UNCANNY SKILL

Evening Star, Issue 21689, 7 April 1934, Page 18

UNCANNY SKILL

THINKING MACHINES WITH LIKES AND DISLIKES WORK IN PSYCHOLOGY The “ thinking machine ” exhibited by Xor man B. Krim, of the Massachusetts Institute of Technology, before a section of the American Institute of Electrical Engineers at Cambridge. Massachusetts, deserved all the publicity that it received, writes Waidemar Kaempffert in the 1 New York Times.' But to sav that it is the first and only one of its kind or that it actually “ thinks,” despite its uncanny ability to remember and forget and to express a sort of electrical pleasure in what it likes and a sort of hatred for what it dislikes, is simply to abuse words. To tell just what these alleged thinking machines arc we must go back to Pavlov and his conditioned reflexes. That great Russian scientist put a tube into a dog’s mouth to measure the amount of saliva generated by an attractive piece of meat. The saliva flowed not because the dog willed it, but because mechanisms over which he apparently had no control made it flow. Physiologists and psychologists therefore speaw of reflex actions. When a doctor taps the nerve.just below your knee and makes your lower leg kick lip calls reflex action into play. There are other reflexes, such as involuntarily swallowing and winking. Biit all these are unconditioned reflexes. Experience has a great effect upon them—“ conditions ” them. Pavlov decided that ages ago a dog had to have meat in his mouth before saliva began to flow. .As a result of inherited experience a signal ultimately sufficed —the mere sight of a morsel, for example. PAVLOV’S WORK. On the basis of this supposition, Pavlov built up in a dog strange behaviour. With disagreeable acids he cultivated responses of avoidance; with agreeable food, responses of attraction. Merely by mixing responses he could reduce a dog to a state of nervous hysteria. Bromides restored him. He got dogs to distinguish clearly between circles and ellipses, to rejoice in the former and to run away in horror from the latter because of- their associrw tions. In all these cases a message from a sense organ passes to a sensory centre of the brain and thence .to a motor centre by which ah order is given to the salivary gland to secret juice it eating is in question. Conditioned reflexes of this kind are usually less powerful than the natural or unconditioned kind. The essential difference between the two is the deliberate switching of nerve impulses from a sensory to a motor centre. Hundreds of experiments have led Pavlov to the conclusion that different kinds of habits, based on training, discipline, and education of any kind, are nothing but a long chain of conditioned reflexes. Associations once established become automatic and are persistently reproduced. If conditioned reflexes are indeed mechanical then it ought to be possible to construct a machine that will reproduce them. Reasoning thus, about a score of psychologists and physicists have constructed extraordinary combinations of electric cells, lamps, batteries, and switches that behave in a very human way, however inhuman the models may look. Professor Clark L. Hull, , of Yale, has done the most notable work in this field. Dr H. D. Baernstein, R. G. Krueger, George K. Bennett, Thomas Ross, and Lewis B. Ward have followed in his steps. Norman Krim' is therefore hut one of a bright hand of kindred spirits. CAN BE TAUGHT. To describe these machines without the aid of a technical diagram is impossible. The reader must take the writer’s word for it that any of the dozen systems developed conduct themselves much like Pavlov’s dog. They can be “ taught ” to react to a signal, an external impulse. In fact, they have a curve of learning much like that of a human being, which means that every day in every way they get better and better. Some not only learn, hut irradiate neighbouring sense organs and thus educate them. There is something much like forgetting, too. It is even possible to construct a machine which will thread a maze after making mistakes and which exhibits likes and dislikes. The experimenters in this field hardly imagine that their machines are exact duplicates of the human thinking or reflex apparatus. They call them mechanical, electrical, or electrochemical analogues, depending on the construction of the machine. Yet these machines have their uses. They may be an aid in settling the old controversy between the vitalists and the mechanists, the one believing in such a thing as soul, spirit, mind, consciousness that governs human behaviour, and the other preaching that soul and spirit are but mystical names for actions that .we cannot as yet explain ■ mechanically. To prove their point the mechanist must bring forth operative models that will adapt themselves to a given set of circumstances —shrink from what they do not like, approach what they do like, and behave just as if they were thinking organisms aware of their environment. And that is one object Dr Hull and his co-workers have kept in view. Although reasoning by analogy is a deceptively bad form of logic, there can be no doubt that these collections of lamps, batteries, and switches actually duplicate the mechanisms of response that differentiate human beings from stones. Dr Hull thinks that there may be a practical use for devices designed to settle the old controversy between the vitalists and the mechanists. “It is not inconceivable,” says he, “ that in the demands for higher and higher degree of automaticity in machines constantly being made by modern industry, the ultra-automaticity of the type of mechanism here considered may have an important place.”