lukeprog comments on Stuart Russell: AI value alignment problem must be an “intrinsic part” of the field’s mainstream agenda

[lukeprog]

I worry about the phrase “provably aligned with human values.”