If your AI made the wrong decision tomorrow, could you prove what it saw, what it was allowed to do, what it actually did, and why?
If your AI made the wrong decision tomorrow, could you prove what it saw, what it was allowed to do, what it actually did, and why?