When I studied artificial intelligence (AI) in the mid-1980s, an argument was raging about whether or not it could ever be solved. That argument continues to occur, but it’s intriguing to understand why. It’s not that we have made no progress, it’s about the real, underlying definition of what we think of as AI. If both academics and practitioners were honest, they’d admit that the definition of AI is “Getting machines to do the rest of what we call human intelligence that we haven’t yet figured out how to do with machines.” Notice that it’s a negative definition, AI is what we still don’t know. Therefore, until we have AI it’s always unsolved.
Vision, robotics and other full-fledged disciplines were considered AI in the early days. When we understood the problems well enough to solve chunk or at least to investigate them as specific units of study, they became their own area. What was left was still AI, the “unsolvable problem.”
I bring this up because my area of interest in AI was expert systems. I had the honor of working on a project under Bruce Buchannan, one of the creators of Dendral and MYCIN (two of the earliest expert systems). It was a business application of expert systems, using rules to get a computer to plan and budget. Working to get the professor’s brain onto the white board is was what made me realize I wanted to be on the business side of computing rather than the computing side of business.
Expert systems as a whole were oversold, and earned a bad reputation. So, just as rap became hip hop, other terms were used. Today we talk about “rule based systems,” “knowledge systems” and other similar techniques to help analyze Big Data. Companies talk about “intelligent agents” for customer support and prospect advice. There are a plethora of terms to describe what are, essentially, expert systems.
Why does that matter? While we might be able to solve new problems in business, adding significant value to software, most of what we do is evolutionary rather than revolutionary. That’s a good thing, as IT and most of the mass market want to know that they can add new capabilities without having to spend time, money and mental anguish over “transforming your organization!”
When looking at the new techniques for better understanding data, for predictive analysis and for other areas of business applications, know that knowledge systems have a long and strong history, regardless of a founder’s or a marketing organization’s addiction to a revolutionary message. Spend time to see if the vendor sees past his revolutionary message to the evolutionary solution needed by most firms.