Another possible use of quantum implementations of neural networks might be as a way to understand what the brain is doing. The hypothesis would be that we might get a better idea of the function of some of the brain's features if we view them as implementing a quantum learning machine. But is there any evidence so far that the brain is sensitive to quantum effects? Not really. There is the well-known study [Baylor et al., 1979] that shows that a single photon striking the retina of a toad is sometimes sufficient to trigger a nerve impulse, but in humans, this phenomenon seems to be suppressed by noise filtering [Hecht et al., 1941].
But as Penrose [Penrose, 1989, p 400,] points out, this does show
that there are some cells in the human body that are sensitive
to individual quanta, and therefore the possibility of quantum-
mechanical effects in the brain is still tenable. But we would
be making the
task unnecessarily difficult if we, like Penrose, required that we
find neurons that are sensitive to a single quantum. As discussed
in 
3.4, quantum computation can still occur in the cases
where an aggregate of phenomena (an entire interference pattern,
rather than one quantum) is required to yield an output. Many of
the advantages of quantum computation would still apply in such a
situation, such as communication via instantaneous forces, rather
than wires.
If quantum learning networks are a more plausible model of brain activity than mere quantum computation, it may have little to do with the fact that the learning algorithm is currently called a neural network learning algorithm. Most likely, the bits of the brain that would correspond to the ``neurons'' in the algorithm would be sub-cellular phenomena. The extra advantage of quantum learning as a model of brain activity would rather derive from its sub-symbolism, and from the fact that it emphasizes learning.