The 1960 Symposium On Self-organization: an Overview
The period between 1950 and 1970 was a time of significant development in the field of computing. The diversity of artificial intelligence research that grew from this period laid the foundation for technologies that we continue using today. At the 1960 Symposium on Self-organization (the book indicates 1961, but this appears to be in error that is corrected in the text; the wiki as well as Hancock's article indicated 1960), with many researchers from the field that had been dubbed, machine learning, explored topics of self-organizing systems. The research topics largely developed from research in cybernetics and neural networks that had seen significant development during the 1940s. A list of names of presenters includes Ross Ashby, Warren McCulloch, Frank Rosenblatt, and Jack Cowan, among others. The work of researchers at this conference is differentiated from those engaged in programs committed to symbolic logic in their emphasis on self-organization. As Don Lavoie reflects in his discussion of programming dominant into the 1980s, the logic supporting programs was stated explicitly in model foundations. Research in artificial intelligence left the programmer responsible for all aspects of the program.
Machine learning researchers dealt with what is now commonly called subsymbolic systems. Neural nets are such a system. Neurons respond to and represent inputs from the environment. Conveniently, McCulloch and Pitts had modeled nervous systems as deterministic finite automata. Thus, the neural net was a natural model for the self-organizing system. Its states were determined by its past (initial conditions) and inputs from the environment.
It is not a great leap in logic moving from self-ordering neurons to self-ordering systems of autonomous agents. Ross Ashby, in his presentation at the symposium, outlines exactly this mode of thinking. Agents in a complex system are finite state machines. A system, itself, then can be modeled as a machine. This is along the lines of what I have recently suggested with regard to Hayek's work. And, yes, Hayek was at this conference. He was supposed to present a paper, but he indicated that illness prevented his completing a paper for presentation. Again, one must take care to distinguish between the use of a deterministic model to understand the logic of a system and the ontological claim that a system is deterministic. These are not equivalent. Even if they were, let us remember that 1) any model of a complex system observed in reality will likely be missing elements that impact its function and that 2) complex systems may only be understandable from a position of methodological dualism.
Once you realize that these studies in self-organizing systems can be broadly interpretted as studies of self-organizing finite state machines (whether that machine is an individual agent or a system of agents), the content from and significance of research from the conference become more accessible. With this in mind, I am providing outputs that aggregate concerns from the conference and texts presented at the conference. I also have provided lightly edited summaries generated from Llama 3.3.
Particular terms of note include system, machine, computer, automata, organization, state, neural, and network. These are all terms that reflect concerns of machine learning researchers and their approach to model-building. In this light, I have not only created representative plots that indicate which words appear the most, but also plots that compare the occurrence of these words within and between texts.