Steven Pinker -- How the Mind Works - Hampshire High Italian ...

176 | HOW THE MIND WORKSthe point where you are so close that further subdivision would not beworth the cost. Similarly, if you were trying to crack a combination lock,every number you bought would cut down the number of possibilities totry, and could be worth its cost in the time saved. So very often more informationis better, up to a point of diminishing returns, and that is why somelineages of animals have evolved more and more complex nervous isystems.Natural selection cannot directly endow an organism with informationabout its environment, or with the computational networks, demons,modules, faculties, representations, or mental organs that process theinformation. It can only select among genes. But genes build brains, anddifferent genes build brains that process information in different ways.The evolution of information processing has to be accomplished at thenuts-and-bolts level by selection of genes that affect the brain-assemblyprocess.Many kinds of genes could be the targets of selection for better informationprocessing. Altered genes could lead to different numbers of proliferativeunits along the walls of the ventricles (the cavities in the centerof the brain), which beget the cortical neurons making up the gray matter.Other genes could allow the proliferative units to divide for differentnumbers of cycles, creating different numbers and kinds of corticalareas. Axons connecting the neurons can be re-routed by shifting thechemical trails and molecular guideposts that coax the axons in particulardirections. Genes can change the molecular locks and keys thatencourage neurons to connect with other ones. As in the old joke abouthow to carve a statue of an elephant (remove all the bits that don't looklike an elephant), neural circuits can be sculpted by programming certaincells and synapses to commit suicide on cue. Neurons can becomeactive at different points in embryogenesis, and their firing patterns,both spontaneous and programmed, can be interpreted downstream asinformation about how to wire together. Many of these processes interactin cascades. For example, increasing the size of one area allows it tocompete better for real estate downstream. Natural selection does notcare how baroque the brain-assembly process is, or how ugly the resultingbrain. Modifications are evaluated strictly on how well the brain'salgorithms work in guiding the perception, thought, and action of thewhole animal. By these processes, natural selection can build a betterand better functioning brain.But could the selection of random variants really improve the designof a nervous system? Or would the variants crash it, like a corrupted byte