This is a short paper, running about 6 pages of actual text. The paper deals with collections of autonomous communicating machines trying to cooperate in order to produce intelligent behavior. The main point is that such systems, which will become more and more important, present some problems that do not arise in more familiar centralized systems, and that some of these problems are difficult. Of course, one might reasonably take the view that it’s not the system that is intelligent, but rather the components. As the paper points out early on, the field of management seems like the place to look for some of the solutions. Part of the paper is devoted to a discussion of the new problems, e.g., inevitable inconsistency and lack of coordination. A fair amount of space is devoted to discussion of inadequacies in traditional models of computation, knowledge, and intelligent behavior. (I find room for argument here.) There is also some space devoted to discussion of planner and logic programming, which may be interesting to some people, but which I thought was rather beside the point.