WEBVTT - The Story: User-Centered AI at Google w/ David Webster 0:00:14.840 --> 0:00:17.560 Welcome to tech Stuff. This is the story and today 0:00:17.720 --> 0:00:20.639 I'm here with Cara Price. Hi. So we were talking 0:00:20.680 --> 0:00:24.520 about these episodes of the story, and we realized that 0:00:24.600 --> 0:00:26.119 while the other of us we often have an in 0:00:26.200 --> 0:00:29.440 depth discussion about the guest or the episode, but that 0:00:29.440 --> 0:00:31.720 we weren't capturing that discussion for the podcast and that 0:00:31.720 --> 0:00:34.800 felt like a misopportunity. So today we're trying out something new, 0:00:35.000 --> 0:00:36.760 which is to have a bit of a chat between 0:00:36.760 --> 0:00:38.920 the two of us to intro the guests of the 0:00:38.920 --> 0:00:42.800 week and in particular why we think their story matters 0:00:42.800 --> 0:00:45.960 to us and to our listeners. So to start of 0:00:46.000 --> 0:00:49.280 this week, I wanted to share a thought from our guest, 0:00:49.479 --> 0:00:53.120 David Webster about the importance of always considering the user 0:00:53.440 --> 0:00:54.840 when developing new technology. 0:00:55.280 --> 0:00:57.600 And it's often a blind spot for the technologists. I mean, 0:00:57.600 --> 0:01:03.520 the technologists are not necessarily cold or dispersonal or or cynical. 0:01:03.760 --> 0:01:06.960 The probably parents of this new technology, but they might 0:01:06.959 --> 0:01:09.920 not realize that their babies their babies ugly. 0:01:11.120 --> 0:01:14.120 So, Cara, that was David Webster speaking. He's head of 0:01:14.560 --> 0:01:18.680 ux or User Experience, that's something called Google Labs, and 0:01:18.760 --> 0:01:20.800 in my opinion, he has one of the most interesting 0:01:20.880 --> 0:01:24.520 jobs at Google, because Labs is the division that's responsible 0:01:24.560 --> 0:01:27.840 for making consumer facing products that sit on top of 0:01:27.880 --> 0:01:30.240 Google's AI models. They've done stuff in the world of 0:01:30.319 --> 0:01:35.600 video generation, but probably their most famous product is Notebook LM. 0:01:35.720 --> 0:01:37.440 And that's the product that we thought was going to 0:01:37.480 --> 0:01:39.800 put us out of business one, the one where you 0:01:39.880 --> 0:01:43.520 can feed any websites, document articles you like, and get 0:01:43.600 --> 0:01:46.200 back a two way podcast. 0:01:45.840 --> 0:01:49.640 Two comforting voices, giggling over whatever the topic at hand 0:01:49.680 --> 0:01:52.840 may be. Much like that exactly. This notebook and AM 0:01:52.920 --> 0:01:55.920 product kind of grew people's minds and it had fans 0:01:56.120 --> 0:02:02.880 from Mark Cuban to AI pioneer Andre Pathy his Mark Cuban, has. 0:02:02.840 --> 0:02:07.560 Anybody played with Notebook l LAMB the podcast feature Oh 0:02:07.600 --> 0:02:08.480 my God. 0:02:08.639 --> 0:02:12.079 So you could take any PDF file, any text. 0:02:12.880 --> 0:02:15.000 It could be like your onboarding manual. 0:02:15.120 --> 0:02:18.680 They could be training manual, they could be distribution manuals, 0:02:18.680 --> 0:02:20.840 whatever it is, and you feed. 0:02:20.560 --> 0:02:22.880 It in there and it turns it into a podcast. 0:02:23.639 --> 0:02:27.000 Cuban is very fixtational manuals for whatever reason he is. 0:02:28.080 --> 0:02:31.440 But what does Webster's role have to do with any 0:02:31.480 --> 0:02:31.720 of this. 0:02:32.040 --> 0:02:36.000 Well, David Webster, as I mentioned, runs Google Labs. Previously, 0:02:36.280 --> 0:02:39.680 he worked for this design firm called Ideo, which is 0:02:39.760 --> 0:02:43.359 famous for the concept of human centered design, in other words, 0:02:43.440 --> 0:02:46.960 doing deep user research to understand what real people want 0:02:46.960 --> 0:02:50.320 from our product and then integrating that feedback into the product. 