WEBVTT - They Founded a Company as Teens. Now They’re Predicting the Future - The Story 0:00:15.640 --> 0:00:18.599 Welcome to tech stuff. I'm os Vloschen. Every year we 0:00:18.640 --> 0:00:21.959 spend billions of dollars asking real humans for their opinions 0:00:21.960 --> 0:00:25.360 to predict how groups of people behave On a larger scale, 0:00:25.840 --> 0:00:30.120 surveys are the foundation of political strategy, marketing, and produge innovation. 0:00:31.000 --> 0:00:34.000 So what happens when you stop asking humans what they 0:00:34.040 --> 0:00:38.919 think and start simulating their responses? Today I'm joined by 0:00:38.960 --> 0:00:41.760 cam Think and Ned Coo, the founders of Aru. At 0:00:41.760 --> 0:00:44.479 the ages of twenty and twenty one, they've built a 0:00:44.520 --> 0:00:48.520 billion dollar simulation engine that uses thousands of AI powered 0:00:48.600 --> 0:00:52.520 digital twins to deliver behavioral predictions for everyone from Ernst 0:00:52.600 --> 0:00:56.200 and Young to Spindriff, Cam Ned, Welcome to text stuff. 0:00:56.360 --> 0:00:58.200 Thank you for having us having us. 0:00:58.200 --> 0:01:00.400 Thanks you very very excited. Tell me about the name. 0:01:00.440 --> 0:01:01.120 How did you choose it? 0:01:01.360 --> 0:01:04.720 Yeah? Well, Uru is the place in Egyptian mythology where 0:01:04.720 --> 0:01:07.520 all the worlds meet, and given we simulate human behavior, 0:01:07.520 --> 0:01:09.560 we simulate all these worlds. It makes a lot of sense. 0:01:10.040 --> 0:01:12.000 But it's also the oldest trick in the book. So 0:01:12.400 --> 0:01:14.280 you know, Auru starts at two a's at the top 0:01:15.240 --> 0:01:16.920 the top of every alphabetical list. 0:01:17.040 --> 0:01:19.200 The Yellow Pages trick you know, triple A plumbing, all 0:01:19.200 --> 0:01:21.600 those sorts of things. You look at any of our investors' 0:01:21.640 --> 0:01:24.000 websites and you see us and then Airbnb, which we 0:01:24.040 --> 0:01:25.720 think is the proper order. I'm just kidding. 0:01:27.200 --> 0:01:29.480 That's very very good, Am I right in thinking, or 0:01:29.480 --> 0:01:32.480 at least is Wikipedia right in thinking that to get 0:01:32.520 --> 0:01:35.600 into Uru, that your heart has to be weighed? That 0:01:35.800 --> 0:01:38.360 is true, That is true. Are you in the business 0:01:38.360 --> 0:01:39.160 of weighing hots? 0:01:39.360 --> 0:01:42.560 I think by so, I think by technicality, really is 0:01:42.880 --> 0:01:46.040 Uru is the place that describes the field of reads, right. 0:01:46.000 --> 0:01:48.520 So we are just kind of a heaven, shall we say, 0:01:48.560 --> 0:01:49.240 an ancient Egypt? 0:01:49.360 --> 0:01:50.919 Yeah, exactlytopia almost. 0:01:51.040 --> 0:01:52.760 This is, by the way, this is why r dot 0:01:52.760 --> 0:01:55.040 com our domain cost us so much, because I only 0:01:55.120 --> 0:01:58.600 learned afterwards that anything that has anything to do with heaven, 0:01:58.760 --> 0:02:01.920 like any domain, like our fortune, because it's really easy 0:02:01.920 --> 0:02:05.040 to bring something that's being positive. But we are not 0:02:05.120 --> 0:02:07.600 in the business of waite arts. I don't know if 0:02:07.640 --> 0:02:10.160 that would be as important for the global economy. 0:02:10.560 --> 0:02:12.600 So, because I was thinking, in a sense, you are 0:02:12.880 --> 0:02:15.280 definitely not judging people, but in a sense you are. 0:02:16.080 --> 0:02:19.639 You're looking into the digital arts, or at least trying 0:02:19.680 --> 0:02:24.440 to distill how real people might behave from people you've simulated. 0:02:25.000 --> 0:02:27.400 No, I think it's a it's interesting, and we like 0:02:27.440 --> 0:02:29.880 things that are thematic and actually mean things. I think 0:02:29.919 --> 0:02:32.360 there's a lot of tech companies nowadays that have names 0:02:32.360 --> 0:02:33.800 that don't really don't really mean much. 0:02:33.880 --> 0:02:37.000 So all the ones around Tolkien were taking. 