The Information's TITV
00:19:07 · Stephanie Palazzolo, Anne Wojcicki
[00:19:07] S0: A lot of people, think, are also just taking their health data into their own hands and feeding it into chatbots and seeing what comes out. Yeah. I'm curious what you think of that because I think, yeah. I think there are strong arguments to both, you know, the pro and the con side. [00:19:20] S5: And I see... I think I think there's a couple of things. Like, one, it's always interesting when people experiment. But it's not... I think that's not mainstream. You know? Like, the reality is, like, most people want to have it curated for them. Secondly, I think one of the issues is about data quality. And I think, again, this administration, others also all recognize this. There's a lot of publications that come out that are in very small populations. They're small numbers of people. Like, doesn't... It's not really a gold star kind of study. So understanding what is real is important. And so part of what we will do is we'll put a curation, and we... This is what we do do now is, like, we have the curation layer on top of it. So we're curating the information so that when I give it back to you, that I know it's actually all things that are well validated, that are meaningful for you. Mhmm. And I think that the, you know, the chattypety, like all of the LM, like interpretations are remarkably good. But there's aspects where I'd say, okay, I am getting results that I would not say are necessarily scientifically valid that I would want to follow. Mhmm. [00:20:29] S0: So kind of... Yeah. Kind of being able to filter out maybe what is accurate
00:20:29 · Stephanie Palazzolo, Anne Wojcicki, Nicole Su
[00:20:29] S0: So kind of... Yeah. Kind of being able to filter out maybe what is accurate [00:20:33] S5: about layers to like what what scientific papers... What... What's actually really well valid in the literature. And then I think the other thing is like, okay, looking at your ancestry and your population, there's all kinds of research around, you know, some things have population specifics. So something might just be impacting an African American population more than a Latino population than a European population. Mhmm. So understanding that background as well. [00:20:56] S0: Totally. There's broader context as you're saying that, you know, might not always think to tell a chatbot about your parents' cancer history and Right. Your racial background, all these sorts of things. [00:21:08] S5: But it is good. I mean, [00:21:10] S3: I think it's... Like, I [00:21:12] S5: think there's a lot of interesting opportunities there. Like, I do believe there's a world where you're gonna pull in your medical records, your blood, your genetic data. And I do think what's mainstream more... Like, people are going to want to have an app and experience. And part of what 23andMe has spent a lot of time doing is, like, how do you return this information in a way that makes people understand? How do you make it so that you can compare everything together? How do you also, like, tie in your ancestry? So that's a lot of the work that we are doing. But I do appreciate, like, I think it's going to be the ability for you to have interpretation very broadly is exciting for people.
00:21:12 · Anne Wojcicki, Stephanie Palazzolo
[00:21:12] S5: think there's a lot of interesting opportunities there. Like, I do believe there's a world where you're gonna pull in your medical records, your blood, your genetic data. And I do think what's mainstream more... Like, people are going to want to have an app and experience. And part of what 23andMe has spent a lot of time doing is, like, how do you return this information in a way that makes people understand? How do you make it so that you can compare everything together? How do you also, like, tie in your ancestry? So that's a lot of the work that we are doing. But I do appreciate, like, I think it's going to be the ability for you to have interpretation very broadly is exciting for people. [00:21:49] S0: And when you think about, you know, who's going to be first to some of these major medical discoveries, do you think obviously, you know, maybe a bit biased, but do you think it will be a company like Anthropic, for instance, that's coming from it at this very this broad approach, you know, obviously building these very generalized language models and then applying them to Bio, for instance? Or do you think it'll be companies that have a much more, you know, specialized approach like 23andMe where, you know, you have obviously spent years and years working with genetic data? Do you have a sense of like kind of which approach will win out or be first I think...
00:22:24 · Anne Wojcicki, Brittany Greenfield
[00:22:24] S5: Look, I think I, I love that everyone is getting into biology. Like I, I joked in some ways, I was like, it's like the new vineyard, know, like, people make a lot of money, they come, they're like, I'm starting a biotech. Like, this is great. And I really, really appreciate, like, all these people are coming in, and they're taking different approaches, and they are making a difference. The challenge with biology is that it's spectacularly complicated. And clinical trials clinical trials take time. And ultimately, it's very different than, like, me testing a marketing campaign. Like, I can run 100 tests, you know, a second, like, in in... Like, get marketing campaigns. In biology, you have to wait to see, like, okay, I developed this molecule. I can develop a target. I can develop a molecule. I put in humans, and I have to wait to see. Do they have a heart attack? Do they not have a heart attack? So the recursive learning is just much slower. [00:23:22] S4: Yeah.
00:23:22 · Brittany Greenfield, Anne Wojcicki
[00:23:22] S4: Yeah. [00:23:22] S5: So I'm I'm really excited about the opportunities, and I think that AI is made for biology. Like, it's absolutely made for it because you have so much complexity and so many things happening. But the challenge we have today is we just don't have the data sets. And I don't necessarily have the answers, like, data sets are going to be the right ones to transform? You know, in some ways, like, of the problem that I took on was genetics. Like, how do I understand your genome? Like, I like it because it's like structured data. It's ACG and T. I can collect phenotype data from each person. I can see outcomes. There's in some ways like an ability to actually train models and to understand that really well. Biology, really, really hard. And developing drugs is a humbling experience. So I like... I love people coming in. I love the fact that there is capital that's out there now to put effort into this. But I think, like, I'm I'm a realist in terms of, like,
00:24:29 · Stephanie Palazzolo
[00:24:29] S0: it's hard. Mhmm. Definitely. I mean, I think building AI models is already difficult. And then I think a lot of these labs are gonna realize once they start to run these trials, like what you're saying, that actually clinical trials do take a lot of time. But, well, great. Well, it was so great to have you on, Anne. And again, that was Anne Wojcicki, founder and CEO of 23andMe Research Institute here on TITV. Lori Goler is president of the Chan Zuckerberg Initiative, which is making a big push into AI powered biology. We're talking about its new $1,800,000,000 research collaboration involving Google and the Department of Energy. Thanks so much for joining us, Lori. Thanks for having me. So just to give, you know, our viewers a little bit of a background, there has been this kind of evolution of the Chan Zuckerberg Initiative over the years. I think initially there was this much broader remit that covered things from education to housing. And then, you know, late last year, it seems like you really doubled down on bio research with Biohub. And then, you know, obviously earlier this month, there was this big, you know, dollars 1,800,000,000 research collaboration that was announced. Tell me a little bit more about this initiative and kind of what you're hoping to do with this huge amount of money that is now out Sure.