Automata & Automation

by Hugh M. Lewis

The problem of automata and the problem of automation are not the same, though they are generally confused on an implicit level and often the two terms are used interchangeably. In a sense, automata refer to the theory of automatons, or theoretical machines that are capable of producing a certain kind of modulated output, arbitrarily determined, depending or independent of the nature of the input into the system. Usually the output is determined by a set of rules, or a "discrimination" structure that somehow logically determines the final choices among a range of alternates. 

Automation refers to a machine that is capable of independently performing complex functions, and demonstrating some measure of self control, without direct manipulation or involvement by a human agent or human mediation. In general, automation tends to simplify and remove human involvement from the direct action or consequences of the behavior of the system or machine. With increasing automation, we expect that human involvement in a system becomes less and less directive, and possibly minimized to a point of merely turning such a system on or off or providing basic starting/stopping operations, etc. Of course, an automated system is not merely a "remote-controlled system" though in general it entails some degree of remote control. An engineer on earth piloting a vehicle on Mars surface is primarily an example of a process of extremely "remote" control. Systems automation entails that the system becomes increasingly self-controlling, and the control function of the human input is simplified to the minimal number of choices necessary to make the system fully operational.

The purpose of automation is not to displace workers from the labor force, but to free human activity and interest from the drudgery of performing routine-operational tasks, in order that they may spend their time doing more meaningful work and cultivating more productive life-styles.

Automation has in general proceeded slowly and incrementally one small bit and piece at a time--it is usually articulated in very narrow and tightly delimited contexts. Machines are best at performing one set of functions, over and over again, at very rapid rates. They are not at their best when they are multi-function by design. Conventional CPU's are single processing devices that perform one stream of continuous information processing in a linear but very rapid manner. Creating multi-or parallel processing devices, like the connection machine, associated primarily with supercomputing and the handling of very large informational demands at very high rates of speed, are a kind of solution to this problem, but perhaps we can consider as well non-linear processing functions that work in fundamentally different ways than straight forward parsing of strings of information. Analog and hybrid computing models have been developed that to some extent address these issues, but the problem still remains open, especially if we consider it in light of the challenge of creating human-like intelligence, or machines that can be automated across a fairly broad range of different tasks.

Of course, the two concepts, automata and automation, converge when we think of an automated system that is intelligent, like a fully functioning robot that is self-directive and independent in its behavior. Given time, we may perhaps achieve such models in a manner that would measure up to what we would expect of our current stereotypes of them. The general trend of the future is of course one toward greater convergence, but this convergence is very broad based and comprehensive in form, and proceeds in a relatively self-organized and piece-meal way. It will not come overnight, or even in a year. It will be measured by relative degrees of integrated distribution and achieved progress compared to measures of the past.

It would be expected to see increasingly intelligent automation in areas that involve the greatest and most intensive inputs of human labor, and this is in areas of greater and more reliable complex pattern recognition, manual manipulation of tools to make fine products of varying design, and in possibly genuine "auto-mobiles" that are self-guided and self-steering. Achievement of semi-intelligent automation in any one of these areas, much less all of them, would represent significant advancements in the development of humankind.

General Systems Essays, Vol. I