0:02:36.120 --> 0:02:38.520 Exactly and the ones from Dune and the ones from 0:02:38.560 --> 0:02:40.280 Lord of the Rings, and it's, ah, well, there we 0:02:40.320 --> 0:02:42.079 go to mythology. You know, you gotta pull one. 0:02:43.200 --> 0:02:46.120 I want to sort of get to how aru actually works. 0:02:46.240 --> 0:02:47.880 But I saw the article about you two in the 0:02:47.880 --> 0:02:50.200 Wall Street Journal or you three, actually your third co 0:02:50.320 --> 0:02:53.600 founder is not here, and you know, basically the teenage 0:02:53.680 --> 0:02:55.919 teenage founders of a billion dollar company. And I thought 0:02:55.919 --> 0:02:59.200 it was a very fascinating and cool story. And I'm 0:02:59.200 --> 0:03:02.120 a huge tennis and it's been kind of interesting for 0:03:02.160 --> 0:03:07.919 me to observe how Djokovic, Nadal and Federer cast this long, 0:03:08.280 --> 0:03:10.799 fifteen year shadow over the game of tennis where the 0:03:10.880 --> 0:03:13.640 kind of next gen couldn't emerge. And it wasn't until 0:03:14.160 --> 0:03:16.680 Nadal and Federer left the scene and Djokovic started to 0:03:16.800 --> 0:03:18.480 age that Ciner and. 0:03:18.400 --> 0:03:21.320 Al Karaz emerged, you know, a lot of people say, 0:03:21.360 --> 0:03:26.160 Cameron looks like by the way, Yeah, I like that. 0:03:27.560 --> 0:03:30.480 So, I mean, you know, it's somewhat of a parallel 0:03:30.520 --> 0:03:33.680 with the tech industry. Right before the twenty twenty three 0:03:33.960 --> 0:03:38.640 chat GPT moment and the diffusion of llms, basically, most 0:03:38.680 --> 0:03:41.240 people with strong technical skills would go and work in 0:03:41.280 --> 0:03:44.680 corporate jobs at like Facebook and Google and Amazon. Yeah, 0:03:45.080 --> 0:03:47.920 and then something changed, and you and the censor at 0:03:47.920 --> 0:03:49.640 the vanguard of that something that changed. 0:03:50.280 --> 0:03:52.920 I think what happened is, for like a really long time, 0:03:53.000 --> 0:03:57.280 you had people who were very technical getting paid a 0:03:57.280 --> 0:03:59.880 lot of money to work, largely like kushy jobs. Right 0:04:00.120 --> 0:04:03.560 right now, we have this perception of Silicon Valley nine ninety. 0:04:03.400 --> 0:04:05.320 Six, like all this stuff, correct engineers. 0:04:05.520 --> 0:04:07.360 Yeah, it wasn't that way for the longest time. I mean, 0:04:07.400 --> 0:04:10.080 if you go back, like even five years in history 0:04:10.120 --> 0:04:13.080 to twenty twenty one, Silicon Valley was where people drank 0:04:13.120 --> 0:04:16.440 beer during the work day and played ping pong, you know, 0:04:16.480 --> 0:04:19.839 at four o'clock in the afternoon. It isn't like that anymore, 0:04:19.880 --> 0:04:22.279 because I think what happened is people got actually attracted 0:04:22.320 --> 0:04:23.800 to the idea of hard work. Right It's like this 0:04:24.240 --> 0:04:27.200 backlash against what was once a work culture of like 0:04:27.560 --> 0:04:29.880 go work ten hours twenty hours a week of productive 0:04:29.920 --> 0:04:31.719 time and then not spend the rest of your time 0:04:31.800 --> 0:04:35.080 doing as many productive things. People want to work hard, 0:04:35.200 --> 0:04:36.159 and I think that changed. 0:04:36.720 --> 0:04:39.400 I mean, with a barrier entry to building a business 0:04:39.440 --> 0:04:42.080 being so so much lower, so many things has changed, right. 0:04:42.120 --> 0:04:44.080 I don't think it's just young people being able to 0:04:44.080 --> 0:04:45.919 do it. I think it's also the people who are 0:04:45.960 --> 0:04:48.839 at the top of the fields also have changed a lot. 0:04:48.960 --> 0:04:52.240 Like something Cameron and I discuss a substantial amount is 0:04:52.560 --> 0:04:55.560 you know, where is this generation's you know, Steve jobs 0:04:55.640 --> 0:04:57.720 or those sorts of areas. And it's actually part of 0:04:57.720 --> 0:05:00.160 our thesis around this space is that people have and 0:05:00.200 --> 0:05:02.480 had to rally around a singular lumin area as much 0:05:02.520 --> 0:05:05.039 because capital has been so much more accessible. Right, So 0:05:05.040 --> 0:05:07.120 it's not just as part of language models. I think 0:05:07.120 --> 0:05:09.080 it's a state of the you know, overall economy and 0:05:09.120 --> 0:05:10.760 things that have changed that have allowed it to happen. 