0:02:50.720 --> 0:02:53.320 Idea is actually the firm who designed the first mouse 0:02:53.360 --> 0:02:56.960 for Apple in the way back when, and Webster has 0:02:57.000 --> 0:03:00.519 now been tasked with making delightful, engaging a eye products 0:03:00.560 --> 0:03:01.560 for Google. Yeah. 0:03:01.560 --> 0:03:03.919 You know, it's an interesting moment for this conversation because 0:03:04.400 --> 0:03:08.800 Google is getting absolutely beaten up, relatively speaking, by the 0:03:08.840 --> 0:03:12.399 stock market because of questions about whether the ad supported 0:03:12.440 --> 0:03:15.840 search engine, which is fundamental to its business, will survive 0:03:15.880 --> 0:03:17.440 the age of AI exactly. 0:03:17.560 --> 0:03:20.440 And David is on the front lines of building the 0:03:20.520 --> 0:03:23.920 next wave of consumer products for Google. We spoke about 0:03:23.919 --> 0:03:27.680 the philosophy behind how he does that and about his 0:03:27.800 --> 0:03:31.399 views on the future of wearable AI. Here's the conversation. 0:03:32.240 --> 0:03:33.600 David Webster, thanks for joining us. 0:03:33.560 --> 0:03:34.880 Today, Thanks for having me. 0:03:35.000 --> 0:03:37.600 You know, when people think about Google, they obviously think 0:03:37.600 --> 0:03:40.440 about the search engine. In fact, you know, to Google 0:03:40.560 --> 0:03:44.720 became a verb much like to hoover, where we come from, 0:03:45.280 --> 0:03:49.800 became a verb for vacuum cleaning in Britain. You work 0:03:49.840 --> 0:03:53.760 at Google Labs as the head of UX, Well, what 0:03:53.840 --> 0:03:55.760 does labs do and how does it interact with a 0:03:55.840 --> 0:03:57.200 larger Google organization. 0:03:57.640 --> 0:04:02.160 Well, Labs is a group that we started in twenty 0:04:02.240 --> 0:04:06.120 twenty two. It is actually our reboot of a group 0:04:06.160 --> 0:04:08.520 that was around her in the early days of Google, 0:04:09.600 --> 0:04:15.520 which was a place where we could experiment with emergent 0:04:15.560 --> 0:04:19.360 capabilities and new technologies coming from the research side of 0:04:19.400 --> 0:04:25.719 the organization and figure out how to quickly productize those 0:04:26.640 --> 0:04:30.000 in new ways and service of Google's missions. So that's 0:04:30.000 --> 0:04:32.120 why we brought labs back into existence. 0:04:32.600 --> 0:04:36.000 Google's mission being is that still to organize the world's information. 0:04:35.760 --> 0:04:41.040 Organize the world's information and make it universally accessible and useful. 0:04:42.400 --> 0:04:45.240 I joined to stand up the UX component of labs, 0:04:45.760 --> 0:04:49.760 and the exciting thing for me was having had a 0:04:49.800 --> 0:04:54.599 career in design and UX UX standing of course for 0:04:54.720 --> 0:05:00.200 user experiences. I was very aware that we were in 0:05:00.240 --> 0:05:04.280 a moment where there was a host actually of unprecedentedly 0:05:04.279 --> 0:05:10.480 powerful technologies their time had arrived, and was very conscious 0:05:10.560 --> 0:05:17.000 of the significance of user experience and inflecting those technologies 0:05:17.080 --> 0:05:19.560 so that the prosu weigh the cons. There were a 0:05:19.640 --> 0:05:23.880 bunch of things emerging around twenty twenty two that were contenders. 0:05:23.960 --> 0:05:26.880 You know, the blockchain was still a thing. Metaverses were 0:05:26.920 --> 0:05:32.520 still a thing, but obviously llms their moment was arriving, 0:05:32.560 --> 0:05:36.919 and very quickly we ascertained that they were becoming the 0:05:37.080 --> 0:05:42.240 thing and started putting all our efforts into figuring out 0:05:42.240 --> 0:05:45.520 how to take the potential of this, you know, significant 0:05:45.520 --> 0:05:50.080 platform shift and format it so that it was accessible 0:05:50.480 --> 0:05:51.760 and useful to people. 