by Hugh M. Lewis

I propose the development of alternative models of intelligent systems that do not strive to meet the rigorous hard AI standards of the Turing Test, and that serve other functions than the mere anthropomorphization of machine behavior and the projective mimicry of our own behavior in the input-output feedback loop that defines a machine as a form of automaton. If we cast about the natural world for examples of such alternative intelligence, they are not very difficult to discover--we find them in ant-colonies, that can forge a trail in the most efficient manner between food sources and other resources and their underground chambers, even if no single ant by itself, or even a small group of ants, is capable of achieving such a solution by itself. I would recommend that any living system, or species, that has successfully worked out for itself an evolutionary solution in a given ecological niche represents from a form of alternative intelligence in its design and behavioral patterns. I would suggest furthermore that if we look at intelligence primarily as the complex manipulation of information, then even genetic transmission and recombination systems that result in adaptive evolution represent an alternative form of natural intelligence.

Given these considerations, I would proffer a definition of alternative intelligence as any system involving the manipulation, storage, change and updating of information in a semi-autonomous manner that serves to solve particular functional problem sets that are the intention of their designs, including but not limited to the performance of some kind of work, as a machine, or the acquisition and mediation of information for its own sake.

I therefore propose alternative standards and measures of machine or systems based intelligence that might include any subset of the following kinds of criteria:

1. non-anthropomorphic designs in form or function.

2. capacity for consistently solving certain complex problem sets in a reliable manner.

3. performance complex operations in an automatic and routine operational manner with minimal human control inputs, manipulation or mediation.

4. is capable of being integrated into larger systems frameworks that are themselves intelligently integrated upon another level of functioning.

5. is capable of some degree of generalization and modularization of form and function of information.

We should be willing to accept more relative standards of what constitutes an "intelligent system" or machine, and to understand that by this term we do not mean the same thing as a "thinking" machine or a machine that can actually think for itself in an independent manner. Perhaps eventually such a sophisticated device can be somehow engineered, but in the meantime we might content ourselves with more reasonable and realistic sets of standards and definitions regarding machine intelligence.

An intelligent system is, therefore, any complex design of information representation, that is capable of functioning in relatively sophisticated ways in the manipulation and handling of information, given the caveat that what might seem sophisticated from the point of view of an ant or a dog might not seem so from the standpoint of a dolphin or a human. Not being constrained by anthropomorphic or anthropocentric conventions, the stereotype of Robby the bipedal, walking/talking robot is perhaps something best kept in the 1950's. Intelligent systems can take virtually any form and design required of their function, without regard to such standards. Of greater interest are the integration of intelligent systems and subsystems into larger meta-systems framework, with the idea of the distribution of automated control functions and tasking to various components of a larger framework, and the capacity for multi-tasking across a range of problem solving areas.

General Systems Essays, Vol. I