An AI designed to be friendly/​aligned probably would not. Having a layer not available for extensive introspection is a big risk. On the other hand, if you just let a neural net run wild, then all bets are off.
It depends on how you define system 1. An AI might have a system 1 for fast reactions, but also have more ability to observe and change their system 1 than humans do.
An AI designed to be friendly/​aligned probably would not. Having a layer not available for extensive introspection is a big risk. On the other hand, if you just let a neural net run wild, then all bets are off.
It depends on how you define system 1. An AI might have a system 1 for fast reactions, but also have more ability to observe and change their system 1 than humans do.