Computer And The Lion
The Computer And The Lion

by don kelly- -Oct 1994



The University of Alaska has a very fine computer science department offering degree track coursework in artificial intelligence. In 1992 the computer science department contained a core of post-grad students and professors from other departments who wrote computer programs called, "artificial neural networks."

As a mid-term exam our professor challenged the class to write a database to validate or invalidate the AI program. We formed four teams, selected a subject and divided up the work, then tried to decide what to do next. None of us had individually written a complete database before. It was going to be a team effort.

The idea was to build a comprehensive database that would include everything we knew about lions, africa and men, taking care to include all possible solutions to all possible problems, (not possible). We were convinced we would run out of information to write about, (wrong). We wrote 10,000 lines of code in two weeks.

Our scenario played out in the wilds of Africa where a lion came face to face with an unarmed man on foot. What would the lion do when it saw the man? What would the man do when he saw the lion? We split up the code three ways and the fourth member of our team wrote the data flowchart against which to test our code.

The habitat part of the scenerio was fixed, the landscape, winding trails, bushes, trees, etc. Man and lion interacted in the context of that habitat. Upon meeting, the lion depending upon factors of hunger, fear and other prior experiences with humans, might stalk, immediately attack, observe, retreat, etc., or ignore the man.

The man on the other hand, depending upon actions taken by the lion, had numerous alternatives to protect himself from the lion. He could become invisible, (possible only to a computer), attack the lion, (not very likely), hide, (good only if the lion couldn's smell him out), run away, (not fast enough), climb a tree, (not bad).

On test day the professor loaded our many times tested database. We held our breaths while the computer compiled the code. The computer determined from the facts that the lion was friendly. The man saw the lion and climed a tree. The lion circled the tree, looked at the man, and sang, "I, am the king, of the forreeesssst."

We didn't expect that result. We thought it was a joke and cracked up. Our professor who rarely displayed a sense of humor, had tears streaming from his eyes. It seems that we had two slightly lazy jokers working on our team. They ran out of ideas to write about lions and ended their code with a story about the Wizard of OZ.

The moral of the story, if there indeed is a moral, is that a computer's decision making abilities, unlike that of humans, are totally dependent upon the information in their databases. Databases are as good as the humans who wrote them. Computers will one day write their own programs and make intelligent decisions, but not yet.