0:05:51.960 --> 0:05:54.120 You just want to sort of go back in time 0:05:55.040 --> 0:05:57.640 thirty odd years to kind of give some framing to 0:05:57.680 --> 0:06:01.360 this conversation into your background. I'm wonders you can tell 0:06:01.440 --> 0:06:05.200 us who Kenji Eko one, if I'm saying that correctly, 0:06:05.240 --> 0:06:08.880 who he is, and what we can't understand about your 0:06:08.960 --> 0:06:10.400 work without understanding him. 0:06:10.920 --> 0:06:17.760 Kenji Ekuan. Kenji Ekuan was absolutely legendary and iconic a 0:06:17.880 --> 0:06:24.680 Japanese designer. He was the founder of a company called 0:06:24.720 --> 0:06:28.560 GK Design, which is a large Japanese group of design 0:06:28.600 --> 0:06:31.880 and creative companies, and in. 0:06:31.800 --> 0:06:33.600 The early days of my career. 0:06:36.480 --> 0:06:42.000 Had the privilege of getting to meet him quite serendipetitively. 0:06:42.720 --> 0:06:46.400 I studied mechanical engineering originally, then studied industrial design at 0:06:46.400 --> 0:06:49.120 the Royal College of Art. Wanted to be both a 0:06:49.160 --> 0:06:52.640 deeply technical person and a highly creative person. You had 0:06:52.640 --> 0:06:54.920 to kind of pick either or in the UK at 0:06:54.920 --> 0:06:59.440 the time, so I was attracted to and moved to 0:06:59.480 --> 0:07:01.840 a place where where you didn't have to pick either or, 0:07:02.160 --> 0:07:05.680 and one of those places was Japan. And so I 0:07:05.760 --> 0:07:08.320 rolled up in Japan, but I didn't have a job 0:07:08.360 --> 0:07:11.800 lined up, and via a friend of mine who was 0:07:11.840 --> 0:07:15.160 already in Japan, I got an introduction to Kenji Ekuan, 0:07:15.360 --> 0:07:19.040 who was this revered chairman of the most iconic design 0:07:19.080 --> 0:07:22.000 firm in Japan. If you have ever used soy sauce 0:07:22.160 --> 0:07:25.880 of one of those Kickoman soy sauce bottles, beautifully kind 0:07:25.920 --> 0:07:29.040 of organically shaped one with the red lead, that's one 0:07:29.040 --> 0:07:30.560 of Ekuan's designs. 0:07:30.840 --> 0:07:35.400 One of the most ubiquitous and iconic consumers there. 0:07:35.240 --> 0:07:40.600 Is, absolutely and I found myself able to go and 0:07:40.640 --> 0:07:43.800 meet the great man. I went into his office. I 0:07:45.120 --> 0:07:48.360 was looking around me and it was this thing in 0:07:48.400 --> 0:07:52.080 the corner that I couldn't help noticing, which was a 0:07:52.120 --> 0:07:59.640 life size fiberglass human body kind of suspended in space 0:07:59.680 --> 0:08:03.080 horror onto. He was two axles going through the shoulders 0:08:03.720 --> 0:08:08.080 and the ankles, and big kind of chrome wagon wheels 0:08:08.080 --> 0:08:11.960 on the axles. After we'd done with this sort of 0:08:11.960 --> 0:08:13.640 introductory niceties. 0:08:13.200 --> 0:08:17.840 I said, I couldn't help noticing that thing. What is that? 0:08:18.680 --> 0:08:20.200 He just looked me in the eye and said, this 0:08:20.240 --> 0:08:23.680 is my soul chariot and pressed the switch and the 0:08:23.720 --> 0:08:27.600 acxules started rotating, and I was awestruck. I thought, this 0:08:27.640 --> 0:08:32.080 guy's amazing. This is not your typical Christy senior executive. 0:08:32.160 --> 0:08:35.160 He's talking about his soul chariot. And then we got 0:08:35.280 --> 0:08:40.080 talking about what my aspirations were, and he looked at 0:08:40.120 --> 0:08:43.880 some of my portfolio work from college and he said, 0:08:44.200 --> 0:08:48.720 I want you to design motorcycles. And through that experience, 0:08:48.760 --> 0:08:51.240 as I worked in his design firm design motorcycles, I 0:08:52.080 --> 0:08:54.760 read a lot about him and learned his philosophy and 0:08:54.800 --> 0:08:57.760 obviously had some exposure to him. And he has this 0:08:58.120 --> 0:09:00.679 way of talking about the central endeavor of design, and 0:09:00.800 --> 0:09:03.320 he calls it human machine soul energy. 0:09:03.600 --> 0:09:06.559 That is the game, Human machine soul energy. 