Information Technology Reference
In-Depth Information
which new declarative knowledge is generated. The paradigm of this “creativity
engine” at work is the evolution of mathematics and mathematical knowledge. If
Leibniz hadn't used what he did know to create the concept of an infinitesimal, what
we know in knowing analysis via knowing the theorems that constitute it, might
never have arrived. So in Hutter's work we have a proposal for what the nature
of intelligence is, in the abstract—but nothing in that proposal yields that general
intelligence entails creativity.
The somewhat odd thing, though, is that Hutter [ 12 ] does mention creativity, and
indeed does so in a context that seems quite relevant to the present essay. For we
read:
The science of [AI] might be defined as the construction of intelligent systems and their
analysis. A natural definition of a system is anything that has an input and an output stream.
Intelligence is more complicated. It can have many faces like creativity , solving problems,
pattern recognition, classification, learning, induction, deduction, building analogies, opti-
mization, surviving in an environment, language processing, knowledge, and many more. A
formal definition incorporating every aspect of intelligence, however, seems difficult. Fur-
ther, intelligence is graded
So, the best we can expect to find is a partial or total order
relation on the set of systems, which orders them w.r.t. their degree of intelligence (like
intelligence tests do for human systems, but for a limited class of problems). Having this
order we are, of course, interested in large elements, i.e., highly intelligent systems ([ 12 ],
2-3; bold text from me).
...
The contrast between this and AIMA is quite interesting. AIMA defines an agent,
as we've noted, as an input-output device, with inputs as percepts and outputs as
actions. So what Hutter says about systems fits the AIMA framework well. But
then the list of “faces” that he gives, and casts aside as infeasible targets for tar-
geted formalization, include many things that AIMA in fact provides computational
definitions of (save, as we've noted, for creativity!). I see no reason to despair of
formalizing all of these parts of human cognition, and for the life of me don't under-
stand why Hutter rules such a project out as too difficult. The problem that I see,
from the standpoint of truly general intelligence, abstracted away from us and our
machines to cognizers in general, is that many of these parts of human cognition
aren't necessarily part of highly intelligent cognizers in the abstract case. I take up
this problem below (Sect. 3.2.1 ), and suggest a solution.
It's also interesting to note that Hutter is to this point roughly in line with what I
shall propose, which is a hierarchy of intelligence (and one inspired by my psycho-
metric tendencies, which renders Hutter's comment about human intelligence tests
welcome)—but for reasons that remain utterly mysterious to me, he takes maximiza-
tion of some utility function to be the essence of intelligence, to which all the “faces”
he lists are supposed to be reducible. He writes: “Most, if not all, known facets of
intelligence can be formulated as goal driven or, more precisely, as maximizing some
utility function.” [ 12 , p.3]. But no proof or argument is offered in support of this credo.
Any concern regarding the absence of such a systematic case is perhaps magnified
by the fact that on some accounts of creativity, for instance on some interpretations
of what Boden [ 2 ] calls P-creativity , to be creative is to somehow produce something
Search WWH ::




Custom Search