To Simulate a Cat Brain

Consider me completely unsurprised that IBM was the company that got the DARPA contract to build a cat brain simulator. All too often, DARPA "open" BAA's sound basically like this: "DARPA seeks a contractor to build a Lockheed Fighter Plane, built in Lockheed's factory by Lockheed employees."
This is because companies have people buttering up DARPA employees 24/7 for contracts. One way to do that is to suggest research to DARPA, which they might fund if they find it suitable to DARPA's mission and also quality research with high-risk/high-reward potential.

Anyway, so IBM got tax dollars to build a cat-brain simulator using their supercomputers, and their lead researcher announced last week that they had "10 billion neurons connected by 10 trillion synapses." Ignoring the entire part of this article where other neuroscientists blast the research as a "scam" and a "hoax" let me just make my own criticism: the whole project is occurring in virtual space.

What good does a virtual cat brain inside a building-sized supercomputer do me? The real research is in making brain-mimicking hardware the size of an actual brain. And though my understanding of electronics is low, I believe the problem with that is the non-plasticity of integrated circuit layouts coupled with the low number of junctions between processing units.

My solution: nanobots. Build nanobots with hundreds of legs, each about the size of a neuron. Each nanobot connects to other adjacent nanobots via its legs. Control signals sent to the nanobot can alter the leg connections of that bot (or double/triple/etc the connections between two bots to strengthen the electrical path and lower resistance) leading to memory making pathways. The nanobots don't need their own power supply, they could each contain a radio frequency receiver that would generate power merely by being in contact with radio waves (essentially the way RFID works).

Nanobiomecharobotic neuroengineers: ready, GO!


_