to a desk, two chairs, one bookshelf, and one table. There was no way that he could cope with so little storage; a bricks-and-boards bookshelf bore two years of old journals, and an unhung door served as a second table. AJ opened the bar-sized refrigerator that doubled as one base of the improvised table; the other support was a dented two-drawer file cabinet topped with the 1988 Spokane Yellow Pages. The phone book was on permanent loan from the university library. There was little demand for it.
His guest shook his head when offered a soda. "Here is where you lost me. Where did you start?"
"Remember the simulated maze we saw?" Rewarded by a nod, AJ continued. "It's far simpler than the maze in any video game, but that's okay. I'll never admit it to my daughters, but even Quake, Doom, and Halo addicts are fairly high on the evolutionary ladder.
"So we begin with a maze. A supervisory program sees the headway made by other programs trying to navigate the maze. The maze runners are the programs we're evolving." Fred steepled his fingers thoughtfully. "And you start with maze runners made of random bits, and see which ones can solve the maze?"
"No, we've bootstrapped the process a little. To save time, we wrote the very most basic software. Think of it as stopping Creation a little early, with only some particularly stupid bacteria on hand." AJ sipped his Mountain Dew as Fred entered some notes.
"Got it." More scribbling with his stylus. "By selecting the fastest maze runners after each experiment, and randomly varying their software, you get ever ... oh, piss on it. . . better performance over time."
"I beg your pardon?"
Fred tilted the palmtop toward AJ; a red LED glowed bale- fully. "The batteries are almost drained, and I left my power converter at the hotel. Got any spare triple As?"
"Sorry." AJ shrugged. "Maybe a pencil?" He got only a puzzled expression in response.
"I guess I'd better wrap it up. How good are they now? The maze runners, I mean."
AJ set his soda can on a pile of ungraded quizzes. "Still dumber than dirt—but they learn a lot faster."
Linda shifted her weight from foot to foot, the nervous dance hidden by the bulky podium. Working in AJ's house felt odd, but it made sense. His in-home setup was preconfigured; configuring one of the school's communal distance- learning labs would have burned more of her time than giving the lecture.
Simpler still would have been for her to meet with the reporter and AJ to give his own damned lecture. Also scarier. She had no interest in meeting with the press.
The highlight of the lecture was a maze-runner demo. Annoying glints of light, reflections off a workstation screen, showed the video had been shot with a camcorder. Networking their lab directly to all the class sites would have been kinder to the eyes. It would also have exposed the ongoing experiment to every worm, virus, and Trojan loose on the Internet. Over my dead body, Linda thought. Only last week, a virus infestation had taken the university two days to eradicate. Damned eco-nuts.
On the classroom displays, a thousand specks crept about an elementary labyrinth. The dots turned and veered at random, bumping enthusiastically into imaginary walls. Most specks remained clustered near the beginning of the maze; only a handful had navigated around more than one corner. None was anywhere close to the exit. Students cheered when the front-running mote successfully negotiated the fourth turn, a 180-degree switchback.
Takagawa, on the main-campus display, was the first to get the point. "It's a constrained optimization problem ... ma'am."
Sigh. He didn't remember her name. She wondered if any of AJ's students did. Linda knew what the young man meant, but it was his job to express himself. " What is a constrained optimization problem?"
"Solving a maze, ma'am."
"Go on."
"Much problem solving involves optimizing some value, subject to a set of constraints." The student bounced excitedly in his
Ellie Dean
Glen Cook
Erin Knightley
Natalie Anderson
Zoey Dean
John Fusco
Olivia Luck
Ann Shorey
Thomas Ryan
Dawn Chandler