It’s a regular explosion! The term “artificial intelligence” and its tantalizing, film title acronym, “AI,” suddenly are everywhere. While far from an expert in computer science, machine learning, automata theory, neural networks, and the like, I am much more an expert than many swatting the concept around like a shuttlecock, and I am sufficiently vain to want my say-so, too.
Along with a small team I helped create one of the first business platforms to revolutionize the banking system for small banks around Chicago. I worked long nights and weekends to take the banking business out of human hands, computerize it, and make it more efficient and easier for both clients and employees. Previously, I had been among a larger group that computerized virtually all the work of the Midwest Stock Exchange. We didn’t imagine that our systems were intelligent; they just encapsulated the intelligence of the people’s creating them.
Circa 2000-2001, anticipating retirement, my wife and I made a series of sorties into quite a few states to find a place to enjoy that phase of life. One of these visits was to the vicinity of Choteau, Montana and an isolated house on the verge of the Bob Marshall Wilderness. While driving to the location, the realtor tantalized us with one of the home’s endowments, an electronic oven! “Whatever can that exotic appliance be?” we pondered. Imagine our disappointment to find a garden variety range with keypad and digital display. Strictly speaking, one doesn’t prove anything with anecdotes, but you can guess what this one is hinting.
What one needs to understand is that, because a computer chip is calling, or helping to call, the shots, does not qualify a system as “intelligent.” More likely than not it is merely following a set of rules with which some programmer has impregnated it. To be in the class of a human mind, a true AI, a system must be able, on the basis of experience, to enlarge its bases of data and knowledge, as well as to change the rules by which it makes decisions. The everyday computer program lacks this capability.
Two computer examples from the 1960s are instructive, and compare that date with what you know about the history of computing. One system was called “Baseball” and could answer any question of the sort, “In 1948 how many hits did Stan Musial get on Sundays?” or “How many home runs did Sherman Lollar hit against the Yankees in 1955?,” the kinds of information contemporary sports announcers babble to conceal their ignorance of the game’s finer points. Certainly superficially impressive, its success depended only upon a database of all box scores of all major league games. Searches of that could answer any sort of statistical question, but the system couldn’t be called “intelligent,” because, once it understood the question, its brains were simply algorithms for searching, an ordinary computer capability at the time.
I can’t recall the name of my second example, but it accomplished medical diagnoses. Punch in a collection of symptoms (from an admissible set – no free-lancing), and the computer churned out a list of possible causes. This system was adequately “intelligent” to do what a physician (hopefully) would do; i.e., ask for additional information when stumped or if necessary to shorten the list of possible ailments, or, in the best scenario, tell the patient exactly what’s ailing him or her. For example, the system might reply, “What’s the patient’s race?”, “What’s his blood pressure?”, or “Does she take penicillin?” and carry on a bit of a dialog with the interrogator before giving an answer. Have you heard of anything similar lately? What’s new under the sun? This was sixty years ago. Inn this case, observe the potential for a semblance of machine intelligence. If the program can be advise of its mistakes and of new medical findings that occur and endowed by programmers with the rules for incorporating such data and knowledge, it can emulate what a person does.
It seems all fields of endeavor, from medical, including prescriptions, to airline tickets to car rentals or grocery chains are giddy with their supposed AI developments, which allow corporations to cut their work forces and rely upon the intelligence of their computer chips. Is it working? Oh, yes, they are cutting their work forces, which increases profit, but does it help the consumer? Does having to do more of the work oneself and having the system now and then botch it royally constitute help?
How many have spent exasperating intervals pushing cellphone or computer keys to answer irrelevant questions created by so-called AI systems, chatting with corporations’ computer robots, and never getting a resolution to their problems? When the computer reluctantly grants permission to talk to a humanoid, how many have waited not seconds, not minutes, but tens of minutes before reaching one, who must then key in the same information to his or her higher-level, AI system, which has not been programmed to give an answer to the problem?
Does the voice-recognition system fail to recognize your voice? Must you fiddle looking for a magic code in a text message when you’re hoping for help on the cellphone or at a website? Have you argued with a postal clerk that’s trying to tell you your addressee doesn’t actually know his own zip code? Does an “intelligent” email or text system have a vocabulary worse than your own; does it insert nonsense you never intended; has it an inferior knowledge of correct grammar and syntax; will it NOT make the same mistakes time after time?
The last I knew, computer scientists and related experts could not agree upon a definition for AI, so that’s an obstacle in recognizing it and discussing whether System X exhibits it or not. The commercial world, however, is not deterred by annoying technicalities; call it AI, by golly, full speed ahead, and peddle it wherever possible.
Several paragraphs above, I may have put one over on you; that was the qualification, “understood the question.” Inputs to “Baseball” and the medical system could not have been any old sloppy expression of the request but had to be presented in a narrow, limited way, just like a line of a computer program. As far as I’m concerned understanding a natural language – French, Japanese, English, etc. – is the mark of machine intelligence, and by “understanding,” I don’t mean the ability to come back at you with a clever, but irrelevant, reply. A problem posed by a British writer in the 1970s illustrates the point.
He said, “Suppose a machine is told that Jack can venture into the woods alone and gather 3 pounds of raspberries in an hour and that Jill can go out alone and gather 2. When will a machine ever figure out that if they go into the woods together for an hour, they may bring back LESS than 5?”
If that day has arrived, I missed it, and I think it would be premature to be frightened by the notion of artificial intelligence or dizzy with anticipation of it. Lots of people are out to sell us things and make bountiful profits, and sometimes salesmanship, like that of the real estate agent, overshadows honesty. I’m already sick of having spelling, grammar, and simple responses botched by purported “intelligent” systems. It’s unfair that the user become the guinea pig in the AI frenzy.
Larry Stanfel
Roundup
Reader Comments(0)