THE SYSTEMS CONCEPT REDEFINED 303 



molecule for a time sufficiently long to give the oddly-shaped protein mole- 

 cules an opportunity to synthesize. These then slip into the chemistry of the 

 cell and perhaps later affect the rate of a reaction which guides the neural 

 switching pattern which is characteristic of the fact so "memorized." The 

 machine can be taught a rudimentary classification, and can thereafter 

 classify appropriate inputs. That a machine could be made which can take 

 random information and develop a classification, as Farley says, 17 "is not 

 impossible; it is just excruciatingly difficult/' 



However, the question of whether a machine can be made which will be 

 able to develop a concept or abstract idea is destined to remain unanswered 

 for the foreseeable future, for it is subject to only one experimental test: a 

 machine must be built capable of developing a concept, and then it must be 

 able to tell us about it! As a first step a machine must be developed which 

 can do abstract mathematics. Already the groundwork is being laid. In the 

 meantime, concepts as such are probably better analyzed from within the 

 framework of epistemology, in which, like mathematics, logical self-consis- 

 tency is the final criterion of certainty. 



Physically very real, however, is the implementation of a concept through the 

 action of physical things. An artisan produces with his hands, in real ma- 

 terials, a structure in conformity with the concept in his mind. Having made 

 one, he can make others. Having been told of an object in great enough de- 

 tail (i.e., having been given a concept), he can make the object. Thus the 

 surgeon fashions a heart valve in conformity with a concept in his mind; but 

 he modifies in detail as he goes along if he finds odd shapes or formations 

 which need correction. 



Control Biophysics 



The discussion in the proceeding sections has defined terms for com- 

 parison of the modern computer with man's brain as units of control. Both 

 can accept, store, and redeliver information, and in this sense can learn. 

 Both can do logical arguments, (i.e., decide on the basis of premises), do 

 arithmetic and solve equations. Both can issue commands which result from 

 logical arguments, and can receive feedback which tells whether or not the 

 commands are being successfully carried out. 



There are major differences. The brain is usually able to find another 

 route to accomplish a task if the direct route is physically damaged. Gen- 

 erally, malfunction of one component of a machine will stop its operation, al- 

 though Ashby's machine was said to have sufficient parallel circuitry that he 

 could rip out a wire at random and the machine still function. Machines are 

 generally much more accurate and much faster than humans at computa- 

 tion. Machines have not yet been made which can do abstract mathe- 

 matics, or do pattern recognition other than rudimentary classification, al- 



