If you run an LLM (with temp >0) and use a true random generator for seeds, it will still tell you 1+1 is 2.

You mean mind is deterministic and cancels all noise? Then what quantum physics would do there?

You don't need to cancel all the noise to solve 1+1 correctly. The model in the example is non-deterministic.

I don't know Penrose's view deeply enough to defend it further (view which I for the record do not share).

He's just rambling.