10,000 neurons in a jar

According to news, a model of a neocortical column has been simulated in software. However it has no correlation to animal behavior. The simulation has 10,000 neurons and is only at the cellular level, not the molecular level. It would have to be 2000 times more neurons to reach the count of a rat brain, and 400,000 times as many neurons to reach human neural capacity. Let alone the architectural specialization of animal brains. Based on these estimates, I agree with Koch, that full brain simulation in 10 years is "ridiculous."

Also, I believe that simulating human brain is premature. Why not follow the path of animal evolution? As I've said for the Turing test, don't try to simulate a human's intelligence. Not yet. We haven't even successfully simulated a rat's intelligence. I think we'll reach our goal of simulating human intelligence faster by starting with the next incremental step in simulation of biological intelligence from what we currently have. Let us climb up the evolutionary ladder one rung at a time, not blindly leap for the top rung. I personally have not heard that we've even simulated a goldfish's neural capacity that corresponds to actual behavior.

Without linking the brain to a system that operates in an environment, then the neural behavior is crippled. At a minimum, the neocortical column ought to be exercised solving problems that brain faculties are well-adapted to. At a maximum, it'd be fascinating to see a robot with a brain. There's surely some Tin Man in robotics looking for a brain.

/science - permalink - comments