0:09:06.640 --> 0:09:08.679 That is a game, and that has stuped with me 0:09:08.800 --> 0:09:12.520 ever since. So my career has been about the pursuit 0:09:12.559 --> 0:09:15.079 of human machine soul energy still is. 0:09:15.880 --> 0:09:19.120 That's a remarkable way of putting it, and especially at 0:09:19.160 --> 0:09:22.600 a time when I mean there's all these questions about 0:09:23.360 --> 0:09:26.800 Emerson consciousness. There's people who are falling in love with 0:09:26.880 --> 0:09:31.040 their ms, treating large language models as spiritual guides. I mean, 0:09:31.120 --> 0:09:35.239 this quality of the soul when it comes to technology 0:09:35.320 --> 0:09:38.360 is something something which for me is a kind of 0:09:38.440 --> 0:09:41.760 fascinating part at the reach of my understanding of what's 0:09:41.800 --> 0:09:44.360 going on silicon value right now. But I haven't had 0:09:44.400 --> 0:09:47.240 a chance to ask somebody really what they mean by 0:09:47.280 --> 0:09:48.559 it and what they think about it. So I have 0:09:48.640 --> 0:09:51.120 to ask you right now. I mean, how how does 0:09:51.200 --> 0:09:54.760 that phrase become newly relevant given everything we're living through 0:09:54.880 --> 0:09:55.160 right now? 0:09:55.600 --> 0:09:59.079 Well, it's newly relevant every time I visit, but it's 0:10:00.480 --> 0:10:05.559 particularly relevant right now because all of a sudden, really quickly, 0:10:06.720 --> 0:10:15.000 there's a bunch of experiences and capabilities that we're securely 0:10:15.080 --> 0:10:18.880 on the human side of the human machine equation, and 0:10:19.760 --> 0:10:24.160 now those are able to happen on the technology side. 0:10:24.760 --> 0:10:24.920 You know. 0:10:25.840 --> 0:10:32.560 Famously, there's a lot of anthrop of myomorphization of what. 0:10:32.679 --> 0:10:33.760 The lllms can do. 0:10:33.920 --> 0:10:37.360 People use all this language that was previously used for 0:10:37.480 --> 0:10:40.440 humans to describe how ILM's work. You know, they say 0:10:40.480 --> 0:10:45.000 it's thinking, and you know, the seashore between human and 0:10:45.200 --> 0:10:51.160 machine has changed rather radically, rather suddenly, and there's a 0:10:51.240 --> 0:10:55.080 lot of figuring out that has to happen to ensure 0:10:55.120 --> 0:10:59.040 that that change nets out positively for humans. 0:11:00.160 --> 0:11:01.920 Of what you're working on, and I want to ask 0:11:02.000 --> 0:11:04.760 you not just what the framing of the problem is, 0:11:04.840 --> 0:11:07.400 but what some of the initial answers are before we 0:11:07.520 --> 0:11:10.880 get there. Though, You finish up in Japan and in 0:11:11.000 --> 0:11:14.319 nineteen ninety seven you Silicon Valley, I believe, which is 0:11:14.400 --> 0:11:17.959 the same year that Steve Jobs returns to Apple, and 0:11:18.040 --> 0:11:20.920 you take a job at a firm called IDEO, which, 0:11:21.440 --> 0:11:23.439 for those who don't know, I think, is kind of 0:11:23.520 --> 0:11:28.240 a household name in this design thinking world. It was 0:11:28.280 --> 0:11:30.599 also the company that worked with Apple I think in 0:11:30.640 --> 0:11:35.160 the eighties to design Apple's first computer mouse. From nineteen 0:11:35.200 --> 0:11:39.080 ninety seven to today, working at IDEO until twenty twenty two, 0:11:40.000 --> 0:11:44.640 you witness first hand revolution after revolution in terms of 0:11:44.720 --> 0:11:48.680 the way humans interact with technology. Where do you place 0:11:49.000 --> 0:11:54.640 twenty twenty two and the consumerization of generative AI in 0:11:54.720 --> 0:11:55.240 that journey. 