2 ms·

Google/Deepmind research still has a fairly large subgroup working on health problems, but most of that work is not given away (or even easily licensable) or pu

by dekhn 3y ago

Google/Deepmind research still has a fairly large subgroup working on health problems, but most of that work is not given away (or even easily licensable) or published in a way that competitors could reproduce. What Isomorphic, the spinoff, learned pretty quickly is that protein structure prediction is not, has never been, and is unlikely to be, the most critical blocking step in developing new drugs. That won't stop pharma from investing biobucks (virtual dollars that are gated by milestones) in them. Right now pharma is terrified because their pipelines are drying up, their blockbusters are going generic, and the recent rates of new discoveries leading to new candidates (target diseases/bio pathways) are dropping, even as they invest more and more into automation and machine learning. (By the way Mike- I always did wish you had been able to run your "math problems" on exacycle, as a way of monetizing idle cores)

mike_hearn 3y ago
Ahhh, I wondered about the importance of structure prediction for a long time, ever since reading those blog posts by a PSP researcher in academia who was complaining that big pharma didn't invest in the field. He was asking "why was it Google who solved this problem and not a Pfizer or Bayer?" His posited explanation was a sort of cultural/institutional disinterest in advanced R&D, but that didn't sound very plausible compared to the other possibility that they just didn't care much about the problem to begin with. > That won't stop pharma from investing biobucks (virtual dollars that are gated by milestones) in them. I didn't quite follow this, could you maybe elaborate? What is a virtual dollar in this context? Do you mean they make promises of future payments if certain goals are achieved, but the dollars never become real because the goals aren't achievable? Re: math problems. I remember looking at Exacycle and wondering about that :) but concluded it would have been self defeating, as it'd just have resulted in Google dominating the network and that'd have been seen as an attack.

by dekhn 3y ago

mike_hearn 3y ago

Ahhh, I wondered about the importance of structure prediction for a long time, ever since reading those blog posts by a PSP researcher in academia who was complaining that big pharma didn't invest in the field. He was asking "why was it Google who solved this problem and not a Pfizer or Bayer?" His posited explanation was a sort of cultural/institutional disinterest in advanced R&D, but that didn't sound very plausible compared to the other possibility that they just didn't care much about the problem to begin with. > That won't stop pharma from investing biobucks (virtual dollars that are gated by milestones) in them. I didn't quite follow this, could you maybe elaborate? What is a virtual dollar in this context? Do you mean they make promises of future payments if certain goals are achieved, but the dollars never become real because the goals aren't achievable? Re: math problems. I remember looking at Exacycle and wondering about that :) but concluded it would have been self defeating, as it'd just have resulted in Google dominating the network and that'd have been seen as an attack.