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).