I'm curious how you'll distinguish between these outcomes:

• Human assistance to the AI is allowed

• AI assistance to humans is not sufficient for resolution

If an AI writes most of a proof but the result has slight issues requiring a human to think hard to fix, who did most of the work? Will you try to identify the hardest crux of the problem and determine whether it was overcome by the AI or the human? What if the human does not reveal exactly which parts were AI-driven?

I think a likely scenario is that a problem could be solved by a human working in tandem with LLMs, and that it would be implausible for either party to have solved it on their own. I'd put mana on Yes if it's deemed that "the human would've been unable to solve it without the assistance of AI" resolves to Yes.

No, and it will not be close to either.