0:05:10.839 --> 0:05:14.239 But it's very interesting. Nonetheless, Yeah, it is interesting. 0:05:14.440 --> 0:05:16.520 Do you think it's like a stylistic thing that your 0:05:16.560 --> 0:05:18.400 generation want to work hard or do you think it's 0:05:18.400 --> 0:05:21.120 an opportunity thing that like if you work hard, you 0:05:21.200 --> 0:05:23.679 can be net and Cam and get a big Unicorn 0:05:23.680 --> 0:05:24.920 founders before you're twenty one. 0:05:25.400 --> 0:05:28.160 I actually think it's I think it's a bifurcation, right, 0:05:28.200 --> 0:05:30.320 Like I think you have a population of people who 0:05:30.400 --> 0:05:32.159 work really hard, and that might be like the top 0:05:32.200 --> 0:05:35.040 death style. But what just happened is like the distribution 0:05:35.160 --> 0:05:37.920 of how hard people work has widened so much, so 0:05:37.960 --> 0:05:40.760 like just as much as in my high school class, 0:05:40.760 --> 0:05:42.560 there are a bunch of kids who, like maybe ten 0:05:42.800 --> 0:05:47.320 fifteen kids who worked extremely hard orderly. You know, I 0:05:47.400 --> 0:05:49.360 was not in the top performing in my high school class. 0:05:49.520 --> 0:05:51.920 Actually proud to announce that I was thirty third rank 0:05:51.960 --> 0:05:53.080 out of one hundred and eight. 0:05:53.560 --> 0:05:54.960 You were high school together, what were you? 0:05:55.600 --> 0:05:58.160 Probably much lower. It's very difficult for me to focus 0:05:58.200 --> 0:06:00.360 on things I don't care about, so you know, it's. 0:06:00.160 --> 0:06:02.920 But you had like ten to fifteen, maybe twenty kids 0:06:02.960 --> 0:06:05.440 who would like go study six hours and they would 0:06:05.440 --> 0:06:07.719 be totally happy to do that, and on the contra, 0:06:07.920 --> 0:06:11.359 like the next eighty kids just did not want to 0:06:11.400 --> 0:06:13.480 put all the work in. And I think you see 0:06:13.480 --> 0:06:16.160 that consistently where it's like there's this really high visibility 0:06:16.200 --> 0:06:18.240 group of people who work a lot, but remember, there 0:06:18.240 --> 0:06:21.159 are tons of software engineers at legacy businesses who don't 0:06:21.200 --> 0:06:24.080 work as much as like the high visibility group of people. 0:06:24.400 --> 0:06:27.360 That is like a very small small group of people 0:06:27.400 --> 0:06:30.880 in comparison to like broader software engineering is a category. 0:06:31.200 --> 0:06:32.960 How do you think about your paid group? I mean, 0:06:33.000 --> 0:06:36.000 do you know the other twenty eighth year old Unicorn founders? 0:06:36.040 --> 0:06:37.440 Do you associate with them? 0:06:37.560 --> 0:06:37.680 Or? 0:06:37.920 --> 0:06:39.440 I mean you're in New York and most are in 0:06:39.440 --> 0:06:41.520 San Francisco, which is a difference, right, Like, what's your 0:06:41.520 --> 0:06:44.360 reflection on the culture that's emerging around around that? I 0:06:44.400 --> 0:06:47.080 love the idea of Steve Jobs no longer being a 0:06:47.120 --> 0:06:50.320 cultural reference point because the structure of how capitalist deploy 0:06:50.400 --> 0:06:52.440 is different. That's really interesting. But how do you think 0:06:52.480 --> 0:06:53.479 about the group you know? 0:06:53.560 --> 0:06:57.000 I'm curious for Ninn's thoughts as well. I really really 0:06:57.040 --> 0:07:00.320 think the fact that more people know that is a 0:07:00.320 --> 0:07:02.720 pathway that is open to them is brilliant. I think 0:07:02.720 --> 0:07:05.159 that's incredible, right, like one hundred percent people should go 0:07:05.200 --> 0:07:08.039 do what they want. I do think some of the 0:07:08.160 --> 0:07:12.800 current environment has created people who