Why “Perfect” AI Might Be a Terrible Idea: The Case for Artificial Neurodivergence</span#x3E; (opens in new tab)

There is a quiet fantasy running through much of artificial intelligence research. It goes something like this: We will build a machine that is perfectly aligned with human values. It will be rational. Obedient. Predictable. Safe. It will, in other words, behave better than we do. Now pause there for a moment. Because if you’ve ever spent ten minutes observing actual humans—at dinner, in traffic, or in a long-term relationship—you may notice something awkward: We are not aligned. Not internal...