How Models of Creativity and Analogy Need to Answer the Tailorability Concern - Computational Creativity Research: Towards Creative Machines - page 101

they do not register as a match here, and as a result the concepts BaseballGame and
Tennis-TheGame are not linked. Nevertheless, were the database more complete (as
one would hope it will be eventually), such a psychometric problem would likely
have been trivial. 6
5.4 Conclusion and Future Work
The concerns reflected by TC have been more or less echoed in many criticisms
of analogical systems over the past 30years or so, but little to no discussion has
attempted to precisely state the conditions under which an analogical systemwould no
longer be subject to this common complaint. We humbly acknowledge the limitations
of our formulation here, and remind the reader that even a systemwith a strong claim
to meeting the conditions of TCA 4 (e.g., an implementation of Fig. 5.2 ) may not be
completely immune to TC-based criticisms. But it seems that such criticisms would
be somewhat difficult to justify. We hope that this paper inspires further discussion,
and invite researchers to pick up from and sharpen TCA 4 , or to steer away from
Psychometric AI and therefore develop TCA 3 in a different direction.
Those who strive for something deeper than a test-based anchor for a response to
TC will presumably be more sanguine than Bringsjord about the ultimate reach of
computing machines. Bringsjord holds that it is impossible for any mere computing
machine to originate anything [ 10 ], or even to genuinely understand anything [ 12 ].
But he holds on the other hand that for any determinate behavioral test, a computing
machine can be programmed to pass it—even if the test is one which, if passed,
suggests to most human observers that the passing machine is not simply the product
of human intelligence, harnessed [ 8 ]. In short, the turn to TCA 4 as a way of fleshing
out TCA 3 reflects deep pessimism about computing machines, and specifically a
deep affirmation of the penetrating and profound nature of TC. Licato, in contrast, is
more optimistic about the potential of analogical-hybrid systems, but is not prepared
to take any philosophical positions on whether a system that passes TC demonstrates
true understanding or creativity.
What about AI work devoted to building creative systems on the strength of ana-
logical reasoning? Our work continues along this path. We are currently investigating
the potential of using our analogical matching systemMETA-R, which is designed to
allow for a flexible application of heuristics across a wide variety of domains, with the
community-built dataset Freebase [ 7 ]. We are also investigating the use of META-R
in modeling the automatic generation of logical theorems and high-level proofs from
semi-formal and formal domains [ 25 , 32 ]. Future work in this direction includes the
investigation of Analogico-Deductive Moral Reasoning (ADMR) using a hybrid of
6 Of course, this discussion does not even mention the difficulty artificial systems have in finding
an overlap between symbols on the basis of the semantics of the referents of those symbols. Such
a problem is somewhat out of this chapter's scope, but is certainly something that researchers
interested in transcending the TC should eventually tackle.