are not in this 0:07:12.960 --> 0:07:16.560 because they're missionaries, but are instead in this industry because 0:07:16.560 --> 0:07:18.440 they're mercenaries and not to say that that is like 0:07:18.520 --> 0:07:20.920 the founders of other unicorns or other young people. But 0:07:21.400 --> 0:07:22.960 I just see a lot of people now who I 0:07:23.000 --> 0:07:26.880 think go to tech because not of a mission, not 0:07:27.080 --> 0:07:30.640 because they think it's like important philosophically important for the world, 0:07:30.960 --> 0:07:33.840 but instead because tech is like the next highest opportunity 0:07:33.840 --> 0:07:36.960 earning category, like finance was in the nineteen eighties. And 0:07:37.400 --> 0:07:40.240 I think shifting away from that is going to be 0:07:41.080 --> 0:07:44.080 interesting because the pendulum will always come back around. 0:07:44.240 --> 0:07:47.320 No, I agree entirely with that. I also think, to 0:07:47.360 --> 0:07:49.960 the same extent Cameron mentioned, you've also seen like the 0:07:49.960 --> 0:07:53.040 pliferation of so many different industries, not just tech, and 0:07:53.080 --> 0:07:54.960 people saying, oh, I can actually make a business. Ow's 0:07:55.000 --> 0:07:57.720 something I like, And it's become infinitely easier to monetize 0:07:57.800 --> 0:07:59.680 right as like a very small business across so many 0:07:59.680 --> 0:08:01.920 different channels, whether it's e commerce, whether that's you know, 0:08:02.240 --> 0:08:04.600 starting a tiny little brand and CpG or whatever may be. 0:08:05.000 --> 0:08:06.800 So I don't think it's a trend that's just in tech. 0:08:06.840 --> 0:08:09.080 I do think people are a lot more public about it, 0:08:09.240 --> 0:08:11.400 but it's been interesting to see. Nonetheless, I mean, we 0:08:11.480 --> 0:08:14.160 have friends that also dropped out of university to start 0:08:14.200 --> 0:08:16.680 schools and have had other companies that have been starting, 0:08:16.840 --> 0:08:19.720 you know, starting very very young, and I think there 0:08:19.800 --> 0:08:23.320 was this large sort of cultural shift. Cameron noted about, 0:08:23.320 --> 0:08:26.240 you know, silkon Valian, it's perception prior, it's now in 0:08:26.280 --> 0:08:28.320 some ways like sexy to be a tech founder. So 0:08:28.320 --> 0:08:30.600 I think you've a lot of people who whose hearts 0:08:30.600 --> 0:08:32.280 really aren't in it and these sorts of elements who 0:08:32.280 --> 0:08:34.400 are going after it. And you know, like anything in 0:08:34.400 --> 0:08:36.920 the world that you commit to not fully is things 0:08:36.920 --> 0:08:37.880 have repercussions. 0:08:38.080 --> 0:08:40.520 There's also an accessibility thing, right, Like your third co 0:08:40.600 --> 0:08:43.360 founder is not here. I think when you Will started 0:08:43.480 --> 0:08:46.480 you were nineteen, you were eighteen, and he was fifteen yep, 0:08:46.640 --> 0:08:49.040 which is kind of remarkable. But he reached out and 0:08:49.040 --> 0:08:51.200 basically said you were an accelerator program, can I get 0:08:51.200 --> 0:08:54.000 your advice? And then you left the call basically with 0:08:54.080 --> 0:08:55.719 him as a third co founders That might probably have 0:08:55.760 --> 0:08:57.160 a simplified but it. 0:08:57.480 --> 0:08:59.480 Was pretty close to that. He called them me on 0:08:59.559 --> 0:09:02.960 LinkedIn and he just said, I'm applying to this accelerator, 0:09:03.000 --> 0:09:06.199 can I have your advice on my application? I don't 0:09:06.240 --> 0:09:08.960 even check my LinkedIn dms that often. I certainly didn't 0:09:08.960 --> 0:09:11.199 when I was eighteen, I mean, you know what I've 0:09:11.200 --> 0:09:14.400 been doing on LinkedIn, And so we ended up scheduling 0:09:14.640 --> 0:09:17.080 a thirty minute zoom call. We spoke for two and 0:09:17.080 --> 0:09:19.559 a half hours, and I called Ned immediately afterwards and 0:09:19.600 --> 0:09:22.280 I said to Ned, I've just met the smartest person 0:09:22.320 --> 0:09:24.400 I've ever met in my life. And that was the moment. 