This is a nice point that I haven't seen before. It's interesting to regress AI to the simplest form and see how we treat it as a test for the more complex cases.