Ecce Signum

Had a long, interesting conversation with Scott today, regarding the problems inherent in duplicating the creative process. He is building an application which, he hopes, will be able to write poetry based on an understanding of the concepts behind language. I cannot imagine a more difficult task than to teach a computer to ‘think’ in metaphors. There is so much we don’t understand about our own thought processes when it comes to recognition and cognition, that modeling such behavior can easily devolve into educated guesswork. Questions come up; hard questions, like: What does it mean to ‘perceive’? To ‘conceive’? To ‘recognize’?

Isaac Asimov once stated that any sufficiently advanced technology would be indistinguishable from magic. Borrowing this idea, could it be said that any sufficiently complex pattern of behavior would be indistinguishable from intelligence? Computers do not ACT. They await input, in whatever form it may be, and then do what they are told to do with that input. They do not autonomously decide what to do with unfamiliar data. They can search for patterns which match patterns of familiar data, but they will not search for patterns which we have not told them to search for. It goes back to my comment regarding Fluid Concepts and Creative Analogies : a computer will not try to creatively figure out a problem. It does not care that {1,4,9,16,25} is a series of perfect squares. It will spend eternity trying to untie the knot where Alexander would simply cut it with his sword.

The questions about intelligence which arise from this train of thought tend toward the unsettling. Is there such a thing as action, or is there only re-action? Is human behavior a reaction to a profoundly complex set of behaviors, or, in being self-aware, do we transcend re-action to the point that we behave autonomously?

Throwing in the question of free will vs. predestination complicates the process of teaching a computer to recognize poetry. But without teaching a computer to think symbolically, the best machine-written poem will, in reality, be the result of complex pattern-matching.

My project is, for the near future, much less complex than Scott’s. I am building a machine to model evolution and genetic drift. Ultimately I plan to explore the question of emergent behavior and hive-mind patterns. I say ‘less complex’ because the a-life I work with does not need to think; it only needs to re-act.