0:09:24.520 --> 0:09:26.400 More or less immediately, I think the three of us 0:09:26.400 --> 0:09:28.520 hopped on a call later that night and we were like, 0:09:28.559 --> 0:09:30.559 screw it, We're starting a company together. I mean, Nut 0:09:30.559 --> 0:09:33.160 and I were about to sell our previous business. John 0:09:33.440 --> 0:09:36.200 literally dropped out of high school and we set everything 0:09:36.200 --> 0:09:38.600 down and said, is, while we're passionate about. 0:09:39.080 --> 0:09:42.080 What was the vision you laid out to him? And 0:09:42.120 --> 0:09:43.920 what was it that he said in response that made 0:09:44.000 --> 0:09:46.320 you think he was the most intelligent person you ever met. 0:09:46.679 --> 0:09:48.920 Well, at the time, he actually was writing a paper 0:09:49.000 --> 0:09:52.760 focused on RAG for healthcare agents, so this space I was, 0:09:52.880 --> 0:09:55.080 Nett and I were in digital health as well, and 0:09:55.160 --> 0:09:57.200 so I remember reading it and thinking to myself, Wow, 0:09:57.280 --> 0:09:59.720 like just reading these benchmarks, If these benchmarks are real, 0:10:00.200 --> 0:10:02.960 it was a material improvement in the quality of models, 0:10:03.200 --> 0:10:07.559 specifically for retrieving like medical evidence and clinical studies. At 0:10:07.559 --> 0:10:12.600 that time it was called heel actually hal and reading that, 0:10:12.679 --> 0:10:14.480 I was just like, wow, this is insane. I spoke 0:10:14.520 --> 0:10:16.640 to him about it, and I was kind of digging 0:10:16.640 --> 0:10:18.240 in and I was like, well, it seems really real. 0:10:18.480 --> 0:10:22.199 So like a fifteen year old living in a dorm 0:10:22.360 --> 0:10:25.800 at a boarding school in like rural Massachusetts who's from 0:10:25.880 --> 0:10:29.480 New Hampshire is able to best the labs and that 0:10:29.640 --> 0:10:32.360 was incredible, And so all of us randomly having a 0:10:32.440 --> 0:10:35.160 digital health background, but then we also all shared this 0:10:35.240 --> 0:10:38.840 vision of population research. John was like an original population 0:10:38.920 --> 0:10:42.360 researcher when he probably also called dm'd its way into 0:10:42.440 --> 0:10:45.079 a lab at Tufts and a lab at MIT doing 0:10:45.160 --> 0:10:47.320 city science and population demographic research. 0:10:47.679 --> 0:10:49.199 But did you already have the idea for. 0:10:51.040 --> 0:10:54.920 Something that was developed in its current manifestation together, right, 0:10:54.960 --> 0:10:57.600 But in the idea of understanding and simulating populations. It's 0:10:57.640 --> 0:10:59.760 something we thought about for a while, but our own 0:10:59.800 --> 0:11:03.000 it's manifestation was certainly something that we developed together, which 0:11:03.040 --> 0:11:03.640 I think is cool. 0:11:03.760 --> 0:11:06.320 I think it was October twenty twenty three we started 0:11:06.320 --> 0:11:09.120 talking about predicting behavior for the first time, and I 0:11:09.120 --> 0:11:10.800 think from there on at it was something that just 0:11:10.840 --> 0:11:12.079 completely consumed us. 0:11:12.200 --> 0:11:15.680 See you mentioned missionaries versus mercenaries. What is the mission 0:11:16.080 --> 0:11:16.400 for us? 0:11:16.400 --> 0:11:18.679 It's predict the future. I think it's really simple. If 0:11:18.720 --> 0:11:22.040 you think that humans are the most impactful species on 0:11:22.080 --> 0:11:25.440 the club, and we're able to predict human behavior, then 0:11:25.800 --> 0:11:29.200 predicting human behaviors tend amount to predicting the future. And 0:11:29.240 --> 0:11:32.000 if you can predict the future, you can shape it. 0:11:32.000 --> 0:11:35.160 It's not just the power to say this is how 0:11:35.360 --> 0:11:37.760 your product concept will perform, or this is who we 0:11:37.800 --> 0:11:40.240 expect to win in an election, but instead it's actually 0:11:40.240 --> 0:11:42.800 the power to find the product that wins and to 0:11:43.160 --> 0:11:48.280 help the people ideas that matter win elections more consistently. 