Things To Ponder: Artificial Intelligence And Computer Programming

*Artificial Intelligence And Computer Programming*
**By Roger Gudobba**

***Last time I talked about artificial intelligence having an impact on creative computer programming in the early 1960s. Let’s explore that further. In those first years the majority of computer systems were devoted to scientific number crunching, performing accounting functions or some other specific mathematical task. I was attending college and was fortunate to get a part-time job with a psychiatrist analyzing data for a research project. He was using a Bendix (I bet you thought they made brakes) G-15 computer. He was very frustrated with the time and effort required to get results. I had absolutely no idea what to do but looked at this as a great opportunity.

****What does this have to do with artificial intelligence and computer programming?

****One of the first programs I developed was used to guide the psychiatric residents’ thought process in determining the correct diagnosis for a patient. I presented questions and then modifying the sequence of questions based on user responses. I didn’t realize it at the time, but such an application might be considered a “rules-based decision engine” today.

****The program was modeled after ELIZA, a computer program and an early example (by modern standards) of primitive natural language processing. ELIZA was written at MIT by Joseph Weizenbaum sometime between 1964 and 1966.

****ELIZA has almost no intelligence and instead mimics the observable signs of human intelligence with tricks like string substitution and canned responses based on keywords. Nevertheless, when the original ELIZA first appeared in the 60s, some people actually mistook her for human.

****Simply getting a text sequence into a computer is not the same as imparting on it an understanding of natural language, and programming it to parse sentences is not the same as giving it the power to converse. The computer must be provided with a precise understanding of the domain to which the text relates, and this is possible for very limited domains.

****I was working at the Lafayette Clinic, a research and training facility connected to the Wayne State University School of Medicine and the Michigan Department of Mental Health. The clinic was recognized as one of the top three research facilities in the world for the study of schizophrenia. That reputation was gained in part by the fact that we had the largest and most complete collection of published articles on schizophrenia. The psychiatric residents read and cataloged articles and identified keywords. Researchers from around the world would request information based on different criteria and the computer program would read a tape file and list the article number that matched the criteria. We would then copy the articles and distribute them to the researcher making the information request. I developed this application based on a product from IBM called KWIC (Keyword in context). It sounds a little primitive based on what you can do today with online search engines, but remember that every antiquated process was once an innovative step forward.

****I was very passionate about the work and felt we were making strides in identifying the cause, effect, and potential treatments for mental illness. We were making a difference.

****During my 18 years there, however, the environment slowly changed. Technology was rapidly evolving and the public sector struggled to keep pace. Research grants became fewer and scientists fought to get funding. The state hospitals in Michigan were downsizing or closing, in part from cost cutting, in part from the desire to decentralize mental health care. Along the way, public policy shifted to allow psychiatric patients greater rights and responsibilities in determining their own course of treatment. I don’t disagree with that initiative, but the pendulum swung too far in the other direction. Now too many people who need help are on the street. I was discouraged and left to start a computer consulting firm specializing in working with small businesses. Eventually, this led me to the mortgage industry.

****Based on my experiences I came to ask: can computer programming ever replace the human intelligence factor? What do I mean? For example, when you think of a bird you immediately picture the bird flying. But then you think of the Penguin, which is a bird that doesn’t fly. Throughout my career I realized that computer programming was a tool to solve problems. But the key concept as I have always seen it was to understand the problem you were trying to solve before you attempted to develop the solution.

****Next time I will address some specific mortgage solutions.