Short answer: Dov Gertz says the current bottlenecks are data and model architecture, not compute. He predicts shorter times to patients, cheaper drugs and true personalization, with cancer likely benefiting first because it has the most data.
Watch this answer from Episode 32 (starts at 10:20) · Episode 32 notes and transcript
From Episode 32: From DNA to Drugs: How AI Is Rewriting Human Biology, starting at 10:20.
Jordan Metzner [10:20]: That's awesome. And so what does this look like three to five years, seven years out as the chips get better, as the labs get, faster to adopt, as, your technology improves, etcetera? Like, are we gonna see all new types of drugs? Are we gonna see new types of implementation of drugs? Are we gonna see just I don't know. I mean, I'm sure you guys think about the future probably all the time.
Dov Gertz [10:39]: Yeah. So I'll say first of all, I'll I'll explain the bottlenecks even before I explain the future. There are three bottlenecks. There's a data bottleneck. There's a architecture algorithm bottleneck and a compute bottleneck. So the compute bottleneck currently, we're not at a scale this industry is not at a scale that that that's the bottleneck. Text and image is in essence. But in our industry, we're still not there in terms of a bottleneck. We are in a data and, correct architecture bottleneck. So we're rigorously working on creating a massive dataset where we pay for massive data collection, and carry data for, creating the best dataset in the world and always innovating on the model architecture front to solve the best architecture, for that. So we have bottlenecks currently. Given that we solve the bottlenecks and the pace that we're moving forward, us and the industry, the bottlenecks are starting to we're seeing in the foreseeable future leaps in terms of these bottlenecks being solved. The future ultimately looks like is much shorter time for drugs to reach patients, much cheaper cost for drugs for patients because there's just gonna be more and more competition for drugs because it's gonna be easier to deploy them and much more personalized. Because now, Evan, that's the last that's gonna come, the true personalization. But we right now, drugs are, in essence, sometimes called personalized, but they're not.
Dov Gertz [12:10]: We are because of the way FDA approvals are built, we have to have, like, very, very robust inclusion exclusion criterias, and they aren't, in essence, very personalized. But as the time to market will and cost to market will reduce, we will see more and more personalization. That's my prediction. Obviously, these are predictions, and I'm a very optimistic person. But I do believe that we're going towards there. I also believe that that's the number one thing that we should be harnessing our capital and brainpower to solve. I mean, chatbots are really, really nice, and it's really cool that you can create a cell game in the spirit of the nineties within a minute. But can you actually improve your quality of life and length of life? I think those are the real questions that we should be asking ourselves.
Sam Nadler [13:02]: I was just gonna ask, is there maybe all potential illnesses and conditions will benefit from this technology, but is there a particular illness that we'd be familiar with that, is in the crosshairs on the horizon that, the patient population will immediately you know, once used properly, will see immediate impacts?
Dov Gertz [13:19]: Yeah. So that's a very good question. I think here, I can only hypothesize, but my hypothesis is based on data. So we do have the most data on cancer, and that's where I believe we'll be seeing the most value. Also, we see the just in terms of the market potential diabetes and is is a massive market and inflammatory bowel disease related issues and autoimmune disease in essence. I think neurodegenerative will be relatively later, and there's a lot of molecular barriers there and whatnot. But these are all predictions. It's very, very hard to tell at this point, and it's it's a balance of complexity of the disease and how much data we have available to model it.