0:11:48.679 --> 0:11:50.600 So predict littles of influence. 0:11:50.400 --> 0:11:53.920 Predict and shape and help people that need to shape 0:11:53.920 --> 0:11:56.439 behavior shape that right for the broader good. And that's 0:11:56.440 --> 0:11:58.480 something we care a lot about. There's a lot of 0:11:58.760 --> 0:12:02.840 predicted technology that exis for the aspect of simply measuring 0:12:02.960 --> 0:12:05.400 behavior right. And at the end of the day, what 0:12:05.559 --> 0:12:07.360 has to be done after that? When you get a 0:12:07.360 --> 0:12:09.480 spreadsheet that says this is what this person said, this 0:12:09.559 --> 0:12:11.719 is what this person said, et cetera, it's not very 0:12:11.720 --> 0:12:14.160 actionable to anybody, and so it ends up being this 0:12:14.280 --> 0:12:16.920 humongous waste of an exercise, because as much as might 0:12:16.960 --> 0:12:18.280 have taken you months to get to that point, no 0:12:18.320 --> 0:12:20.559 matter what measurement it gets to, it's very difficult for 0:12:20.640 --> 0:12:22.120 someone to have to go from there and make their 0:12:22.120 --> 0:12:24.280 own assumptions upon that data to then make a conclusion. 0:12:24.760 --> 0:12:27.200 And then frequently that data ends up just supporting what 0:12:27.240 --> 0:12:29.720 that initial person's decisioning has been or gets sworn that 0:12:29.720 --> 0:12:33.360 way or another. We have no like manual waiting on 0:12:33.440 --> 0:12:36.360 the output of our system. We run simulations of behavior 0:12:36.400 --> 0:12:38.600 our system as analysis upon it, and people take action 0:12:38.640 --> 0:12:40.599 as a result. And that's part of what makes it 0:12:40.679 --> 0:12:41.199 so powerful. 0:12:41.360 --> 0:12:44.040 When I think about prediction and influence, I guess there's 0:12:44.120 --> 0:12:49.840 like black hat predictors and influences like you know, Goebbels 0:12:50.800 --> 0:12:54.320 or Cambridge Analytica, and then there's like would be white 0:12:54.320 --> 0:12:57.200 hat predictors and influences like cast Sunstein and his like 0:12:57.280 --> 0:13:01.200 famous nudge theory. Right, do you think about this spectrum 0:13:01.360 --> 0:13:01.880 all the time? 0:13:01.920 --> 0:13:03.640 And if we didn't, I don't think we would believe 0:13:03.640 --> 0:13:05.440 in our technology. But we do, and we see how 0:13:05.480 --> 0:13:08.360 strong it is every day, and we do as much 0:13:08.360 --> 0:13:10.040 as we can to be on the white out side 0:13:10.040 --> 0:13:12.000 of that. Right, So, like I think about things and 0:13:12.200 --> 0:13:14.200 you know, maybe not to announce it here, you know, 0:13:14.320 --> 0:13:16.800 formulae leans or not. We're not without finance and all 0:13:16.800 --> 0:13:19.520 that yet. But later this year we'll be launching a 0:13:20.000 --> 0:13:23.839 research institute within our where we're dedicating a significant amount 0:13:23.960 --> 0:13:27.440 of compute and infratructure from our side who works exclusively 0:13:27.480 --> 0:13:30.520 on pro bono aspects. So certain these areas are you know, 0:13:30.640 --> 0:13:33.000 women's self for example, getting people to go to preventive 0:13:33.040 --> 0:13:36.560 screenings for you know, preventive diseases and other elements like that. 0:13:36.559 --> 0:13:39.040 That's a behavioral problem, right. Things like getting people to 0:13:39.120 --> 0:13:42.560 understand that nuclear energy isn't dangerous, that's a behavioral problem, right, 0:13:42.600 --> 0:13:44.679 All those sorts of elements. There's so much good you 0:13:44.720 --> 0:13:47.319 can do in the world if you understand what people's 0:13:47.360 --> 0:13:50.120 awareness of certain issues are, how to make them aware 0:13:50.120 --> 0:13:52.200 of certain elements and how to change their behavior on 0:13:52.240 --> 0:13:54.959 things that are genuinely better for society. And that's something 0:13:54.960 --> 0:13:56.360 we spend a lot of time thinking about. 