Working Up

by Dwayne Phillips

Some of us have been here before. “AI” is hot again. Perhaps this time the future will be different from the past, but I doubt it.

I worked extensively in AI in the 1980s. What I see today is remarkably similar. This is due to what I believe is a gross misunderstanding of what AI is, where it has been, and probably where it is going.

At its most basic, AI is an area of research. Sometimes the research releases products into society and those products are no longer called AI, they are just something else we have and use. See, e.g., the thermostat. That is a machine that causes hot air to enter a cold room and cold air to enter a hot room. Whoa! That requires thought and judgement and all that stuff of AI.

In the 1980s, we did lots of AI research. One of the hot fields of the 1980s was the expert system. It was software that stored knowledge of a narrowly defined field and provided advice like “this has a high probability of being cancer” or “this is probably a problem with the carburetor.” Some of those systems went into general use, and no one thinks of them as AI.

Today we are doing lots of research in supervised learning in general and neural networks in particular. These techniques are decades old, but today we have the required computing, and that computing doesn’t cost much money. Hence, those techniques are “hot topics.”

Many of these supervised learning systems are being released into general use. Cars can park themselves between white lines. Maybe they don’t part so well if the lines are blue, but someone is working on that. Systems can recognize faces as long as we don’t have too many different persons come to our front door and the persons aren’t wearing funny clothes and the lighting is pretty good and a few more things.

Once these few more things are solved, we’ll have automatic door openers and we won’t consider it AI. Come to think of it, how about those door openers at stores and such that recognize when a person wants to enter, so they open the door. Once that was AI research, now its just something that opens the door when we approach it.

Research to everyday use. Some of us have seen it before, and some of us will see it again.