http://u.nu/6j7f4

A Battle That Revolutionized Computer: Traveler’s Tale

http://u.nu/6j7f4

To the enemy commander, the 96 huge ships of the Eurisko battle fleet, show and cumbersome and carrying only light arms, seemed an easy target. His own fleet consisted of only 20 small, highly maneuverable craft, but each boasted heavy guns and sophisticated laser weapons. The commander confidently gave the order to open fire.

http://u.nu/8m7f4

Four salvos later, more than half of the Eurisko ships had been sunk but their commander, Douglas B. Lent, was preparing to accept the enemy’s offer of surrender. Despite its huge losses, the lumbering Eurisko fleet had destroyed all but one of the opposition’s highly vaunted high-tech ships. The battle had ended.

Helping Itself to Success

http://u.nu/4p7f4

This confrontation of armed might took place on July 4, 1981, in San Mateo, California. The exercise was part of a futuristic war games tournament called Traveler. Lenat was a computer-science expert from Stanford University; Eurisko was his computer program. What made their joint victory so remarkable was that Eurisko itself had designed the battle fleet and developed its tactics.

Eurisko was a major innovation in the development of artificial intelligence. Before Lent invented Eurisko, “intelligent” programs could rewrite sentences or create new culinary dishes-if supplied with massive amounts of information. But they were incapable of discovering facts or creating new ideas of their own.

Lenat envisioned a program that could not only solve problems from information at hand but also test and improve its answers and the techniques it used to work out. Lenat wanted a program that could think for itself and discover entirely new information in the process.

The Traveler game demanded that each player calculate how to build a fleet of ships that would defeat all opponents. Lenat gave Eurisko enough details of its basic task to begin work, backed up by a huge memory bank of “heuristics”- rules about how to make discoveries, test concepts, and adapt them to from new ones.

Each time is designed a fleet, the program simulated a battle, examined the results, and adjusted its ships accordingly. Then it staged another battle and again made adjustments. Eurisko learned by experience. Lenat, its mentor, claimed to have created only 60 percent of the final program.

The solutions Eurisko devised were unorthodox. In the course of 10,000 simulated battles, it had discovered that a large fleet force of lightweight craft, no matter how elaborate their weapons or how fast they moved. The fleet that lasted the longest won the game. In case an enemy managed to sink all of Eurisko’s ships, the progress developed a fail-safe: a tiny, unarmed and agile “lifeboat” that was deemed unhittable.

Please Stay Away

Following Eurisko’s victory in the 1981 Traveler tournament, the game’s organizers changed the rules only a week before the 1982 event. Undaunted, Eurisko once again defeated the opposition. Politely, Lenat was requested not to participate the following year.

But he had proved that, with the right program, a computer, almost unaided, can make new discoveries. How had Lenat achieved this? He had simply explained to the program the principle that governed its operation and then let it “talk to itself” – a major step, perhaps, toward creating a machine that genuinely “thinks.”