| |||
| |||
| |||
| |||
| |||
| |||
|
The first strand of CRCC work involves the attempt to build faithful computer models of some of the most central mechanisms and features of human thinking -- namely, high-level perception, analogical thought, discovery and creativity, and, most especially, the foundation that effectively underlies all these phenomena -- namely, what we term the "fluidity" of human concepts.
The second strand of CRCC work, by contrast, has little to do with building computer models (at least on the surface), but simply involves FARG members engaging in creative intellectual endeavors, either scientific or artistic, such as poetry translation, discovery in mathematics, the study of human error-making, the study of humor, the study of sexist language and imagery, the creation of various types of art, and so on.
The former strand has led to several Ph.D. theses, many research articles and conference presentations, and a few books, while the latter strand has led to several books, innumerable lectures in universities around the world, some art exhibits, a good number of small symposia at CRCC, and so forth.
It must be mentioned that careful monitoring by FARG members of the creative activities that constitute the second strand informs, in a profound and pervasive manner, the architecture of the computer models that constitute the first strand. In that fashion, the totality of the work at CRCC possesses a coherence that might at first not be apparent.
The cognitive modeling at CRCC is based on the thesis that mental activity consists of many tiny independent events and that the seeming unity of a human mind is merely a consequence of the regularity of the statistics of such large collections of events. Thus the metaphor of the "intelligent ant colony" and the image of "active concepts" (as set forth in the book "Gödel, Escher, Bach") have inspired our models for over two decades.
The models all involve the nondeterministic interaction of many tiny events that take place in simulated parallel. The models also feature both long-term and short-term memories, the former of which houses permanent concepts, and the latter of which is like a stage on which temporary mental structures are built, modified, and eventually razed. Events in each memory profoundly influence the multiple tiny parallel processes, and in that way, each memory affects what goes on in the other.
Our most advanced computer models so far have been the Copycat and Metacat programs (implemented by Melanie Mitchell and James Marshall), which make all sorts of analogies in a micro-world (e.g., "If ABBCCC changes to DDDEEF, then how would JJJII change?") and the Letter Spirit program (implemented by Gary McGraw and John Rehling), which designs the lowercase letters of the roman alphabet in artistically coherent new ways, either starting from seed letters provided by a human, or starting from scratch.
These models, along with several others, are described in our collective book Fluid Concepts and Creative Analogies (Basic Books, 1995), as well as in Melanie Mitchell's book Analogy-making as Perception (MIT/Bradford, 1993), Robert French's book The Subtlety of Sameness (MIT/Bradford, 1995), and the doctoral theses of James Marshall, Gary McGraw, and John Rehling.
Each of these models applies the parallel terraced scan architecture to a microdomain -- a small piece of reality which is easy to define, if not always easy to understand in its fullness. The letter analogy problem microdomain of Copycat and Metacat is probably the easiest to understand, although it is continually surprising in its depth and richness.