Mammalian brains, particularly human brains, are the most complicated objects known in the universe. Every generation compares the brain to the most complicated piece of technology so far invented. In the 17th century, Leibnitz compared it to a water-mill. At the end of the 19th century, Freud compared it to a hydraulic system. In the 1930s it was compared to a telephone exchange, in the 1960s to a digital computer. Most recently it has been compared to a neural network or parallel-processing system, a development in computer technology partly inspired by (but not necessarily intended to simulate) a living brain. How useful is this analogy?
Neural networks, like brains, have no central processing units of the kind found in digital computers. Different areas of the network, as in the brain, engage in democratic dialogue. And like brains, neural networks can detect signals, recognise patterns, interpolate data, make predictions, guide movement on the basis of visual information, and even synthesise speech. But they are not really like brains. A one-year-old child's visual processing capacity far exceeds that of any computer. Different areas of the human brain detect and interpret form, motion and colour in a single visual stimulus. The fine division of labour here, and the vast number of neurones involved, distinguishes the brain from a neural network. Also, a three-year-old child's language production, which again involves very fine division of function among closely related brain areas, is qualitatively different from anything that a machine can do. In particular, neural networks do not remember or learn in anything like the way brains do.
This is not to say that computer models or analogies of the brain are useless - or that there is anything wrong with neural network systems. In the final chapter we shall explore a computer metaphor that recalls our general model of the living state. However, brains have to be studied as objects in their own right. Brains are brains. They are unlike any piece of technology we have or are ever likely to have. They are not soggy computers, any more than they are soggy water-mills.
Was this article helpful?