Theorem: General Intelligence Entails Creativity, Assuming - Computational Creativity Research: Towards Creative Machines - page 49

Chapter 3
Theorem: General Intelligence Entails
Creativity, Assuming ...

Selmer Bringsjord

3.1 Apparent Evidence Against the Claim

There appears to be considerable evidence against the claim that general intelligence
entails creativity. For example, with this claim unpacked as the proposition that
any general-intelligent agent must be creative, the field of AI declares the claim
to be false. To see this, we need but note that the dominant and encyclopedic AI
textbook [ 16 ] defines an intelligent agent as one that computes a mapping from its
percepts to its actions. The range of potential mappings explored in this volume
is non-trivial, ranging from simple arithmetic functions to functions that leverage
declarative knowledge, and beyond. But never is creativity discussed in connection
with any of these functions; indeed creativity is nowhere discussed in the topic,
period; nor for that matter is any synonym (e.g., 'innovative') discussed. In short,
as far as this highly influential and comprehensive volume is concerned, general-
intelligent agents needn't be creative.
Of course, AIMA , as it's known, is a textbook, at the end of the day; a masterful
one, yes, but certainly a textbook. It's on the bookshelf of nearly every single AI
researcher and engineer on our planet, but the tome doesn't purport to provide a
novel account of general machine intelligence. Yet it seems to me that we observe
the same lack-of-entailment result if we examine “research-grade” proposals for what
abstract machine intelligence is. One example is Hutter's [ 12 ] theory of “universal
artificial intelligence.” Whatever virtues this theory may have (and I do think it
has some significant ones), an explanation of creativity isn't one of them. Hutter's
formal foundations are avowedly and indeed proudly in sequential decision theory
and algorithmic information theory; but such things, if the scientific literature on
creativity is any guide, would be top candidates for being in tension with creativity.
Part of the reason for this is presumably that if we know anything about creativity in
the human case, and from that knowsomething about the abstract concept of creativity
that can cover information-processing machines and extraterrestrial lifeforms, we
know that creativity leverages declarative knowledge to produce new concepts, from