ARTIFICIAL INTELLIGENCE: dogmas and metaphors
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(1) |
| ID | ◀ 156 ▶ |
| Object type | Epistemology, ontology or semantics |
As stated by S. GOONATILAKE: “Broadly speaking, it could be said that there are two main aims of people engaged in AI research. The first is building 'smart' machines or what could be called 'technological AI'; the other aim of AI researchers is to build models of cognitive processes, thereby helping one to understand the mechanisms of the mind” (1991, p.108)
Both groups brought into the subject their specific preconceptions (or even prejudices)
P.R. MEDINA MARTINS and L. ROCHA consider that the following dogmas have, until quite recently, “framed not only the kingdom of artificial systems but also — more deeply — their designer's minds:
- the dogma of the logical mind
- the dogma of the disembodied mind
- the dogma of the timeless mind
- the dogma of the adult observer
- the dogma of objectivity
- the dogma of the “in” and the “out” (1992, p.683)
These dogma are rooted — according to the authors — into two main interrelated metaphors, i.e. “the brain as a serial computer” and “the mind like a computer”
The implicit use of this conceptual frame led to reduccionist views on human brain and intelligence, which created a number of senseless problems in psychology and also brought artificial intelligence to a conceptual dead end (If and when one admits that A.I.'s only aim is or should be to mimic natural intelligence, which is not obvious). Gigantic, but rigid basically sequential algorithms can obviously not replace nor reproduce the simultaneous and parallel workings of neural networks, notwithstanding the fact that sequential computers and some expert systems obtain specific results which are out of reach of natural intelligence.
As noted by G.J. DALENOORT: “… artificial intelligence went on for some twenty years in constructing programs that could do something that on the outside was similar to what humans could do, without bothering at all how this capacity could have emerged in the human system” (1987, p.14)
As meanings are always “human” meanings , any symbol used in a classical A.I. program has necessarily been human constructed, and this also true for the program used, which is not endowed with anything more than rigid binary logic , sequently applied in a rigorously predefined way. Binary computers have thus no autonomouslearning capacity and, consequently, are no satisfactory metaphors for natural intelligence.