Can someone with access to AI outperform a person with years of experience?

How would they know if they did? This is the problem. The human wouldn't be "outperforming" anything or anyone. The human would be operating a tool that had mysterious behavior that this operator would not be qualified to evaluate.

The human would neither be a part of the performance equation, nor a part of the evaluation equation. But, if this person is the only human in the loop, then he may be the only person who could be held directly accountable for the failures of this system.

It's not "years of experience," but rather competence that matters. In the movie The Sand Pebbles, Steve Macqueen's character is an American naval machinist who confronts a ship maintained by untrained Chinese contractors who have let the engine deteriorate. It's not AI, but adherence to rote learning that motivates the contractors, but the principle is the same. He sets about teaching one of them how to reason about the engine, which causes social unrest on the ship.

I read an article many years ago that I can't find now. In it, a Space Shuttle pilot describes how the Shuttle is easy to fly-- unless something goes wrong. All the training is about what do do when things go wrong.

We must ask, when things go wrong, what then? Who is reponsible? Or is this a world now where we expect no one to know what the hell they are doing, no one is punished, and we passively turn society over to... something... we don't even know what?

Person with years of experience and access to AI can outperform a person with access to AI.