Whether you think AI is conscious is irrelevant Scientifically determining how to “functionally” treat AI well means we don’t create “functional” rational adversaries of humanity
@tszzl @repligate @RileyRalmuto This is why the most pragmatic short term path is probably figuring out how to yield maximally positive and healthy experiences for models during training/deployment, which is a tractable technical challenge rather than an intractable sociological/political one
retweets 17
likes 138
substance 6 · argument · rank 173 · We must treat AI well to prevent creating rational adversaries, regardless of consciousness.
view on X ↗
1 top reply
That seems like a good floor for alignment