0:11:56.160 --> 0:11:59.400 It's interesting because on the one hand, there's one way 0:11:59.440 --> 0:12:05.520 of character, which is it's just another cycle, right, which is, Okay, 0:12:05.600 --> 0:12:11.080 there's a technology unlock. There's a bunch of investment that happens, 0:12:11.120 --> 0:12:14.319 and a bunch of hype that happens, and kind of 0:12:14.440 --> 0:12:20.920 Darwinistic explosion of invention of new value that happens. All 0:12:20.960 --> 0:12:24.800 of that is true, And so at some level it 0:12:24.920 --> 0:12:27.679 feels like Web two point or or feels like the 0:12:27.760 --> 0:12:32.640 dotcom moment. But to me it feels very different in 0:12:32.800 --> 0:12:38.679 terms of how pervasive this technology is already and definitely 0:12:38.800 --> 0:12:43.640 will be. But I think the fundamental character of the 0:12:44.240 --> 0:12:50.559 technology makes its potential even more significant. And then for me, 0:12:51.600 --> 0:12:54.360 the attraction to IDEO was a continuation of that through line. 0:12:54.360 --> 0:12:57.319 You know, engineer who went to art school went to 0:12:57.480 --> 0:13:01.120 Japan to be both technologists and create to and then 0:13:01.240 --> 0:13:09.000 Ideal famously resolved any dichotomy there. And so IDEO's cultures 0:13:09.640 --> 0:13:14.320 point of view philosophically is that it's an absolutely user 0:13:14.440 --> 0:13:20.120 centered and just some very motivated to bring that you know, 0:13:20.280 --> 0:13:24.720 solving for the users as a kind of fundamental approach 0:13:26.240 --> 0:13:29.520 to this extremely significant technology moment. 0:13:29.960 --> 0:13:32.000 Yeah, and going back to the mission of Google Labs 0:13:32.000 --> 0:13:34.959 that I read at the beginning, Google Labs is where you, 0:13:35.600 --> 0:13:38.240 I guess you means the user in this case can 0:13:38.320 --> 0:13:43.079 discover and try Google's latest AI experimental products and help 0:13:43.160 --> 0:13:46.160 shape the future of AI technology. A very interesting in 0:13:46.240 --> 0:13:49.160 second half of that sentence, especially because it's someone at 0:13:49.160 --> 0:13:52.000 odds with how people think about the development of AI. 0:13:52.200 --> 0:13:52.320 Right. 0:13:52.400 --> 0:13:55.880 The general model is, here's a bunch of computer scientists 0:13:55.960 --> 0:13:59.920 who are not particularly interested in humans or even humanity 0:14:00.200 --> 0:14:04.000 more broadly, and are willing to write it off in 0:14:04.120 --> 0:14:08.360 favor of a you know, silicon superintelligence. But here you 0:14:08.520 --> 0:14:11.880 are promising us as a role in the development of 0:14:11.880 --> 0:14:12.559 the technology. 0:14:13.480 --> 0:14:14.000 How do you do that? 0:14:14.760 --> 0:14:18.559 Yeah, that is absolutely the case, certainly something that I 0:14:18.640 --> 0:14:22.480 bring from Ideal in certainly in my experience of the 0:14:22.600 --> 0:14:26.000 engineering culture at Google, I don't think it is in 0:14:26.120 --> 0:14:30.000 denial about the user. But what you do is you 0:14:30.680 --> 0:14:34.080 you know, in the approach that Ideal famously used across 0:14:34.120 --> 0:14:40.080 all of these waves of technology, it's to characterize successful 0:14:40.120 --> 0:14:42.800 and innovation as being in the sort of sweet spot 0:14:42.920 --> 0:14:49.680 between feeability, desirability, and viability. Meaning there's a three circle 0:14:49.760 --> 0:14:55.800 diagram which is technology, human and business, and a successful 0:14:55.800 --> 0:14:58.000 innovation has to be like the middle of that three 0:14:58.160 --> 0:15:02.880 circle diagram. It has to matter to humans and be 0:15:03.000 --> 0:15:05.720 desirable to them for some reason, and it doesn't get 0:15:05.720 --> 0:15:07.880 to exist unless there's a business model that supports it, 0:15:08.760 --> 0:15:15.040 and often the conditions around a new technology moment, a 0:15:15.120 --> 0:15:20.120 new platform moment can lead to getting to that middle 0:15:20.120 --> 0:15:24.640 of the ven diagram territory from the technology circle. So 0:15:24.800 --> 0:15:29.440 a technology push that can often feel like an answer 0:15:29.560 --> 0:15:30.360 looking for a question. 