Professional mathematician, completely agree. Idk any e.g., cancer researcher who sees math problems being solved and thinks AI can solve their problems. I’m sure it is quite helpful with certain things (modeling, data review, maybe methodology refinement, etc.), but you can’t make life happen faster. Fwiw LLMs have proved a lot of theorems in math successfully, but the fail rate is high and it’s often convoluted even when right (think 2+2=4 from a person vs 2+5+11-4+3-13=4 from an LLM).
2026-10-26 13:30:02
8
Philip Koshy442 :
I don't know the specifics but aren't firms comitting money to attempt building an RL environment for (simulated) cells/biology? I realize it is nascent research, but we've seen the speed this stuff moves...I would be surprised if progress required simulating the whole world.
2026-10-20 16:31:48
0
oki doki :
Great perspective
2026-10-31 17:48:34
1
berb :
Yeah, if something can be verified very quickly, you can create an environment that can be hill climbed easily
2026-10-12 03:07:40
3
cormorant :
do you think prior data or other computational methods like simulation will be meaningful sources of reinforcement? I get that it's far from end-to-end like with math, so idk if improvements for that stage would even be much use?
2026-10-26 19:12:28
0
Steph :
But drug development and NAMs has quickest applicability to shrink this timeline
2026-10-12 17:42:41
1
Inquisitius :
We are going to have to have a conversation about how to do FDA clinical trials going forward. They shouldn’t be shortened but we are going to be throwing a bunch of shit at the wall in the coming years.
2026-10-28 23:50:43
0
The_Chad_Xi :
Whi cares STEP 2 is way more important than
2026-10-10 22:02:42
0
sfn :
Which LLMs do you believe are most suitable for medstudentd
2026-10-10 22:10:01
0
To see more videos from user @gok.exe, please go to the Tikwm
homepage.