It was in an interview with Lex Fridman. I don't understand how an MIT AI researcher could hear this and keep a straight face, maybe he's never used Autopilot...You'll get nowhere trying to suss out any logic from his tweets. They are aspirational, not driven by any real data or breakthroughs.
"Do you see Tesla's full self-driving as still for a time to come requiring supervision of the human being. So, its capabilities are powerful enough to drive but nevertheless requires a human to still be supervising, just like a safety driver is in other fully autonomous vehicles?"
"I think it will require detecting hands on wheel for at least six months or something like that from here. Really it's a question of, from a regulatory standpoint, how much safer than a person does Autopilot need to be for it to be okay to not monitor the car."