If you think of transistors as analog computational devices, 100 is not that small. There are animals with ~200-300 neurons, so it's at least within the range of possibility to have a fairly autonomous robot with only 100.
A neuron contains thousands if not millions of transistor-equivalents in the form of gene level and ion level switches. I'm not saying that you couldn't do it with 100 transistors, or that everything in a neuron is needed for this application, but I just wanted to clarify.
Right, the transistor is not fully simulating a neuron, just a first order approximation of the electrical impulse response. It's close enough for government work, as they say.
