Suddenly, the series of decisions she'd made to coming to Gray's island seemed flawed. What was she doing there — what was Gray doing there? She wanted to know, to learn, to sate her natural curiosity, especially after the small glimpse she'd got of his island. But Laura's rising sense of foreboding evoked in some saner corner of her mind the desire instead to go home, return to the comfortable confines of her prior life.
"I have built a sixth-generation computer," Gray said, standing beside Laura at the top of the steps. The tone of his voice indicated a degree of importance that was lost on Laura. "It's a fully functional, massively parallel neural network." She looked up at him.
He was gauging the effects his words had on her. "The world's first true neurocomputer."
And those words did have an effect. "Are you serious?" He remained silent, his face devoid of expression. "I mean, pardon me, but… but a truly high-speed, highly distributed neurocomputer is a technological impossibility. It was a red herring. They tried it in the fifties, and it didn't work."
A smile — that self-satisfied look from before — spread across his face. "That was before fuzzy logic, genetic algorithms, and me."
Laura didn't know how to respond. On its face, his statement was rife with egotism. But he didn't seem egotistical. Then again, she didn't want to believe him to be an egotist. She had come here to help Joseph Gray, the misunderstood orphan, not Joseph Gray, the egomaniac.
She knew she should be on guard against projecting — against imagining in this strange and intriguing man those traits she desired to find. She was left not knowing what to think about Gray.
But she had not been brought there for judgments about Gray's mental condition. He'd brought her there for something else — for something else. Laura looked down into the dark, foreboding hollows at the bottom of the steps. She didn't want to go downtown to meet what the creator had encased in a heavy tomb of concrete.
Obviously sensing her hesitation, Gray turned from the steps.
"Let's take a walk," he said, touching Laura's elbow lightly and leading her away. They strolled alongside the roadbed that led across the great lawn to the assembly building. The enormous structure at the other end of the road formed a man-made wall stretching from one side of the field to the other. A gentle breeze rid the air of the noxious fumes. Laura breathed deeply of the smell of the ocean, slowly feeling cleansed of her earlier apprehension.
"We didn't know what we'd end up with," Gray continued as they strolled, "but we knew that we needed extraordinary computational power. You see, we wanted automation, but the main limitation on robotics was not the hardware of the robots themselves, but the horsepower of the computers that operated them. Take just the vision system. We use retinal chips in the robotics instead of cameras. They're particularly good at adjusting for both light and shade — which is a big problem — and in tracking objects across their field of view. But even with the advantage that retinal chips gives you, the processing demands are tremendous. Simply discerning edges — identifying the visual cues that tell where the boundaries of objects are — could take minutes for even the world's fastest supercomputers. Add hearing, touch, kinesthesia, problem solving, goal constraints — everything else that's required to actually function in the real world — and the processing needs of the system vastly exceeded capacities."
His voice had a certain quality to it, commanding attention without raising its volume. It was clear, deeper in tone than the voices of ordinary men of his build and age. And his thoughts welled up in such a fluid fashion that they seemed rehearsed. But he spoke naturally. He seemed at ease.
"Digital computers are very precise, which is good for tasks that involve numbers," he said. Laura listened more out of enjoyment of the experience than of the content. "But the best digital computer we've been able to produce can't surpass the information-processing capabilities of a three-year-old, and they never will. Digital computers are a dead end."
Gray lapsed into silence and walked on with his hands thrust deep into his pockets.
"You sound so sure," Laura finally said.
He stopped and turned to face her. "This" — he held out his arms, but looked straight at her—"is not a digital world; it's purely analog. The brain that we humans possess evolved, therefore, to process analog information — not to crunch numbers. If your brain were a digital computer, you'd store a description of all the faces of the people you knew in a long list. Every time you saw your mother, you'd have to search that list of faces for a match before you recognized her. As you got older, and your list of faces grew, it would take you longer and longer to find that match and recognize your mother's face. But your brain's not digital. You don't have a list of faces, You have a network of highly interconnected neurons. When you see your mother's face, those connections in your brain recall not only her identity, but everything about her — instantly! Your love for her. Her warmth to you as a child. The smell of her perfume. The fact that her birthday is next week. Everything, all at once. And a neural network can retrieve all those things from just a sliver of data — the sight of your mother's face or the unmistakable tone of her voice as she calls out your name. Unlike digital computers, it excels at handling fragmentary data—'fuzzy' data."
Laura's gaze shot up to Gray on his use of the word "fuzzy."
Gray had been working on a neural network to solve "fuzzy" problems way back in the 80s.
Could all this be the result, two decades later? she wondered, recalling the milestones of Gray's biography. His prediction of market demand for PVCs in 1984—his first billion. His prediction of the great stock market crash of 1987—tens of billions in wealth. His cornering of the high-definition TV market. The neural network he'd been developing was intended to discern patterns — to aid in market analysis.
Laura was suddenly so keyed up — so absorbed in the intriguing possibilities — she hadn't noticed that Gray had fallen silent.
Their eyes met briefly, causing Laura's skin to tingle eerily. It was irrational, she realized, but in that momentary look she felt as if she'd been touched. Not physically, but mentally — as if he had peered behind the curtain to share in her innermost thoughts. As if in his look he'd said, "I know what you're thinking." It was unnerving, and it scattered those thoughts to the wind.
When he continued, he spoke quietly, softly — as if not to jar Laura from her reverie. "Something special happened. Something magical."
Without knowing why, Laura had to muster her energy to ask, "What?" It was energy she used to brace herself for his answer.
"It learns," he said simply. Her first reaction was disappointment. She had prepared herself for something more, and her expectations weren't met. That initial reaction, however, changed with every word he spoke. "It generalizes, Laura. But when it generalizes, it does so by analogy not to the human world, but to the new world in which it exists. We missed it! When the computer first concluded that a wire is to electricity like a vein is to blood, we rolled our eyes and moved on. Its intelligence began to transcend our own. And that's because, you see, it was freed of this… shell," Gray said, looking down at his body, "and with it was freed of the limitations of our perspective — our experiences in life."
Laura was afraid to look at him now. Afraid that her face might betray some private thought or feeling. Afraid that his might betray something she didn't want to see. Her mouth was dry when she asked, "What exactly are you saying?"