Mae saw it, too. "New behavior ..."
"Yes," I said. "Their strategy isn't working, so they're trying something else."
"Not going to do shit for them," Charley said. "They can zigzag all they want, it won't open any doors."
Even so, I was fascinated to see this emergent behavior. The zigzags were becoming more exaggerated; the swarms were moving farther and farther away from the buildings. Their strategy was shifting progressively. It was evolving as we watched. "It's really amazing," I said. "Little fuckers," Charley said.
One of the swarms was now quite close to the rabbit carcass. It approached within a few yards, and swirled away again, heading back to the main building. A thought occurred to me. "How well do the swarms see?"
The headset clicked. It was Ricky. "They see fabulously," he said. "It's what they were made to do, after all. Eyesight's twenty-oh-five," he said. "Fantastic resolution. Better than any human." I said, "And how do they do the imaging?" Because they were just a series of individual particles. Like the rods and cones in the eye, central processing was required to form a picture from all the inputs. How was that processing accomplished?
Ricky coughed. "Uh ... not sure."
Charley said, "It showed up in later generations."
"You mean they evolved vision on their own?"
"Yeah."
"And we don't know how they do it ..."
"No. We just know they just do."
We watched as the swarm angled away from the wall, moved back near the rabbit, then returned to the wall once more. The other swarms were farther down the building, doing the same thing. Swirling out into the desert, then swirling back again. Over the headset, Ricky said, "Why do you ask?"
"Because."
"You think they'll find the rabbit?"
"I'm not worried about the rabbit," I said. "Anyway, it looks like they already missed it."
"Then what?"
"Uh-oh," Mae said.
"Shit," Charley said, and he gave a long sigh.
We were looking at the nearest swarm, the one that had just bypassed the rabbit. That swarm had moved out into the desert again, perhaps ten yards away from the rabbit. But instead of turning back in its usual pattern, it had paused in the desert. It didn't move, but the silvery column rose and fell.
"Why is it doing that?" I said. "That up and down thing?"
"Something to do with imaging? Focusing?"
"No," I said. "I mean, why did it stop?"
"Program stall?"
I shook my head. "I doubt it."
"Then what?"
"I think it sees something."
"Like what?" Charley said.
I was afraid I knew the answer. The swarm represented an extremely high-resolution camera combined with a distributed intelligence network. And one thing distributed networks did particularly well was detect patterns. That was why distributed network programs were used to recognize faces for security systems, or to assemble the shattered fragments of archaeological pottery. The networks could find patterns in data better than the human eye. "What patterns?" Charley said, when I told him. "There's nothing out there to detect except sand and cactus thorns."
Mae said, "And footprints."
"What? You mean our footprints? From us walking over here? Shit, Mae, the sand's been blowing for the last fifteen minutes. There's no footprints left to find." We watched the swarm hang there, rising and falling like it was breathing. The cloud had turned mostly black now, with just an occasional glint of silver. It had remained at the same spot for ten or fifteen seconds, pulsing up and down. The other swarms were continuing in their zigzag course, but this one stayed where it was.
Charley bit his lip. "You really think it sees something?"
"I don't know," I said. "Maybe."
Suddenly, the swarm rose up, and began to move again. But it wasn't coming toward us. Instead, it moved on a diagonal over the desert, heading back toward the door in the power building. When it came to the door, it stopped, and swirled in place. "What the hell?" Charley said.
I knew what it was. So did Mae. "It just tracked us," she said. "Backward." The swarm had followed the path we had originally taken from the door to the rabbit. The question was, what would it do next?
The next five minutes were tense. The swarm retraced its path, going back to the rabbit. It swirled around the rabbit for a while, moving in slow semicircles back and forth. Then once again it retraced the route back to the power station door. It stayed at the door for a while, then returned to the rabbit.
The swarm repeated this sequence three times. Meanwhile, the other swarms had continued their zigzagging around the building, and were now out of sight. The solitary swarm returned to the door, then headed back to the rabbit again.
"It's stuck in a loop," Charley said. "It just does the same thing over and over again."
"Lucky for us," I said. I was waiting to see if the swarm would modify its behavior. So far it hadn't. And if it had very little memory, then it might be like an Alzheimer's patient, unable to remember it had done all this before.
Now it was going around the rabbit, moving in semicircles.
"Definitely stuck in a loop," Charley said.
I waited.
I hadn't been able to review all the changes they'd made to PREDPREY, because the central module was missing. But the original program had a randomizing element built into it, to handle situations exactly like this. Whenever PREDPREY failed to attain its goal, and there were no specific environmental inputs to provoke new action, then its behavior was randomly modified. This was a well-known solution. For example, psychologists now believed a certain amount of random behavior was necessary for innovation. You couldn't be creative without striking out in new directions, and those directions were likely to be random-