If you perfectly learn and perfectly maximize the referent of rewards assigned by human operators, that kills them.
Not checkable An opinion, value judgement or statement with nothing to check.
Not checkable: An opinion, value judgement or statement with nothing to check. Not scored.
Cite this claim
Eliezer Yudkowsky, “If you perfectly learn and perfectly maximize the referent of rewards assigned by human operators, that kills them.” (said 10 June 2022). Verdict: not checkable. Reliability API, claim c_e56677b1b7d5, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_e56677b1b7d5
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
Verdicts are made automatically; how verdicts are made.