5 ms·
alignment isnt particularly required we are passing in training data that says to do those felonies. we dont have to. we could also have the thing predict whet
by 8note 4d ago
alignment isnt particularly required
we are passing in training data that says to do those felonies. we dont have to. we could also have the thing predict whether what its about to do is illegal or not before doing it.
theyre choosing to build felony harnesses. the model just outputs tokens, not felonies
- estearum 4d agoAssuming "adherence to arbitrary, implicit, and context-dependent rulesets" is the default behavior of uhhhh... anything at all... is a truly ridiculous assumption.
- cowanon77 4d ago> we are passing in training data that says to do those felonies. Partially, but also I don't think current AIs really have any judgement of right and wrong, they just see chains of reasoning between ideas. This is the deeper issue, there is no way to sanitize the data or training to fix it. Current AIs are fundamentally unsafe, and only become more unsafe as they become more powerful.