[barelyauser]

One thought about AI. Testing for correct answers is not a useful metric for AI. People can learn something that is wrong as easily as something that is "less wrong", as long as it makes sense. Sometimes things that are very counterintuitive are proven correct, and our intellect has to kind of reason a way to believe it.

Also, AI doesn't need to be "human" to be very useful. The argument of birds vs. planes comes to mind.