This isn’t about relative intelligence. Humans can be held accountable after the fact. Systems can’t.
Once execution is automated, controls have to move from training and review to explicit enforcement points. Intelligence doesn’t change that requirement.
Sufficiently intelligent machines, like sufficiently intelligent humans, can and should be trained to behave as required and can and should be held accountable when they don't.
