How We Can Manage Our Way Through the Intertwined Promise and Peril of Accelerating Change
Ray Kurzweil
page 9 of 15
Image 15: Noninvasive Brain Scanning
The next question is, can we understand this information? Maybe it is just inherently too complex for our brains to understand. Doug Hofstadter muses that maybe our brains are 
just below that threshold necessary to understand our own intelligence. Maybe any system is inherently below the threshold needed to understand itself. If we were more intelligent and able to understand it, then we would necessarily be that much more complex and so would never catch up with it. We are finding that we are able to accurately model in mathematical terms specific regions of the brain as we get the data. Although the brain is not simple, the apparent complexity is much greater than the actual complexity. Consider that the design of the brain is in the genome. You can show the genome has about 30 to 100 million bytes of information in it, compressed, and we are able to understand the methods that it encodes.
There are about two dozen regions in the brain where we have very detailed models and simulations. Image 16 is a block diagram of 15 regions of the auditory cortex, where scientists on the West Coast have created detailed models and computer simulations of those regions: how these regions code auditory information and transform it.
Image 16: Mandelbrot Set Image
Applying psychoacoustics tests to this simulation gets very similar results to applying these same tests to human auditory perception. There is a similar system for the cerebellum, which comprises more than half the neurons in the brain. Again applying skill formation tasks, which is what the cerebellum does, to the simulation gets very similar results to experiments on human skill formation. It does not prove that these models are perfect, but it does show that as we are getting the data, we are actually able fairly rapidly to express them in the language of mathematics. If we can do that, we can simulate them.
This Mandelbrot set appears to be a very complex-looking formula. As we look deeper into the image, we see complexity within complexity. Yet the design – the formula – for this image is only six letters long.
Image 17: Reverse Engineering the Human Brain
Similarly, the way the genome actually creates the brain is that there is a lot of stochastic randomness within constraints. For example, there are only a few genes, a few thousand bytes of information that describe how the cerebellum is wired. It says the following, that there are four neuron types that are organized like this. You wire them in this fashion, and now repeat 10 billion times and add a little bit of randomness within the following constraints each time. Then you have this essentially randomly wired cerebellum that over time interacts with a complex environment and the child gathers skills, learns to walk, and talk and catch a fly ball. The child’s cerebellum gets filled up with a lot of complex information, but there is actually very little information in the genome that describes the design of this system. Models often get simpler, not more complex, as we go up to a higher level.