The Shape of Actions: What Humans and Machines Can Do

Rating:
Summary: Machines only mimic behaviour, not human action.
Review: UNDERSTANDING TECHNOLOGY

Can technology ever duplicate (and perhaps replace) people? This is actually an ancient question which has sparked debate from many positions (yes, no, it depends, etc.). Both recent trends (i.e., modernity ever since Francis Bacon) AND its aesthetic imagination (i.e., modern science fiction) have given new life (and perhaps urgency) to this debate. So?!

Mechanization, automation, and cybernation have certainly replaced people in many repetitive tasks (a process Noah Kennedy aptly calls "The Industrialization of Intelligence"). But what about fully automated factories, or a computer like HAL, or androids, or whatever? Some with generally respected bona fides (Arthur C. Clarke, Raymond Kurzweil, etc.), have speculated that eventual duplication is possible. Others with equally impressive bona fides (John Searle, Hubert Dreyfus, etc.), have argued that duplication is essentially impossible.

If you have already decided an answer for yourself, such debates may not change your mind. But if you really are both interested and open to credible arguments, where does plausibility reside? Collins and Kusch have definitively answered the question (at least for the present). They have developed a theory of what distinguishes human action from machine behaviour, and have shown why there is an impenetrable barrier between the two.

Collins & Kusch's conclusion is basically a theory of contrivance, something which is both very rare and very important. To my knowledge, only Aristotle's Four Causes and McLuhan's Four Effects are in the same class of explanations. The Collins & Kusch typology is different from other classifications of technologies. It essentially distinguishes between contrivances which embody and imply social context (polimorphic), and those contrivances with do not involve such context in use (mimeomorphic).

Technology is limited to the mimeomorphic, and there are four types of possible gizmos:

Simple capability (embodying a single process); Branching capabilities (a limited number of alternatives); Feedback capabilities (built to adjust and regulate performance); Learning capabilities (incorporating the patterns of time-series summaries).

What this amounts to, is that there can be (and are) Expert Systems (knowledgeable technologies), because they are confined to a domain. BUT there can never be Artificial Intelligence (contextualized comparing across domains). And even Expert Systems (and the multitude of lesser gizmos) only function because the faults in their processes are continually corrected by human intervention. Few, if any, gizmos ever live up to their hype! For every other explanation, there is a RAT in the practice. Read the book to see what this means, and why.