“Will do.”
Corrigan went through to his own office and set the briefcase down on one corner of the broad sweep of curved, walnut-topped desk with its terminal, onyx pen-holders and neatly arranged piles of papers and reminders. The floor-to-ceiling windows formed a corner of the building, presenting fine views of the downtown Pittsburgh vista along the opposite shore on one side, and the meeting of the three rivers with the Ohio Valley beyond on the other.
The last few months had seen an intensification of the realscaping program for capturing every facet of visual imagery over the entire Pittsburgh area. Camera teams had been out every day, touring and recording all the streets, expressways, parks, and trails; from vehicles and on foot, from helicopters overflying the city, from boats out on the rivers. Back in Xylog the machines were running day and night, reducing and compiling the encoded scenes into crosslinked hierarchies of field definitions in the huge database that took up half a floor of high-density crystal-array recirculator memory cubicles. Hatcher had told Corrigan that they could reproduce any aspect of any scene out there, from any viewpoint, in any direction. Corrigan had studied the figures and experimented with some samples, and he believed it. The results of the similar but smaller-scale program that Himomatsu had carried out in Tokyo were now incorporated into the main Oz database, as was a part of the Inglewood area of California and a few other places, following experiments by SDC.
Having reviewed his priorities for the day and disposed of the calls, Corrigan went through the mail with Judy and gave her a list of follow-up actions for the morning. Then he went back out to the elevators and down past the Primary Operations Level, where the main banks of massively parallel processing lattices took up almost the entire floor, past the Interface Level with galleries of COSMOS coupling hardware for up to fifty real-world surrogates, past the Monitoring & Control Center, from where the whole operation was directed, and came out on the second floor. Finally he came to a door marked FINAL EVALUATION & TEST, which was where Tom Hatcher’s group ran newly completed system modules prior to operational integration.
Hatcher’s concessions to the new order of things amounted to switching to regular pants in place of jeans, acquiring a jacket, and, on special occasions, adding a necktie. But underneath, the old, easygoing casualness remained unaffected, and he was still more at home sprawled in front of a terminal with his coffee in a Styrofoam cup than listening to investment plans being expounded over pâté de foie gras. When Corrigan arrived, he was waiting with Charlie Wade, one of the old crew from Blawnox, and Des Jorrecks, the head of Xylog’s applied psychology department. There were two broad areas to discuss:
First, results of tests to evaluate different strategies for creating animations that would best emulate people. Like people, the animations would shape their lives and personalities by pursuing goals. The intention was that these goals would arise internally, according to the animations’ individual natures and experiences, rather than be imposed from without. But real people rarely formed distinct goals that they pursued consciously and deliberately all the time, such as to become a doctor, lawyer, physicist, or actor, or to head a country or win an Olympic gold medal; for the most part, they simply lived their day-to-day existences following unconscious drives and desires, and the bigger things just “happened.” How, then, should such a nature best be simulated? What mix of drives, fears, ambitions, aversions was needed, with what kinds of relative weightings? How should such factors be represented as a statistical distribution across a whole population? Opinions on these questions changed constantly, and the short answer was that nobody really knew. A lot would be learned when the first runs were done in full-system mode, with the animation and environmental modules finally on-line and interacting together.
The other thing on the agenda was a subject that it seemed could never be laid to rest: the question of suppressing the surrogates’ memories when they began the full-system tests. Those in favor argued that it would ensure greater authenticity of behavior. Those against, who included Corrigan, maintained that they were scientists running an experiment, and scientists needed to know what was going on. “All we have to do is play role models to a bunch of dumb machines. We’re not trying to impress a panel of Shakespearean critics,” Corrigan said after they had been through the technical arguments yet again. “And on top of all that, it will make it a more exciting experience for everyone: the thought of launching off into the unknown—a bit like going up on a space flight to another planet, or something.”