,160 Speaker 2: that are in space. There's comras in our cars, there's 674 00:33:44,200 --> 00:33:46,719 Speaker 2: commers in our houses, there's comers in our cell phones. 675 00:33:47,120 --> 00:33:49,240 Speaker 2: There's covers everywhere in the world. So the view of 676 00:33:49,280 --> 00:33:53,480 Speaker 2: the cameras a scientific breakthrough that was only using storytelling 677 00:33:53,480 --> 00:33:56,000 Speaker 2: has evolved to become a general purpose machine or a 678 00:33:56,000 --> 00:33:58,480 Speaker 2: tool that allows us to do way more. I think 679 00:33:58,520 --> 00:34:01,080 Speaker 2: of the work where doing not way in a versimilar way. 680 00:34:01,840 --> 00:34:04,400 Speaker 2: Many times describe runways a new kind of camera, and 681 00:34:04,440 --> 00:34:06,200 Speaker 2: when I say a camera, runways were a new kind 682 00:34:06,200 --> 00:34:09,960 Speaker 2: of camera because we're stimulating reality. The first applicable use 683 00:34:10,000 --> 00:34:13,520 Speaker 2: case was art. You went and you use this new camera, 684 00:34:13,560 --> 00:34:15,120 Speaker 2: and the same when they used the first camera to 685 00:34:15,120 --> 00:34:17,600 Speaker 2: make movies, to make stories, to make art. And then 686 00:34:17,640 --> 00:34:20,160 Speaker 2: you start realizing that the camera has way more applications 687 00:34:20,160 --> 00:34:24,720 Speaker 2: than just like filmmaking. It has ways of improving scientific progress. 688 00:34:25,320 --> 00:34:27,760 Speaker 2: We should totally be there, like we have the models 689 00:34:27,800 --> 00:34:31,399 Speaker 2: to allow us to do that. Why not, Chris, thank 690 00:34:31,440 --> 00:34:33,000 Speaker 2: you of course, appreciate your. 691 00:34:32,880 --> 00:34:33,560 Speaker 1: Time on test stuff. 692 00:34:33,560 --> 00:34:35,279 Speaker 2: We should do this again on former years to see 693 00:34:35,280 --> 00:34:36,160 Speaker 2: how much we progress. 694 00:34:36,200 --> 00:34:38,279 Speaker 1: I don't want to wait that long. Yeah, that's very 695 00:34:38,320 --> 00:34:38,680 Speaker 1: so much. 696 00:34:38,719 --> 00:34:39,040 Speaker 2: Thank you. 697 00:34:56,680 --> 00:34:59,040 Speaker 1: That's if it takes stuff this week, I'm as velocian. 698 00:34:59,560 --> 00:35:02,680 Speaker 1: This episod SO was produced by Eliza Dennis and Melissa Slaughter. 699 00:35:03,480 --> 00:35:06,800 Speaker 1: It was executive produced by Me, Karen Price, Julian Nutter, 700 00:35:06,840 --> 00:35:11,319 Speaker 1: and Kate Osborne for Kaleidoscope and Katrina nor velfa iHeart Podcasts. 701 00:35:11,880 --> 00:35:15,320 Speaker 1: The engineer is Matt Stillo and Jack Insley mixed this episode. 702 00:35:15,920 --> 00:35:19,560 Speaker 1: Kyle Murdoch wrote our theme song. Please do rate, review, 703 00:35:19,640 --> 00:35:22,040 Speaker 1: and reach out to us at tech Stuff podcast at 704 00:35:22,120 --> 00:35:24,120 Speaker 1: gmail dot com. We love hearing from you.