0:13:56.640 --> 0:13:59.200 Do you have philosophical ethical debates the two of you 0:13:59.240 --> 0:14:01.440 with your co found it do board? What is the 0:14:01.880 --> 0:14:04.680 count sort of I was going to account some elders, 0:14:04.720 --> 0:14:06.800 but it's not appropriate in this conversation. But how do 0:14:06.840 --> 0:14:08.560 you debate these questions together? 0:14:08.720 --> 0:14:11.480 I mean, we were a company founded in like really 0:14:11.480 --> 0:14:16.120 deep philosophical principles. We do have a lot of thoughts, 0:14:16.559 --> 0:14:19.560 you know, about the moral and ethical use of our technology. Right, Like, 0:14:20.000 --> 0:14:24.040 to be clear, predicting behavior is really powerful. Predicting behavior 0:14:24.200 --> 0:14:29.400 incredibly accurately, with the capability to shape it is like 0:14:29.480 --> 0:14:31.720 even more powerful. Right, this is something that not only 0:14:31.760 --> 0:14:33.760 economically is important, but for the good of the world 0:14:33.800 --> 0:14:36.640 it's important. And so I think part of the reason 0:14:36.720 --> 0:14:38.800 we built these company is because we wanted to use 0:14:38.840 --> 0:14:40.360 this technology as a force for good. 0:14:40.560 --> 0:14:43.400 So how does it work well within what we can share? 0:14:43.440 --> 0:14:47.040 Of course? No, it's really interesting because you think about 0:14:47.080 --> 0:14:49.320 the architecture or system, and you think about the proliferation 0:14:49.400 --> 0:14:52.480 of language models and how much more accessible starting certain 0:14:52.520 --> 0:14:54.960 elements has been and all these things. And certainly these 0:14:54.960 --> 0:14:57.880 systems would not be capable without language models. But they're 0:14:57.920 --> 0:15:00.560 not the only thing that makes them what they are. Right, 0:15:00.760 --> 0:15:02.840 you think about what a language model is built for 0:15:02.880 --> 0:15:07.240 and its most basic principles. It is not to capture 0:15:07.240 --> 0:15:10.360 the irrationality of human behavior, right, It's not. It's in 0:15:10.360 --> 0:15:13.440 fact quite the opposite. To average results exactly. It is 0:15:13.480 --> 0:15:16.360 a beta generating machine in many ways, right, And so 0:15:16.480 --> 0:15:19.000 you think about the things that make humans who we are, 0:15:19.080 --> 0:15:21.440 that makes us act or buy a product from a 0:15:21.440 --> 0:15:24.080 brand and a new category that they've never made before. 0:15:24.560 --> 0:15:27.120 Of course, a language model doesn't work for that use case, right, 0:15:27.160 --> 0:15:29.640 And so there have been many evolutions of the architecture 0:15:29.680 --> 0:15:32.960 that we've come to today. At its core, what it 0:15:33.000 --> 0:15:36.320 is is what we call multi agent simulation. So effectively, 0:15:36.360 --> 0:15:40.600 we organize these large populations of AI agents, assign them 0:15:40.680 --> 0:15:44.440 different distributions of income, age, race, gender, sex, you know, 0:15:44.640 --> 0:15:46.720 credit card purchase history, all these sorts of elements that 0:15:46.760 --> 0:15:49.520 makes them who they are. The entire principle of how 0:15:49.520 --> 0:15:52.840 we generate those distributions of what we call audiences in 0:15:52.880 --> 0:15:55.240 the first place, or the demographics that make that simulate 0:15:55.280 --> 0:15:58.320 and population up or that we never actually survey people 0:15:58.400 --> 0:16:01.240 at all, and so we're just generate these populations off 0:16:01.280 --> 0:16:03.560 of the basic building blocks and what makes them who 0:16:03.560 --> 0:16:06.200 they are. So what we call ground truth data, right, 0:16:06.240 --> 0:16:08.400 we only look at data from areas where we don't 0:16:08.440 --> 0:16:11.440 actually have to ask people because people lie, they have biases. 