The immune system as a prototype of autonomous decentralized systems

Abstract

Some of the salient biology is reviewed, so that it can be seen how suitable the immune system is as a prototype of "bottom up" artificial intelligence. It is stressed that the immune system is entirely distributed (to first approximation). Its agents, the cells, are highly complex. Much is known about these cells and their interaction with each other and with pathogens, and the system is of high biological and medical interest. Major questions to be addressed include the following: 1) what are the "goals" of the immune system, and how can feedback promote these goals? 2) how can spatial organization allow non-specific chemical signals to select specific immune elements that contribute effectively to system goals? 3) how and in what sense can immune system performance be improved in the absence of an overall goal? and 4) how does the immune system compare with other autonomous decentralized systems.