Thinking Machines, Racing Brains, and conversations with ALICE
I've just finished going over this lesson concerning AI. Now, all joking aside about computers taking over our world, and watching a computer and a brain race away, get awarded trophies and frown over losing...Bah, who am I kidding? I laughed at that animation until milk came out of my nose! It was awesome!
Now, since I am not a scientician (sic) in any real way (does watching star trek count?), when I first came across a lesson on AI, I thought: WOOHOOO! Now I'm going to find out how long it will be before I can buy a computer program that will write a publishable paper for me! Much to my chagrin, it turns out that my dreams do not stand a chance of becoming reality. Boo-urns.
What struck me as particularly interesting about the lesson (aside from the responses by the ALICE program to my very inappropriate questions) was the notion of "expert systems". The potential for new diagnostic technology was particularly intriguing to me. At the same time, the complexity required for such a system to be actualized is daunting. The lesson made it clear to me just how much of our cognitive processes are in fact the sum of several thousand at once. The instances of language and visual recognition were very illustrative. In order for a computer to become an expert diagnostician, it would need to be able to analyze huge amounts of data, reconciled by various tasks. However, at the same time it seems that there are certain factors in play in a clinical encounter that the computer could never really get around. For instance: after testing positive for a form of leukemia, would a computer prescribe a high-toxicity chemotherapy to a patient that is already severely weakened by the course of the disease? What factors influence a physician's decision in such a matter, and can they all be reformatted into a useable computer program?
I suppose such a situation could be handled by a logic tree of some sort, but it strikes me that the clinical encounter requires too many implicit value judgements for a program to be able to handle.
At the same time, it strikes me that historical analysis may be quite far beyond the current theories of AI. After all, where does one's inspiration come from when we choose a particular topic to investigate? I know that this is an old "art" argument, but I am surprised that nothing like it what covered in the articles. It is possible to apply fractal algorithms to Pollock's art, but how does one even begin a program to create such a piece? I guess that I'll have to keep looking for an answer...
Then again, it may be the case that I am just a neo-luddite.