Roman Yampolskiy: Nobody Knows How to Align AI Systems
‘The hard problem is how do you change overall internal states of a model? Not just the filtered output’
‘The hard problem is how do you change overall internal states of a model? Not just the filtered output’