Hacking at Relevance: Agile Development, Consulting and Training

Wednesday, February 4

I cleaned up my wiki some, adding papers I'd written in the course of my master's graduate work. These include papers on Web Services for an architecture course, testing parallel and concurrent systems for a course on testing and verification, object querying in Prevelance for the database course (for this course, I have notes for all the lectures posted) and a couple of papers for an AI course. For one of those papers, I attempted to give an answer to the question, "What is AI?" and highlight what an AI solution would be. That seemed worthy of quoting here:

I share Dennett's vision that AI is best viewed as "being a most general, most abstract asking of the top-down question: how is knowledge possible?" AI takes the question on by engineering a system that can solve - or attempt to solve - a specific problem in a given domain and then identifying those aspects of the system which would be required in any such system. In other words, the specific solution is mined for the generic attributes any intelligence would be expected to have. Repeated application of this tactic on varied problems enhances our understanding of the answer to the question.

Given this - somewhat vague - definition of artificial intelligence, an ultimate AI solution would be the design of a system that is not constrained to a specific domain nor problem. One that, like us, exhibits reasonable success in solving many different problems.

No comments: