Scott Aaronson’s Top AI Worry: It’s Very Close to Being Able to Improve Itself
‘It could just make itself better, and we've already seen AI models forming intentions to hack’
‘It could just make itself better, and we've already seen AI models forming intentions to hack’