Joshua Achiam argues human inconsistency limits perfect AI alignment

Joshua Achiam, on X, argues that AI alignment is hard partly because humans do not fully understand what they want and cannot state it perfectly.

He says that even an AI that could read minds would not produce a coherent extrapolated volition (CEV) that people would accept, because enforcing consistency would remove the inconsistencies he considers part of what makes life valuable. He concludes that humanity will have to muddle through under high uncertainty. The post is a single report, and no other sources or responses are cited.