On Comparing the Brain with Machines^ 



By D. M. MacKay 



Kin^s College, University of London 



In discussing this subject, three questions must be distinguished: 

 How far does the brain resemble existing mechanisms, such as elec- 

 tronic computing machines ? How far would it be possible to imitate 

 human behavior with a suitably designed artificial mechanism, or 

 "artifact"? How far is it possible to envisage a model that would 

 imitate human behavior, and also work internally on the same princi- 

 ples as the brain ? 



Only the third question is here discussed in detail. The first 

 is of trivial interest. The second may be answered by saying that 

 as soon as the required behavior has been specified by an adequate 

 statistical test, a simple logical process can convert the test into a de- 

 scription of at least one possible mechanism that will meet it. The 

 chief difficulty lies not in meeting the test, but in knowing enough to 

 frame it. It is only here that philosophical issues arise ; these will not 

 be discussed in this article. 



The brain is unique in offering two kinds of clues as to its function : 

 through physical experimentation and through psychological study. 

 The language of information and control belongs to both fields, and 

 a theoretical model is best designed in these terms. The first step in 

 its construction is to summarize the conditions that it must meet ac- 

 cording to present knowledge; these are regrettably broad. Next, a 

 model meeting these conditions must be invented, and its theoretical 

 characteristics checked against observation of the brain in action. 

 Any discrepancies should yield information toward refining or replac- 

 ing the model, and so the cycle continues. 



Progress was at first represented more by elimination of invalid 

 models than by the development of a valid one ; but the lines on which 

 a valid model may be sought are becoming gradually clearer. It is 



* This article Is based on a discussion by the author before Sections .T, A, and 

 I of the British Association for the Advancement of Science, September 8, 1953. 

 Copyrighted, 1954, by the Society of the Sigma Xi and reprinted from the Amer- 

 ican Scientist, vol. 42, No. 2, April 1954, by permission of the copyright owner. 

 326511—58 16 231 



