Comment by simondotau

A model should understand multiple perspectives on morality and avoid prescribing a single one where there’s no overwhelming prior consensus.

Alternatively, they should be trained on my opinion on everything. That would also be acceptable.