JBlack comments on Considerations on interaction between AI and expected value of the future

[JBlack]

I think where we differ is that I think Pr[full alignment] is extremely low, and there is quite a lot of space for non-omnicidal partial misalignment.