I agree. Also, I’ve heard LeCun arguing that a super intelligent AI wouldn’t be so “dumb” as to decide to do something terrible for humanity. So it will be 100%

I agree. Also, I’ve heard LeCun arguing that a super intelligent AI wouldn’t be so “dumb” as to decide to do something terrible for humanity. So it will be 100% resistant to adversarial attacks? And malicious actors won’t ever train their own? And even if we don’t go all the way to super intelligence, is it riskless to progressively yield control to AIs?