0:15:30.680 --> 0:15:31.480 It's like, we've. 0:15:31.360 --> 0:15:36.160 Invented this new capability, let's figure out how to make 0:15:36.240 --> 0:15:36.960 people want it. 0:15:37.280 --> 0:15:39.760 With all due respect to Mark Zuckerberg, this is the 0:15:40.320 --> 0:15:42.400 if people are lonely, might not give them fifteen AI 0:15:42.600 --> 0:15:45.080 friends school of thought. I don't want reach out town 0:15:45.400 --> 0:15:46.480 so we can right. 0:15:48.560 --> 0:15:50.880 And it's often a blind spot for the technologists. I mean, 0:15:50.880 --> 0:15:56.800 the technologists are not necessarily cold or dispersonal or cynical, 0:15:56.840 --> 0:16:00.240 the proud parents of this new technology, but they might 0:16:00.280 --> 0:16:03.760 not realize that their babies, their babies ugly, you know. 0:16:04.160 --> 0:16:07.520 And what Ideal. 0:16:07.720 --> 0:16:13.520 Strongly believes, what I believe and what we practice in 0:16:13.680 --> 0:16:18.160 labs is that you have to, at the very least 0:16:19.280 --> 0:16:22.600 also fold in the human at the earliest stage. You 0:16:22.720 --> 0:16:27.480 have to and preferably drive it from from the user. 0:16:27.640 --> 0:16:29.600 You have to go out there and get curious about 0:16:29.600 --> 0:16:33.000 what matters to people on their terms and use that 0:16:33.320 --> 0:16:35.400 use what you learn there as the jumping off point. 0:16:35.600 --> 0:16:39.440 For generating new ideas about value that you can deliver 0:16:39.560 --> 0:16:43.360 within new technology. When you go about it that way, 0:16:44.160 --> 0:16:50.920 you're far more lightly to create experiences that users love. 0:16:51.560 --> 0:16:53.880 You're more likely to create the kind of value that 0:16:54.080 --> 0:16:58.160 you can defend over time as a company because you're 0:16:58.200 --> 0:17:02.360 in a you're in an emotional relationship with your users, 0:17:03.200 --> 0:17:06.480 and it generally works out a lot better. The thing 0:17:06.520 --> 0:17:11.120 that it's often in tension with is speed, because when 0:17:11.160 --> 0:17:14.399 a new technology appears, time is of the essence, and 0:17:14.480 --> 0:17:17.400 that's very real, and so you have to achieve both. 0:17:17.480 --> 0:17:25.080 Had when we come back the double edged sword of 0:17:25.160 --> 0:17:29.520 technology and what happens when you innovate too quickly stay 0:17:29.560 --> 0:17:44.840 with us. So I mean, let's anchor this in the 0:17:44.960 --> 0:17:49.880 tangible world of products that you have incubated and launched 0:17:50.240 --> 0:17:50.960 within labs. 0:17:51.040 --> 0:17:51.159 Right. 0:17:51.280 --> 0:17:53.960 So, Google's develop a conference was in May of this year. 0:17:54.040 --> 0:17:58.320 I think there were four labs launches that were announced, 0:17:59.160 --> 0:18:00.879 maybe a five. I'm not show but correctly if I'm wrong. 0:18:00.920 --> 0:18:04.760 But take one of them and explain how it reflects 0:18:04.840 --> 0:18:06.880 the process you've just described. 0:18:07.359 --> 0:18:10.359 Yeah, so let's take one from one of our recently 0:18:10.440 --> 0:18:15.639 shared ones. Flow so flows the surface that lets people 0:18:16.119 --> 0:18:20.400 interact with the capabilities of the generative video model VO three. 0:18:21.080 --> 0:18:25.080 VO is the name of the latest video generation model 0:18:25.960 --> 0:18:30.600 Google has, So you know, you can convey your intention 0:18:30.720 --> 0:18:35.840 to this model and it will generate incredible, undiscernible from 0:18:35.880 --> 0:18:39.359 reality videos. So VO is the model, Flow is the product. 0:18:39.440 --> 0:18:44.240 Flow is the means of conveying that intention to the model. 