Hacker News new | ask | show | jobs
by yuffffley 5 hours ago
I remember that.

That was when they realized the deep learning was largely unnecessary, and they could just use their massive compute resources to brute force the problem space.

Proving that we would greatly benefit from using our compute resources for science rather than showing ads, and then we just kept showing ads.

5 comments

AlphaFold is based on deep learning and it's not brute force.
If AlphaFold really is brute force on known protein problem space, would it then be usable as a model for novel proteins?
What brute force? Any citations?
You could argue that training SOTA LLMs is pre-bruteforcing every problem everywhere all at once.