niplav comments on Provably Safe AI: Worldview and Projects

[niplav]

That argument seems plausibly wrong to me.