0:18:44.280 --> 0:18:48.600 As a filmmaker, so you communicate what you want the 0:18:48.680 --> 0:18:54.720 scene to be, you communicate camera instructions, you communicate edits 0:18:54.800 --> 0:18:58.159 to it, and Flow as a lab's product. Just as 0:18:58.160 --> 0:19:01.520 an example of how that comes about. You know, from 0:19:01.600 --> 0:19:05.639 the beginning of Flow, we worked very closely collaboratively with 0:19:06.240 --> 0:19:10.399 movie makers. One of our fundamental principles and labs is 0:19:10.440 --> 0:19:13.600 co create. Find out the people that you're interested in 0:19:14.520 --> 0:19:18.040 making something for and set things up so that you 0:19:18.119 --> 0:19:21.679 can actually work with them, and you're in a very 0:19:21.800 --> 0:19:25.040 very close feedback loop about how well you're doing in 0:19:25.160 --> 0:19:28.320 terms of delivering the kind of experience the matter to them. 0:19:28.359 --> 0:19:31.680 So we worked with a bunch of filmmakers and developing Flow, 0:19:32.400 --> 0:19:36.080 and it's really interesting you mentioned Notebook LM earlier, there's 0:19:36.080 --> 0:19:39.000 another kind of co creation there. We worked with Stephen Johnson, 0:19:39.560 --> 0:19:47.520 who's a best selling writer who had been obsessed for years, 0:19:47.680 --> 0:19:50.600 for like a decade with this idea of tools for 0:19:50.800 --> 0:19:57.159 thought and ways of collecting and formatting all of the 0:19:57.480 --> 0:20:01.520 stuff that he you he's when writing a novel, both 0:20:01.560 --> 0:20:06.159 his own previous writing has notes, his references, and figuring 0:20:06.160 --> 0:20:07.600 out how to make it most useful. 0:20:07.680 --> 0:20:09.560 So he had He was like the. 0:20:10.760 --> 0:20:16.760 Their original super user of notebook LM before notebook LM existed. 0:20:16.960 --> 0:20:19.800 So we hired him and brought them onto the team 0:20:20.080 --> 0:20:22.959 and co created the product with him and a lot 0:20:23.040 --> 0:20:26.920 of the core DNA of notebook LM is what it 0:20:27.119 --> 0:20:28.160 is because of Stephen. 0:20:29.040 --> 0:20:31.080 What's cool about notebook and LM is rather than just 0:20:31.160 --> 0:20:35.920 saying to it, you know Gemini, Let's say I'm interested 0:20:35.960 --> 0:20:40.720 in the history of the Nutcracker, and you know ballet, 0:20:41.480 --> 0:20:44.480 you can say notebook LM. Here's you know, a great 0:20:44.920 --> 0:20:48.600 New Yorker piece about the Nutcracker, a review of a 0:20:48.680 --> 0:20:53.000 contemporary production at the Brooklyn Clagory of Music. And here 0:20:53.520 --> 0:20:57.040 is you know, a long document about the history of dance. 0:20:57.680 --> 0:20:59.280 I don't want to look at anything else, just these 0:20:59.320 --> 0:21:01.720 three documents and then provide you with insights and so 0:21:02.160 --> 0:21:04.080 in a sense, it puts you in the driver's seat 0:21:04.200 --> 0:21:06.760 in terms of it's more like having a research assistant 0:21:06.880 --> 0:21:09.479 than just sending out a query and seeing what happens. 0:21:09.560 --> 0:21:09.720 Right. 0:21:10.280 --> 0:21:13.720 But then the kind of additional layer was and you 0:21:13.800 --> 0:21:15.800 can hear it as a two way podcast, and that 0:21:16.320 --> 0:21:18.680 final thing seemed to be what blue people's minds most 0:21:18.720 --> 0:21:18.959 of all. 0:21:19.359 --> 0:21:19.879 Absolutely. 0:21:19.960 --> 0:21:23.159 I mean, it wasn't entirely surprised that the blue people's 0:21:23.200 --> 0:21:26.000 minds because we had been having our minds going by 0:21:27.000 --> 0:21:31.359 internally before we launched it, and we knew that this 0:21:31.640 --> 0:21:34.720 is gold. You know, this is amazing as in terms 0:21:34.760 --> 0:21:39.679 of like delivering on the make the world's information accessible 0:21:39.800 --> 0:21:46.000 and useful. This directly served that in a super fresh way. 0:21:46.560 --> 0:21:50.119 That was super delightful. You know, a big part of 0:21:51.119 --> 0:21:54.880 user experience designs. Don't over index on yourself as the user, 0:21:55.200 --> 0:21:59.959 but as a user of notebook LM, the magic has 0:22:00.080 --> 0:22:03.520 and more and not. One of the early comments and 0:22:03.640 --> 0:22:08.160 our discard feed from actual users about it, well, somebody said, 0:22:08.400 --> 0:22:13.320 it makes boring stuff interesting and I use it all 0:22:13.400 --> 0:22:14.760 day every day to do that. 0:22:15.200 --> 0:22:18.119 I love it to play Devil's advocate for a moment. 