Comment by enum
6 hours ago
The real question is how do you come up with a credible 3-5 year research program that is unlikely to be scooped or become pointless overnight.
3-5 years is the period of a grant, and grants have to make research progress, or you don’t get the next grant.
Well, in this case, I think the problem and what should be updated is how grants work. They were stupid before, but now they are even more.
Stupid, but could be fun: grants are given for how amusing your proposed prompts are.
Everyone should have a portfolio of prompts that are indecipherable by other humans but when fed to a frontier model, produces shocking one paragraph english version of a 10000 line lean proof
Obfuscated prompt grant contest
Either fun or totally random. Unfair but at least cheap.
SIGBOVIK with extra steps.