|Intelligent Systems And Their Societies||Walter Fritz|
Creating The Theory
This theory is the outcome of my having built several artificial intelligent systems (ISs) on a computer. This theory started out as a hypothesis, which we tried out by means of using a computer program to represent an artificial IS. We then observed which parts of our system were not functioning properly and attempted to modify the theory and thus the computer program. This we did many times. It was the modification or our program that has taught us most about our hypothesis.
Studying artificial ISs has the advantage over studying human ones that we can readily observe all their internal processes: we can observe the creation and use of concepts and of response rules. Thus, this theory talks about the creation of concepts, the elaboration of the present situation, the elaboration, storage and retrieval of response rules, the selection of an adequate response rule, and finally, the execution of the response part of the selected response rule.
We can observe that a surprising number of the brain functions of the human IS are quite similar to those of an artificial IS. From this, we would probably want to conclude that this is obvious, since most artificial ISs are modeled on the natural, the human one. But this is not quite true. When we review the artificial intelligence literature we can observe that a wide variety of approaches have been utilized in the functional design of artificial ISs. However, the author and others have noted that most of these other approaches do not work well, even if at first sight they seem to be quite reasonable. Further examination shows that many of those that have worked show these amazing similarities to how we currently believe the human mind functions.
We all know that human beings forget. Since humans generally consider forgetting to be unbeneficial or even detrimental, and as a computer never forgets (that is, unless someone trips over its power cord or otherwise pulls it out of its wall socket), we thought that our IS program could dispense with modeling the function of forgetting. We were wrong! We found out that without forgetting, so much unnecessary and unimportant information accumulated in the computer that it soon filled its memory and thus could no longer store any new information, whether useful or not. Also, because of the excessive buildup of unnecessary response rules, the time the program took to select an adequate response rule began to be unreasonably long. So we concluded that even a computer, when used as an IS, needs to forget.
Human beings sleep about a third of their lives. "What a waste!" we thought. Surely a computer does not need to sleep. Again, we were wrong! It turns out -- and of course we learned this the hard way -- that the intelligent system needs a great deal of post event processing time to review and begin to take full advantage of its experiences. A considerable amount of its generation of concepts and of its relating these concepts with each other can only be reasonably done after an experience is finished. This processing takes a lot of time because for each concept the IS has to review all of its memory, and not only the portion that relates to its last experience.
This becomes even more important when we fit the IS into a "real" human environment. For example, let's take the case of a robot, controlled by an IS, walking down a sidewalk just a few steps in front of a person. It could be more than inconvenient for the person if the robot were to occasionally stop suddenly and appear to meditate. Similarly, in an interactive exchange between a human and an artificial intelligent system, for instance the playing of a game, it would be inconvenient if the program stopped its external activity every few minutes.
Thus, we believe that an IS should only review its memory at a time when there are no external demands by the operator. We included a command "sleep" by which the operator can start this mode of internal activity and external inactivity. Also, when the computer is left alone for some time, it automatically enters this mode.
It seems that something similar is going on in human beings. We suppose that in the REM mode of sleeping humans review their experiences. After a full day of new experiences, we are confused and often irritable. But after a good night's sleep, we suddenly see everything much clearer. A certain amount of review and of ordering of information must have taken place. So, it seems, any intelligent system, needs to sleep.
(2003: Prof. Jan Born's researches showed that after an experience, the REM phase of sleeping helps the learning of movements and that deep sleep helps the learning of experiences related to words and emotions.)
For continuous reading, like a book - continue
Jump to the e-book Contents / Intelligent Systems document / top of the page.
Copyright © New Horizons Press