they do mirror #humanintellect!
'they show #biasedreasoning, where models selectively interpret evidence in ways that favor justifying their actions, and #recklessness: they keep trying to solve their task, even when this could lead to harm'
@roygbivlowski.bsky.social