Neural networks and the human mind: new mathematics fits humanistic insight

It is asserted that the reinforcement learning or optimization designs are the only designs of values in understanding the human mind. They are the only designs capable of meaningful planning or foresight. They are also the designs which have led to the most exciting real-world applications in recent years. Within the field of optimization over time, two classes of design have proven useful in practice: (1) direct optimization, using generalized backpropagation to calculate derivatives of utility or performance or cost: and (2) adaptive critic designs, which approximate dynamic programming. After a brief review of neurocontrol and an explanation of reinforcement learning, the author asks what implications these designs have for the understanding of the human mind. He argues that this new mathematics is fully compatible with older deep insights into the human mind, due to humanistic thinkers East and West. There is no real conflict between new ideas from neural network theory or physics and deep, classical ideas from Eastern and Western cultures; in fact, the new ideas may provide a basis for helping understand and appreciate the roots of the classical ideas at a deeper level.< >