Thinking - Robert Hamill

Only 13% of Brain is Used

   In common discourse I occasionally hear that only a small percentage of a person's brain is used, usually the number is 10% or for sophisticated common discourse I hear 13% used.
   Let's look at some of the things going on between our ears. First, note that words and the stream of consciousness is just the tip of the iceberg. The human brain is a neural network.
   Neural nets excel in pattern recognition. They are innately parallel, yielding speed. Neural nets regularly recognize matches when the comparison images are less than 100% identical.
   Using the neural net model, let's consider recognizing the face of a friend - the classic visual matching problem.
   The human optical system is marvelously complex. You will have to allow me some latitude to simplify both the visual system and the associated neural network processing. A typical person has 125 million rods in each eye. These rods are on/off detectors. Using a computer model we could say that the eye's resolution is 125 million pixels - but this does not go directly to the brain.

   In about 4 layers, the 125 million bits of visual information are condensed down to 1 million bits, which travel the optic nerve to the brain. A biological neural net performs this concentration of information. Approximately every 5 photoreceptors feed their output into an accumulator cell at each layer. The accumulator cell also has cross-links with adjacent accumulator cells. Grossly simplifying, if 3 inputs to an accumulator are on and 2 are off, then the accumulator is on, albeit with an associated probability.
   These accumulators also rise to another layer of accumulator cells, which in term sum to another layer, until finally the optic nerve is reached. These 1 million bits of information are transmitted to the brain. There we save images or more accurately representations of images, of which one is the representation of the face of our friend, stored in a neural net capable of holding 1 million bits of information.
   A straightforward match between the two one million-bit images can be easily envisioned. Not every bit (or node) will match. There are daily changes to one's face. There are reflections and shadows. Our visual recognition system must be able to ignore those irrelevant distractions. Neural nets