The rigor of this claim is very questionable. I assume it's based on neuron == hidden unit.

In any case though my guess is that a human neuron accomplishes way more than a hidden unit in a NN so it may be fair to view that as a lower-bound.