Ricky shook his head. "I'm not sure. But they did this before, too."
"It almost looks like they're eating it."
"I know," Ricky said.
Of course that was absurd. PREDPREY was just a biological analogy. As I watched the pulsing cloud, it occurred to me that this behavior might actually represent a program hang. I couldn't remember exactly what rules we had written for individual units after the goal was attained. Real predators, of course, would eat their prey, but there was no analogous behavior for these micro-robots. So perhaps the cloud was just swirling in confusion. If so, it should start moving again soon.
Usually, when a distributed-intelligence program stalled, it was a temporary phenomenon. Sooner or later, random environmental influences would cause enough units to act that they induced all the others to act, too. Then the program would start up again. The units would resume goal seeking.
This behavior was roughly what you saw in a lecture hall, after the lecture was over. The audience milled around for a while, stretching, talking to people close to them, or greeting friends, collecting coats and belongings. Only a few people left at once, and the main crowd ignored them. But after a certain percentage of the audience had gone, the remaining people would stop milling and begin to leave quickly. It was a kind of focus change. If I was right, then I should see something similar in the behavior of the cloud. The swirls should lose their coordinated appearance; there should be ragged wisps of particles rising into the air. Only then would the main cloud move.
I glanced at the timeclock in the corner of the monitor. "How long has it been now?"
"About two minutes."
That wasn't particularly long for a stall, I thought. At one point when we were writing PREDPREY, we used the computer to simulate coordinated agent behavior. We always restarted after a hang, but finally we decided to wait and see if the program was really permanently stalled. We found that the program might hang for as long as twelve hours before suddenly kicking off, and coming back to life again. In fact, that behavior interested the neuroscientists because"They're starting," Ricky said.
And they were. The swarms were beginning to rise up from the dead rabbit. I saw at once that my theory was wrong. There was no raggedness, no rising wisps. The three clouds rose up together, smoothly. The behavior seemed entirely nonrandom and controlled. The clouds swirled separately for a moment, then merged into one. Sunlight flashed on shimmering silver. The rabbit lay motionless on its side.
And then the swarm moved swiftly away, whooshing off into the desert. It shrank toward the horizon. In moments, it was gone.
Ricky was watching me. "What do you think?"
"You've got a breakaway robotic nanoswarm. That some idiot made self-powered and self-sustaining."
"You think we can get it back?"
"No," I said. "From what I've seen, there's not a chance in hell."
Ricky sighed, and shook his head.
"But you can certainly get rid of it," I said. "You can kill it."
"We can?"
"Absolutely."
"Really?" His face brightened.
"Absolutely." And I meant it. I was convinced that Ricky was overstating the problem he faced. He hadn't thought it through. He hadn't done all he could do. I was confident that I could destroy the runaway swarm quickly. I expected that I'd be done with the whole business by dawn tomorrow-at the very latest. That was how little I understood my adversary.
DAY 6
10:11 A.M.
In retrospect, I was right about one thing: it was vitally important to know how the rabbit had died. Of course I know the reason now. I also know why the rabbit was attacked. But that first day at the laboratory, I didn't have the faintest notion of what had happened. And I could never have guessed the truth.
None of us could have, at that point.
Not even Ricky.
Not even Julia.
It was ten minutes after the swarms had gone and we were all standing in the storage room. The whole group had gathered there, tense and anxious. They watched me as I clipped a radio transmitter to my belt, and pulled a headset over my head. The headset included a video camera, mounted by my left ear. It took a while to get the video transmitter working right. Ricky said, "You're really going out there?"
"I am," I said. "I want to know what happened to that rabbit." I turned to the others. "Who's coming with me?"
Nobody moved. Bobby Lembeck stared at the floor, hands in his pockets. David Brooks blinked rapidly, and looked away. Ricky was inspecting his fingernails. I caught Rosie Castro's eye. She shook her head. "No fucking way, Jack."
"Why not, Rosie?"
"You saw it yourself. They're hunting."
"Are they?"
"Sure as hell looked like it."
"Rosie," I said, "I trained you better than this. How can the swarms be hunting?"
"We all saw it." She stuck her chin out stubbornly. "All three of the swarms, hunting, coordinated."
"But how?" I said.
Now she frowned, looking confused. "What are you asking? There's no mystery. The agents can communicate. They can each generate an electrical signal."
"Right," I said. "How big a signal?"
"Well…" She shrugged.
"How big, Rosie? It can't be much, the agent is only a hundredth of the thickness of a human hair. Can't be generating much of a signal, right?"
"True…"
"And electromagnetic radiation decays according to the square of the radius, right?" Every school kid learned that fact in high school physics. As you moved away from the electromagnetic source, the strength faded fast-very fast.
And what that meant was the individual agents could only communicate with their immediate neighbors, with agents very close to them. Not to other swarms twenty or thirty yards away. Rosie's frown deepened. The whole group was frowning now, looking at each other uneasily.
David Brooks coughed. "Then what did we see, Jack?"
"You saw an illusion," I said firmly. "You saw three swarms acting independently, and you thought they were coordinated. But they're not. And I'm pretty certain that other things you believe about these swarms aren't true, either." …
There was a lot I didn't understand about the swarms-and a lot I didn't believe. I didn't believe, for example, that the swarms were reproducing. I thought Ricky and the others must be pretty unnerved even to imagine it. After all, the fifty pounds of material they'd exhausted into the environment could easily account for the three swarms I had seen-and dozens more besides. (I was guessing that each swarm consisted of three pounds of nanoparticles. That was roughly the weight of a large bee swarm.) As for the fact that these swarms showed purposeful behavior, that was not in the least troubling; it was the intended result of low-level programming. And I didn't believe the swarms were coordinated. It simply wasn't possible, because the fields were too weak. Nor did I believe the swarms had the adaptive powers that Ricky attributed to them. I'd seen too many demos of robots carrying out some task-like cooperating to push a box around the room-which was interpreted by observers as intelligent behavior, when in fact the robots were stupid, minimally programmed, and cooperating by accident. A lot of behavior looked smarter than it was. (As Charley Davenport used to say, "Ricky should thank God for that.") And finally, I didn't really believe that the swarms were dangerous. I didn't think that a three-pound cloud of nanoparticles could represent much of a threat to anything, not even a rabbit. I wasn't at all sure it had been killed. I seemed to recall that rabbits were nervous creatures, prone to die of fright. Or the pursuing particles might have swarmed in through the nose and mouth, blocking the respiratory passages and choking the animal to death. If so, the death was accidental, not purposeful. Accidental death made more sense to me. In short, I thought that Ricky and the others had consistently misinterpreted what they saw. They'd spooked themselves.