Are intelligence and evolution aspects of the same process?

By comparing computers with brains, we can easily see that adaptation and evolution are not the same. A system that adapts to changing circumstances by switching processes according to predetermined rules, as a computer program does, is not self-evolving. A self-evolving computer would have to do more than switch from one chunk of code to another. It would have to write its own software.

Because the views that inspired the term 'artificial intelligence' have generally been discredited, the term itself has also been deprecated. Nevertheless, the notion that machines can in some sense be intelligent still lingers. Perhaps the term 'intelligence' should be cast in a slightly different light to avoid confusion with any form of automatic processing of data, no matter how robust or nuanced.

The conventional view is that intelligence involves translating experience into representation and then manipulating the objects in that representation according to a set of logical rules.

A more useful description of intelligence might be that it is the momentary instantiation of an evolutionary process. This definition applies to instances of biological evolution that might seem superficially unlike human intelligence. In the opposite sense, it might be also true that brains create perceptions and new ideas through a process that resembles biological evolution but that takes place in a matter of milliseconds, as some scientists have speculated. Given this view of intelligence, the earth's ecosystem would be among the most intelligent things in the known universe because of the novelty it has created, including the human nervous system itself.

Taking the notion of a homeomorphism between intelligence and evolution in both directions, we would say that a system that does not evolve is an automaton and has no intelligence whatsoever, regardless of how complex that system may be. We should use intelligence to mean more than just flexible manipulation of data.

Are so-called genetic algorithms used by some computer programmers a simplified form of biological evolution, by virtue of the fact that they are Darwinian, or does biological evolution supersede the bare Darwinian algorithm in some essential ways? Does 'artificial life',  which is ultimately reducible to a set of rules, truly evolve in as deep and precise a sense as proteins have evolved? Are the selective processes that take place in 'artificial life' in some essential way unlike those that take place in the immune system of an animal?

G J Chaitin Home Page

Iain Stewart's home page

Michael Webb, 2000

home ]