The Samurai Strategy Part 15

"Ah, well, for now my work is here." He gestured uncomfortably about the room. "Let me try once more for that tea."

She realized he'd slipped deftly around her quick probe

concerning MITI's new role in the lab. He knew how to be a team player, she thought. Very j.a.panese.

This time he raised a response. A female voice dripping with long-vowel honorifics announced his tea would be delivered immediately.

Next came a small, awkward lull as they both sat there remembering Kyoto and not sure how to get around that memory. She wondered if it was happening all over again.

Maybe it hadn't been just a fluke, a crazy one-night diversion. She was about to switch to j.a.panese, thinking that might provide the jolt needed to break the ice, but just as the silence swelled between them, there came a knock on the door and tea.

She was half annoyed, half relieved.

He rose to walk over and began chatting as though they'd been interrupted in the midst of some intense technical exchange.

"Are you scheduled to present a paper at that Kyoto conference coming up?" He pushed a b.u.t.ton beside the door, and it swung wide. "There's sure to be quite a crowd. Everybody here's excited about supercomputers these days."

"No, this is strictly a pleasure trip. With maybe a few interviews thrown in to make it a tax write-off for a book I'm planning on robots." She hesitated. "Though I actually might go down and try to see a few people."

"Then this could turn out to be a pleasant coincidence." He took the tea, and the bowing girl vanished. Again the door clicked shut. "I have to go too, but I'm hoping to steal a few moments of freedom."

"You're chairing a session this year?"

"Absolutely not." He turned serious. "I'm not allowed time for anything like that anymore, Tam. This new project is top priority." He poured her a cup of the pale green liquid and pa.s.sed it over, seemingly relieved that the tension had abated. "There's a lot of work here at the Fifth Generation lab now that we're coordinating this program with the supercomputer effort."

"You mean with MlTl's supercomputer project?" Caught your little slip, she told herself. You are still with MITI. Which means they have taken over this lab.

He didn't blink. "As you probably know, MITI has the goal of creating a machine capable of a hundred billion computations a second, targeted just down the road. Which means we have to come up with entirely new computer languages and architecture."

"Parallel processing."

"Exactly. Handling multiple streams of information at once. Now that we finally understand what's required for a superfast computer, this work in AI just happens to be very relevant. It turns out we humans are already walking around with parallel processing in our heads, able to handle words, images, ideas, all at the same time. So if we want to create machines that operate as fast as possible, then it's crucial to understand how our brain manages things like recognition, learning, inference. Our hope is that by utilizing the studies here in those areas, incorporating them into our supercomputer work, we might be able to put ourselves a major step ahead. . . ."

Good G.o.d, Tam thought, it's elegantly simple. That's why MITI has taken over the Fifth Generation Project. They're going to use this research in artificial intelligence to come up with a computer more powerful than anything the world has yet imagined. Their silicon monsters are about to start replicating themselves, getting smarter as they go, like in some bad fifties horror flick. The difference is, this isn't make- believe.

"So you're here on behalf of MITI."

He paused. "For coordination. As I said, MITI needs the Fifth Generation work to be accelerated." He still hadn't exactly answered the question. "As part of our supercomputer effort."

Tam knew that Hitachi and NEC were both already claiming they had the world's most powerful machines, faster even than Cray's entry, the best American computer. What did MITI want?

He continued. "With 16-megabit chip production already going strong and 64 megabit commercialization in the wings, it did seem the right time to pull all our work together. If you think about it, computer speed and computer intelligence go hand in hand. I'll show you in a second what I mean."

Not kilobit. Megabit. MITI was going for the kill. This was a crash program. Why?

"Does this mean you plan to increase your funding for the Fifth Generation effort?"

"Whatever it takes to do the job," he replied after a moment's hesitation. "I suppose there's never enough money, is there?"

"Ken, why the rush? This sudden drive?"

"It depends on whom you ask." He leaned back and looked at the ceiling.

"Some call it survival, Tamara. Maybe it is that simple. j.a.pan is at a crossroads; we're rapidly losing our edge in the cost of labor. The only possible way to counter that is to step up our use of smart machines."

"Well, it looks as if I came to the right place. I'd like to add your name to my interview list."

His look darkened a moment. "Strictly off the record." Then he smiled.

"And only if we can do it over dinner."

"That sounds like a bribe."

"Call it an offering from an old admirer." He smiled, attempting to ease the tension. "The most I can do, for now at least, is just give you a small peek at a few of our experimental gadgets. Details are strictly proprietary. At the moment we're concentrating on computer vision and voice access. And on that last, by the way, I think we've just about reached AI's Holy Grail, natural language comprehension."

"Good luck." That was one of the mythical dreams of AI research, a computer that could understand the speech of anybody who happened along. Even though millions had been invested in the U.S. n.o.body was anywhere close yet.

"I think we're getting there. Enough so in fact that we're starting to look at applications. Expect commercialization in, oh, say a year, two at most."

Look out IBM, she found herself thinking.

"I probably shouldn't be showing you this, Tamara. So let's just keep this informal. No notes. But here, have a look at one of MITI's new toys. Can you guess what this is?" He pa.s.sed over a small device that had been sitting on his desk, his hand lingering on hers a moment longer than absolutely necessary.

She stared down at what appeared to be some kind of calculator-watch, except there was no watch face, merely a small speaker and two b.u.t.tons.

"That uses advanced versions of MITI's new 64-megabit memory chips.

There's nothing like it anywhere in the world. Without ever having heard the speaker's voice before, no calibration, it can translate ordinary spoken English into j.a.panese." He pointed to one of the b.u.t.tons. "Just press there

and talk. When you finish, push the other b.u.t.ton for the translation."

She did, testing it with the opening paragraph of Pride and Prejudice, her favorite novel. A simulated voice emerged from the small speaker on the face of the device and gave it back . . . in flawless j.a.panese.

"Not bad." She set it carefully onto the desk. The thing was actually almost frightening.

"Using this, linked to our new high-definition video and satellite, you could punch a b.u.t.ton in your living room and bring up people on a wall- size screen from anywhere in the world, then talk to them in your language and be understood in theirs. It's a quantum advance over current technology." He retrieved the device, dropping it into a desk drawer.

"I must admit I'm very impressed."

"Truthfully, so am I. Where's this program of MITI's taking us?" He looked up. "But let me show you something else, which I think is even more astonis.h.i.+ng. Of course you're aware that speech comprehension is easy compared to the really tough nut, duplicating the human eye. Since a visual image can contain billions of pieces of information, it can be very time-consuming for a computer to a.n.a.lyze all those at once and figure out what it's looking at. I've heard people at IBM claim that for a computer to recognize something even as simple as an odd-shaped coffee cup would still require almost an hour of processing, that to match the human eye and brain could take a computer the size of a building. But watch."

He walked over to a black metal installation attached to the wall and held up three fingers before its small lens. Then he pushed a b.u.t.ton and spoke into a built-in microphone.

"What do you see?"

She started to reply herself, then realized he was talking to the lens.

This time the answer took about ten seconds. Finally a voice in pa.s.sable simulation of the Tokyo dialect emerged from a gray speaker beneath the lens. "That is a human hand."

"How many fingers does this hand have?" he continued.

Again the eerie, disembodied voice. "The normal human hand has five fingers. This appears to have only three."