|Intelligent Systems And Their Societies||Walter Fritz|
For this project we have used a Macintosh Plus computer with 4 Mb RAM and a speed of 8 MHz. The compiled GL (General Learner) program uses 97 Kb of space, and the memory of response rules, concepts, and rule patterns varies in size according to the state of the GL's accumulated experience. For the experiments shown previously, the GL needed only a few hundred response rules. Working with 8000 response rules, together with their corresponding concepts, and rule patterns, should be possible. For this system, the speed of reaction is good; there is no noticeable interval between pressing RETURN and receiving an answer.
The GL Program
The GL (General Learner) program was written in Think's Lightspeed C language. The program has a continuous cycle of sense - select response - act - remember experience. When the GL program has not perceived an input for a given extended period of time (or on specific request), the program enters the "sleep" function. The entire program, including all library functions, window, menu, auxiliary, and debugging parts, consists of 220 functions.
In order to facilitate the quick retrieval of the particular response rules that are applicable to a given situation, the program includes in the structure of each concept the numbers of those response rules in which this concept exists in the stimulus part. These numbers are themselves related to the location of the response rule in memory. This method was chosen because it was found to be much faster than even a binary lookup of response rules. As expected, however, this gain in speed comes at the expense of larger storage space requirements.
For continuous reading, like a book - continue
Jump to the top of this document /General Learner /Artificial IS /e-book Contents.
Copyright © New Horizons Press