Whether AI is reasonable is something you can test. Whether it is intelligent or creative is a debate that never ends.

• Intelligence and creativity are undefined enough that the debate never resolves. Reasonableness is closer to measurable.
• An LLM is trained on the sum of written human output, from the worst of the internet to Ulysses, so it has some claim to a centrist, averaging pull.
• Research is mixed: LLMs predict centrist politicians' votes better than fringe ones, but abstract questionnaires can pull them left.
• The mainstream a model gravitates to is itself shaped by training data and alignment choices, so it isn't neutral either.

• #ai
• #bias
• #reasoning