Interesting question of what an AI-policy advising function might or should look like. Some thoughts:
- Advising: Google is a primitive form of AI policy advice that is certainly already used (every policy analyst had better hope to be much better than that Minister’s latest Google search when they enter the Minister’s office).
- Recommending: Would a Minister be any more likely to listen to an AI analyst than a human one?
- Deciding: There are Decisions and there are decisions. It doesn’t seem likely that the Minister could stand up in Parliament and say “it wasn’t me, the Computer did it.” But where it’s a matter of applying the law impartially, that’s one recipe for good governance — even if the decision is made by an AI agent.