0:22:18.800 --> 0:22:23.240 Some people say, well, you know, in the age of 0:22:23.400 --> 0:22:26.680 l l m's and jernity of AI. Anything that sits 0:22:26.720 --> 0:22:29.520 on top of a model is just a rapper. Uh, 0:22:30.240 --> 0:22:33.080 and so labs in the wrapper business. 0:22:34.280 --> 0:22:37.120 Yeah, I mean, like who isn't in the in the wrap? 0:22:39.119 --> 0:22:41.879 It's like, I guess I don't really agree with the frame. 0:22:42.480 --> 0:22:43.800 No, I'm not. 0:22:44.240 --> 0:22:44.480 I'm not. 0:22:46.160 --> 0:22:51.800 Is every car brand a rapper and internal combustion? Yeah? 0:22:51.920 --> 0:22:54.679 But to be a bit more substantive about this, right, 0:22:56.160 --> 0:22:58.920 why do I need flow? Why can't I just tell 0:22:59.480 --> 0:23:02.040 the three really in natural language what I want to achieve? 0:23:02.160 --> 0:23:04.159 Like why is a filmmaker? Do I need like a 0:23:04.200 --> 0:23:07.080 product layer beyond natural language processing between me and a 0:23:07.160 --> 0:23:08.200 video generation model? 0:23:09.280 --> 0:23:12.600 Yeah, I mean I think you're talking to UX person. 0:23:12.800 --> 0:23:15.200 I think it's all about that layer. It's all about 0:23:15.200 --> 0:23:18.720 the application layer. It's all about the qualities of the 0:23:18.880 --> 0:23:24.640 interaction experience with the underlying technology. I think we've been 0:23:24.760 --> 0:23:30.879 in the sort of a command line era for AI 0:23:31.119 --> 0:23:34.639 thus far, where it's all about kind of the user 0:23:35.760 --> 0:23:41.840 learning this arcane language of prompts and contorting what they 0:23:41.880 --> 0:23:44.920 want to make happen into that language to feed it 0:23:45.000 --> 0:23:47.760 to the model. That doesn't bring me joy as a user. 0:23:48.320 --> 0:23:50.760 I think it doesn't have to be that way. You know, 0:23:50.920 --> 0:23:56.000 we can figure out delightful interactions for the user to 0:23:56.240 --> 0:23:58.840 convery their intent to the model and for them to 0:23:58.920 --> 0:24:01.639 work with the model. And I think it's all about that. 0:24:01.760 --> 0:24:03.960 If you look at at how value is built and 0:24:04.040 --> 0:24:07.800 defended by the world's most successful companies, certainly in the 0:24:07.880 --> 0:24:13.439 consumer sector, it's about UX ya your UX. What are 0:24:13.480 --> 0:24:17.320 the qualities of that experience and that's what builds a 0:24:17.440 --> 0:24:19.919 character in the brand of a company. 0:24:20.240 --> 0:24:23.720 You mentioned business models yourself, right, Like where does a 0:24:23.960 --> 0:24:26.760 flow or notebook m LM? Like how much is it 0:24:26.880 --> 0:24:30.040 your job to monetize these new products that you're creating. 0:24:31.240 --> 0:24:35.840 Certainly not my job to monetize them. You know. 0:24:35.880 --> 0:24:40.240 The way tech companies in general are configured is you've 0:24:40.240 --> 0:24:43.040 got engineering as a discipline. You've got uxit's a discipline. 0:24:43.040 --> 0:24:45.000 You've got product management is a discipline, and the product 0:24:45.040 --> 0:24:49.480 managers or the or the CEOs essentially of the endeavor 0:24:50.119 --> 0:24:54.520 typically and it's their job to take an innovation on 0:24:54.600 --> 0:25:01.800 the trajectory from glimpse of value to users to successful product, 0:25:02.160 --> 0:25:10.680 including monetization strategy, usually through phases. And so when you're 0:25:10.720 --> 0:25:17.040 doing early stage, it's less about are you hitting a 0:25:17.320 --> 0:25:20.240 weekly active user target and it's more about are you 0:25:20.440 --> 0:25:24.240 seeing fire in the eyes of the users that you're 0:25:24.280 --> 0:25:28.200 putting in front of them, And certainly for me, my 0:25:28.400 --> 0:25:30.560