Comments on “Motivation, morals, and monsters”

“Instead, it is an justification”

should be

“Instead, it is a justification”

Most of this page absolutely agree with, but I wanted to quibble about one point:

Attempts to specify what “values” we want an AI to respect fail because we don’t have those.

Seems a bit too strong. We obviously have something like “values”, they may not be coherent or consistent, but they are real enough. They may not be top-level goals from which all lesser goals are derived as subgoals, but a value at root is just a preference for some states over others, and I think we have those.

If you are saying that all computational models of values are bad and don’t capture the richness of lived human existence – I’d be down with that, but that is not the same as saying that values don’t exist.