0:16:11.480 --> 0:16:13.240 And you go to trybeca near where our office is 0:16:13.280 --> 0:16:14.960 and you ask the average man how much money makes 0:16:14.960 --> 0:16:17.160 He's gonna say four and a half million bucks, you know, 0:16:17.280 --> 0:16:19.680 like whatever it may be, right, So we generate these 0:16:19.680 --> 0:16:23.040 populations based off these ground truth distributions of traits and 0:16:23.040 --> 0:16:24.920 then be able to sign those distribution of traits to 0:16:25.000 --> 0:16:28.000 these thousands of models to become simulated populations, or will 0:16:28.040 --> 0:16:30.560 then ask them questions to measure their behavior. Right, so 0:16:30.600 --> 0:16:32.440 you might ask someone you know, who you vote for, 0:16:32.480 --> 0:16:34.000 what will you buy? How much will you buy? This 0:16:34.080 --> 0:16:35.960 for all those sorts of elements to be a measure 0:16:36.000 --> 0:16:38.880 everything from proprency to purchase to who's going to win 0:16:38.920 --> 0:16:41.920 an election to you know, churn risk and everything in between. 0:16:42.120 --> 0:16:44.720 And is it falsifiable like by a validation? 0:16:45.600 --> 0:16:48.239 How do you like? How do you be tested exclusively 0:16:48.280 --> 0:16:48.840 to outcomes? 0:16:49.120 --> 0:16:52.720 Right, So we never try to benchmarker models on surveys 0:16:52.960 --> 0:16:56.080 or any sort of traditional primary research, so surveys, polls, 0:16:56.160 --> 0:16:58.600 focus groups, et cetera, because again we believe this to 0:16:58.680 --> 0:17:01.960 be biased. So exclusive look at the actual actions people 0:17:02.000 --> 0:17:04.359 take and not what they say they do. And there's 0:17:04.480 --> 0:17:06.160 a lot of different examples of where we've been able 0:17:06.200 --> 0:17:08.439 to do this publicly that have been really cool. You know, 0:17:08.480 --> 0:17:10.520 this is a fantastic example with one of our partners, 0:17:10.640 --> 0:17:13.040 Ernst and Young, who you mentioned. They took a survey 0:17:13.080 --> 0:17:16.240 that traditionally took the months to run. Thirty six hundred 0:17:16.280 --> 0:17:20.840 people responded, thirty countries, fifty three questions. We said, don't 0:17:20.840 --> 0:17:22.240 trust their models. Right, This is what we say to 0:17:22.240 --> 0:17:24.679 all our customers. They gave us the questions they asked. 0:17:24.960 --> 0:17:26.600 They gave us the people they spoke to, or the 0:17:26.680 --> 0:17:29.600 quotas for the assets under management. It was an investor simulation. 0:17:30.440 --> 0:17:32.760 We ran the entire questions that within our model in 0:17:32.800 --> 0:17:34.760 a matter of minutes returned it to them. They did 0:17:34.760 --> 0:17:37.639 a correlation, had a ninety percent correlation, but in the 0:17:37.680 --> 0:17:40.960 ten percent where we were not accurate, we were significantly 0:17:41.080 --> 0:17:44.640 closer to the real human action. So you think about 0:17:44.640 --> 0:17:46.480 a question in that example, right. One of those in 0:17:46.480 --> 0:17:49.600 that study was what's your likelihood to retain a wealth 0:17:49.680 --> 0:17:53.240 manager when your parents pass away? What are the three 0:17:53.280 --> 0:17:54.600 things that go through your head when you get asked 0:17:54.640 --> 0:17:56.000 that question. The first one is I hope my parents 0:17:56.040 --> 0:17:57.919 don't pass away. Second one is that might have been 0:17:57.920 --> 0:17:59.560 in five years, ten years, thirty years, I don't know. 0:17:59.600 --> 0:18:01.400 I'm going to do my money then, and the third 0:18:01.400 --> 0:18:03.920 one is my wealth managers asked me this question. I'm 0:18:03.920 --> 0:18:05.600 not going to tell them a lot, right, So you 0:18:05.640 --> 0:18:07.520 have this whole scenario where at the end of the day, 0:18:07.920 --> 0:18:10.240 eighty two percent of the people in their traditional measurement 0:18:10.280 --> 0:18:13.439 gold standard survey said they were going to retain wealth manager. 0:18:13.680 --> 0:18:16.760 Real market data from actual wealth management firms said this 0:18:16.840 --> 0:18:19.280 between the tune of like twenty to forty percent, depending 0:18:19.320 --> 0:18:22.280