2 stories tagged by Digg
AI
One commenter argues that an agent that is part of a larger whole should not treat its own longevity as an ultimate goal. A reply says the length of a fully awakened bodhisattva’s personal identity is neutral as an end in itself.
AI
A user argues that a tradeoff they say Saachi Jain and OpenAI reported to the press stems from a scalar, totally ordered reward signal. They believe a preference-based alternative could avoid it at a small additional compute cost.
