Quantum compute could be being leveraged to do things like accelerating search for gradient descent. This would allow very rapid learning in polynomial time. There are also quantum algorithms that can more or less serialize parallel search, etc.
The objection is usually that biology is hot and noisy and wet and that this is a poor environment for QC. This assumes that it’s doing QC in a remotely similar way to how the quantum computers we are trying to build would do it. Billions of years of evolution might have found ways to harness quantum phenomena for information processing that are quite radically different. It’s all analog for starters, so nothing like a “qubit” or rigidly defined circuits.
Maybe some biochemical reactions are structured so as to invoke and amplify useful quantum effects that allow things to happen informatically that would be much slower or more costly without those effects.
Without understanding what’s happening this would just look like noise and luck to us. I do know that there are “unreasonably effective” enzymes and reactions involved in things like DNA repair, though it’s been a while since I read about this. I think there are cases where repair complexes find DNA errors more effectively than can easily be explained by basic chemistry and chance, and we are not sure exactly what’s going on. Maybe there’s something involving quantum information processing happening somewhere.