The Intellects were "massively parallel" computers, computers made up of thousands of smaller computers, all running more or less independently of one another — but manipulating different parts of the same huge data base, that intertwined list of memories Lawrence called the GAT. Within Intellect 24, the largest Intellect, nine-tenths of the circuitry was dedicated to communication between processors. The processors themselves, the Intellect's real brains, were only a small part of the huge machine. Intellect 24 contained six million independent processors. Intellect 39, the portable unit, had nearly a million. And Lawrence knew, as Taylor had only guessed, that most of those processors were doing well to achieve a fifteen percent duty cycle. They spent most of their time waiting for communication channels to become available so they could talk to other processors.
ChipTec had found a loophole in the laws of quantum mechanics that allowed them to send a signal, not through space, but around space. From point A to point B without crossing the distance between the two points. Faster than light. Faster than anything. Instantly.
ChipTec had hoped to open up the stars for mankind (and reap a tidy profit on the deal, Lawrence thought silently). But their effect only worked at distances up to a few miles. It was only really efficient at centimeter distances. What could you do with such a thing? You could build a computer. The fastest computers were limited by the time signals took to cross their circuit boards; this was why supercomputers had been shrinking physically even as their performance grew and grew. It was why Intellect 39, with its million processors and huge switching network, was portable.
"We think you could realize an order of magnitude performance gain with very little effort," Taylor was saying.
"Two orders, if what you've said is true."
"It would be quite an achievement for ChipTec if our technology allowed you to realize your ambition and create a fully capable analogue of the human mind. We would, of course, own the hardware, but we know your reservations about the source code and are prepared to accept them."
Lawrence's eyes flashed. "That's a little unprecedented, isn't it?"
Taylor smiled. "If you succeed, we won't need the source code. Why start from scratch when a finished product is waiting to be duplicated?"
"There are some," Lawrence said darkly, "who aren't happy with the direction the code has taken."
"ChipTec is happy to have any marketable product, Dr. Lawrence. If anybody else wants to be that picky, let them find their own computer genius."
Lawrence's mind was racing, racing. Within each tiny processor in the massive Intellect were special functions of his own design, functions that could be reduced to hardware and done very efficiently with this new technology. Had he said two orders of magnitude? Try three. Or four. He could do full-video pattern recognition. Voice analysis. Multiple worldview pattern mapping. Separate filter mapping and reintegration. These were things he had tried in the lab, in the surreal world of artificially slowed time, that he knew would work. Now he would have the hardware to do them for real in a functioning prototype.
If he had been less excited, he might have wondered about that word "marketable." But the possibilities were so great that he didn't have time to notice.
"When do we begin?" he finally said.
The building had once been a warehouse for silicon billets, before ChipTec had switched to a ship-on-demand method of procurement. Lawrence wasn't vain and he was in a hurry to get started; the metal building would be more than adequate for his purposes.
With his move from the university and this quantum leap in technology, it didn't seem appropriate to continue numbering his computers. What would be Intellect 41 was going to resemble its predecessors about as much as a jumbo jet resembled the Wright Brothers' first plane. It would be the first of a new series of Intellects, the first, Lawrence hoped, to have a truly human level of intelligence.
It would be the Prime Intellect.
The label stuck, and the sign which ChipTec hung on the side of the building within the next month said:
PRIME INTELLECT COMPLEX
The speed of things made Lawrence feel a little dizzy. At the university he had had to make grant applications, oversee procurement, hand-assemble components, and do testing as well as designing hardware and code. Now he had the resources of a major corporation at his disposal, and if he suggested a change to the chipset at 8:00 A.M. he was likely to have the first prototype on his desk the next morning. Talented engineers took even his most vague suggestions and realized them in hardware before he could even be sure they were final.
A crew assembled modules in the warehouse, starting with the power supplies and empty card racks. The amazing thing was that none of this seemed to interfere with ChipTec's main work of churning out CPU's for personal computers. ChipTec had recently built a new plant to manufacture its latest high-technology product. The older plant dedicated to Lawrence's project was technically obsolete, even though it was only a few years old.
The chips being made for Lawrence's project were eerie for their lack of pins. Each tiny logic unit, barely a centimeter across, contained nearly a billion switching elements and yet had only three electrical connections to the outside world; they resembled nothing so much as the very earliest transistors. Unlike most computer parts, they communicated with each other through the "Correlation Effect" rather than through wires. This made Prime Intellect's circuit boards alarmingly simple; the only connections were for power. Even a transistor radio would have appeared more complex.
There were five major revisions before Lawrence declared the design final. Then production stepped up; at its peak, ChipTec was churning out forty thousand tested processors per day. Lawrence's goal was to give Prime Intellect ten million of them, a goal which would take most of a year to fulfill. Since each processor was over ten thousand times faster than a human nerve cell, Prime Intellect would be blessed with a comfortable information processing advantage over any human being who had ever lived.
Long before the goal was reached Lawrence was using the processors that had already been installed; he used them to test and educate his video recognition programs, to integrate experiential records from all his previous Intellect computers, and to perfect some ideas that had been beyond even his slow-time experiments to test. He did not, however, run the full Intellect program in the incomplete assembly. For one thing, it wasn't necessary; Prime Intellect wasn't just «a» program, but a constellation of over four thousand programs, some of which would be running simultaneously in thousands of processors. Each was more than capable of doing its job without the full cooperation of the entire organism, just as a nerve cell can function in Petri dish as long as it is supplied with nutrients.
And there was a kind of superstitious sense of expectation surrounding that final goal which Lawrence didn't want to blow by starting Prime Intellect prematurely. The project was written up in the popular science press, and Lawrence hosted emissaries from TV shows and magazines. Toward the end, there was nothing to do but watch the circuit card banks fill and listen to the growing hum of the power supplies. It was just as well, because Lawrence found himself becoming a bit of a celebrity.
Finally, after eleven months and four days, Lawrence sat at an ordinary looking console and typed a few commands. Four TV cameras and twenty journalists watched over his shoulder. Lawrence had a pretty good idea what would happen, but with self-aware computers you could never be completely sure, any more than you could with an animal. That was part of the magic of this particular moment in time. So Lawrence was as tense as everyone else while the final code compilation took place.