Saturday, November 1, 2014

Pachman on the Cybernetic Machine

     In my 1963 edition of Pachman’s Modern Chess Strategy he wrote a brief section concerning chess and the ‘cybernetic machine.’   By cybernetic machine he meant computers.  He wrote:

“There has been talk in recent years of building a chess automation or more precisely using a modern cybernetic machine (i.e. an electronic apparatus capable of solving extremely difficult mathematical and logical problems) to play a game of chess or solve chess problems…It is a simple matter to teach the cybernetic machine the basic rules about the movement and capturing of pieces, check, mate, stalemate, pawn conversion, etc…But that is not enough; the machine must actually play. For this two methods can be used."

"With the first, the automation examines all possible variations and selects the best move by rejection. This method can be employed for solving problems in which the number of possible move is relatively small…the machine must examine an enormous number of variations and despite the extraordinary rapidity with which electrons work, the process takes a lot of time. It is quite clear that this method cannot be used for playing a game…If the machine had to calculate for only seven moves in advance in a position with thirty alternatives, it would need 10,000 years to select the right move!”

“The second method is to teach the machine the most important principles of strategy and tactics. A cybernetic machine can solve not only mathematical, but also logical operations-even extremely complicated ones. And, after all, the principles of strategy and tactics have a logical form. This attempt was also made in the Soviet Union and the machine was indeed able to play a game of chess: but it played very weakly and was beaten by an average player...how could this result be possible…the reason is that play at the chessboard crosses the limit of logic and enters the field of dialectics (the art of investigating or discussing the truth of opinions). This is a field that is outside the scope of even the most perfect machine: it is a field of activity reserved for the human brain and will always remain so.”

Pachman gave a mate in three problem that appeared in Chess in the USSR in 1956 and observed that a cybernetic machine solved it in 12 minutes but he was able to do it in one minute. Out of curiosity I decided to test it out with today’s engines.


     The solution is 1.e8B Kxf6 2.g8R Ke6 3.Rg6 mate. Using only a single core on my laptop the mate in three was found instantly by Houdini 2, Stockfish 5, Fritz 12, Critter 1.6, and Naum 4.2 while Komodo 8, Gull 3 x64 and Rybka 2.3.2a all found a mate in four instantly and after one minute were still showing only a mate in four. Eventually they all found the mate in three though.
     Pachman’s thoughts on the subject remind me of the time I found a box of Popular Science magazines with dates ranging from 1912 to 1920 at the flea market. One issue had a letter to the editor from a university physics professor explaining why space travel would never be possible. Will chess ever be solved? The usual answer is no, but don’t forget Pachman’s words in 1963 that to calculate for only seven moves in a position with thirty alternatives would take 10,000 years to find the right move.

No comments:

Post a Comment