by Jay Johansen | Feb 10, 2009
Since the earliest days of computers, people have been discussing the idea of "artificial intelligence", the idea that a computer can be programmed to mimic human intelligence.
I readily agree that it's a fascinating question. But I routinely find the claims of "artificial intelligence researchers" to be exagerrated.
For example, a few years ago I read an article about a group of AI researchers that began by boasting that their program, called "BACON1", had "rediscovered Kepler's Law!"
I couldn't find that article, but here's a similar one about the same program:
As an example of a program that can discover things, Simon cited a computer program called Bacon (for Sir Francis, of course). He calls it "an inductive machine.' He went on to explain: "You give Bacon data--raw data--and Bacon's task is to find the scientific laws that are hidden in the data. We have tested Bacon primarily on historical scientific discoveries. We said to Bacon, "well, if you think you're so smart, let's see what you can do by taking the data that Kepler had about distances of planets from the sun and their periods around the sun.' Bacon discovered Kepler's Third Law in about 59 seconds. Computer Intelligence: Unlimited and UntappedSee also, Discovery of Scientific Laws. This article points out that Kepler took 20 years of research to discover this law.
Sounds impressive, doesn't it? Kepler's discovery of his Third Law (along with his First and Second laws) led to him being hailed as one of the most brilliant astronmers in history. A computer, fed the same data that Kepler had but not told his conclusions, rediscovered this same law. And the computer did it in only 59 seconds while Kepler took 20 years! Surely this means the computer is even smarter than Kepler, and Kepler is widely hailed as a genius! Artificial Intelligence is a reality!
Well, not so fast. As I read the details of what the computer had accomplished, it quickly became less impressive.
Kepler's Third Law states that, if D is the distance of a planet from the sun, and P is its period, that is, the length of time it takes to travel around the sun, then
Does this mean that the computer was as smart as Kepler? I think not.
What made Kepler a genius was not that he was able to study two columns of numbers and figure out the formula that relates them. I suspect that many bright college or even high school algebra students, if given these two columns of numbers, could also discover the formula. It's not that complicated. Are they all geniuses on a level with Kepler? No.
What made Kepler a genius was not that he found the formula relating these two columns of numbers. His first and greatest stroke of genius was to consider that there might be mathematical relationships describing the motions of the planets. His second stroke of genius was to consider that there might be a formula relating the distance from the sun and the period, and nothing else. Even today, centuries after Kepler published his discoveries, when we know a great deal more about the planets than Kepler did, I wonder what response most people would give to a question like, "How could you calculate the amount of time it takes a planet to travel around the sun? What information would you need?" Well, I suppose most people would have difficulty even understanding the question. Many would probably suppose that there is no way to find out short of taking the measurement. They would assume that it depends on how fast the planet is travelling, without ever realizing that the speed might be directly related to the distance from the sun.
Oh, and by the way, Kepler had inaccurate numbers for the orbit of Mercury, so his formula didn't quite work in one case. Nevertheless, he was insightful enough to realize that, as his formula worked for the other planets and was close for Mercury, the discrepancy was likely due to errors in observation and not the theory. That was a tough judgement call: When can you legitimately dismiss that sort of discrepancy as experimental error, and at one point does "proving" your theory by ignoring any data that contradicts it become an exercise in self-delusion or outright scientific fraud?
When someone writes a program where you can tell the computer simply, "Find me some formulas describing the motions of the planets", and the computer decides for itself what data to ask for, I'll be more impressed. If it comes up with formulas that were previously unknown so we know you didn't build in your knowledge of astronomy or physics, I'll be more impressed still.
No one today says that Kepler was a genius because he was able to solve an algebra problem. We call him a genius because he considered the idea that there was an algebra problem to solve, constructed the problem, and made insightful conclusions about the results. Solving the algrebra problem was a trivial exercise. The computer was given a problem where all the parts that required intelligence and creativity and already been done, and all that was left was some arithmetic and algebra. To say that the computer had "rediscovered Kepler's Law" because it could solve the remaining, easiest part of the problem, is like saying that anyone who buys a plane ticket to the United States has "discovered America" and is thus an explorer on a par with Columbus.
© 2009 by Jay Johansen