Critics compare her skepticism to early deep learning doubts.
History suggests that experts consistently underestimate exponential technologies. Biology won’t be the exception.
AI isn’t innovating… yet. Nobel Prize winner Jennifer Doudna pours cold water on claims that AI will end disease and cure cancer in 48 hours. Find out why, on the latest episode of The Circuit.
Full show: https://youtu.be/n-hWHV2ZKOA?is=oD7WT3xzU2HOVwq6
Positive users defend AI's medical potential by citing Hassabis's successes over Doudna's skepticism, while negative users dismiss rapid cures as overhyped and demand concrete proof after repeated tech failures.
No Digg Deeper questions have been answered for this story yet.
Award-winning career NLP researchers were wrong first about what deep learning, and later about what LLMs could/could not do with human language. Ignore experts and do your thing.
AI isn’t innovating… yet. Nobel Prize winner Jennifer Doudna pours cold water on claims that AI will end disease and cure cancer in 48 hours. Find out why, on the latest episode of The Circuit.
Full show: https://youtu.be/n-hWHV2ZKOA?is=oD7WT3xzU2HOVwq6
@9far_gg The most expensive phrase in science isn’t “AI is overhyped.”
It’s “this field is too special.”
History hasn’t been kind to that prediction.
@QuanquanGu Historically the opposite has been true. Tech often overestimates its ability to turn into cures. Human genome project, CRISPR and many others claimed to cure vastly more diseases than they actually have. AI might be different, only time will tell.
@deliprao LLMs? They are just stochastic parrots. You disagree? You don't understand how "research" works. - NLP Expert
@QuanquanGu Exponentials only exist for very short periods in nature for the most part. There won’t be a singularity because there will be a bunch of bottlenecks that show up along the way.
@QuanquanGu Biology has been around for much longer than history though.
@deliprao I remember psychology seniors perseverating about rigorous stats and reproducibility hyper accounting.
We were engineers; we shouldn't be sloppy, but also we are not the same.
@deliprao Experts in domain A and experts in domain B are very, very different.
@vineettiruvadi Yet majority of human-conducted psychology studies are not reproducible.
@QuanquanGu Sure, just cure one disease with AI first then talk big
@hu_yifei My biggest Twitter achievement was using my platform to counter the parrot crowd. Glad they migrated elsewhere.
@QuanquanGu Sorry, I am aligned with Jennifer Doudna
cc: @emilychangtv
@deliprao The AI guys are almost always wrong when it comes other fields. Listen to the experts in the core fields, not the AI guys, especially not the AIinfluencers on social media.
@deliprao That's a them problem - why would I ignore expert bridge builders and engine engineers because psych sucks at stats?
@deliprao But another Nobel prize winner Hassabis clearly used AI to transformed the field. I trust Hassabis more
@QuanquanGu 💯🚀
@deliprao I mostly agree with Doudna's take in the interview. She didn't claim that LLMs were incapable of innovation like Gary Marcus or Yann Le Cun. She just said they weren't there yet, especially at biology where LLMs lag compared to their capabilities at math and coding.
@deliprao 高手往往被成见绊住，还得是实干出真知
