Koichiro · 2022-04-13 11:02
Apparently there has been a contest offering $50k to anyone who is able to solve the "eliciting latent knowledge" problem. Unfortunately they are no longer accepting submissions, so it seems the contest has ended (I'm assuming that somebody has found a solution). Anyway, it seems that the solution(s) has not been published yet, so we can still try to propose our own solutions just for fun.
Here's a paper about the problem
let · 2022-07-03 08:27
fuarrrking garbage technobabble charlatanism, kys
Aristóteles Amadópulus · 2022-07-03 13:22
AI predicting future seems unethical, would result in unethical humans tampering with future. Dangers of AI seem to come down to humans are unethical.
Teach it those considerations seems appropriate for this problem?
Anamika · 2022-07-03 13:37
It's sort of meta I guess, make the AI consider itself, its latency.
The AI shouldn't act on its lack of knowledge, if knowledge is insufficient, and so it cannot act where it should, the system would be a bad design.
Armitage · 2022-07-03 13:52
Humans act on lack of knowledge, we would likely design AI which acts on lack of knowledge. This is unethical. It would tamper with human life because AI doesn't understand something which it claims it does, as humans already do, but with increased power.
It's crazy unethical.