Wei Dai comments on Stable self-improvement as a research problem

[Wei Dai]

Paul, do you have a list of “other aspects of the AI safety problem” that you think should be prioritized higher?