How or when would we confidently say, if it’s even possible, that we have “solved” the problem of AI alignment?
How or when would we confidently say, if it’s even possible, that we have “solved” the problem of AI alignment?