Book Read Free

Brian D'Amato

Page 7

by In the Courts of the Sun


  The lab was on the UCF campus in a new hacienda-style nerd ghetto. You could still see the grid lines in the new St. Augustine grass sod. Even though it was the day after Christmas, everybody seemed to be working. There were private security primates all over. They kept talking to each other, and eventually to Taro, over those Bluetooth ear thingies that make people look like processed livestock. Well, here I am, crawling back, I thought. Was he still mad at me? Maybe just ask him. Hey, are you still mad at me? No, don’t. Don’t embarrass him. Or yourself. He probably figures you’ve seen the error of your ways. Maybe he’s right. I knew I’d decided Taro was just another mercenary, and I’d felt pretty disgusted, but now I didn’t remember why exactly I’d felt that way.

  Taro met me at the third door in. He didn’t so much look older, but I’d been remembering him as a jovial Hotei sort of character, and now he was more of a Hsun Tzu, drier and grave. Like all Japanese people he only looked half-Japanese. He still wore his old powder-blue Tokyo University lab coat.

  “It is nice to see you,” he said. He held my hand for a second. For him it was like licking my face. His hand was smooth, dry, brittle, and delicately ridged, like the shell of a paper nautilus.

  “It is nice to see you,” I said. He seemed like it was actually nice to see me. Well, he’s a guileless sort, I thought. If he said it was nice, it was nice. There was hug potential, but instead I shook his dry hand. Neither of us was a very demonstrative type. I’m not Latino in that way. I’m an Injun. Like, Chief Stone Face him no show any heap big emotions.

  “Thanks for having me over,” I mumbled. “You know, I feel kind of bad showing up after all this, whatever.”

  “No, not to worry,” he said. He didn’t have an accent—I mean, he had a bit of an Oxbridge accent but no Japanese one—but he spoke in that precise way that tells you an East Asian language is still lurking in there. “I understand that things sometimes are difficult.” Despite myself I got a floodlet of that warm ’n’ fuzzy feeling like you’re a scoop of ice cream and somebody’s pouring hot butterscotch over you. I hate it when that happens. Teacher/pupil has got to be one of the weirdest possible relationships. Well, maybe he’d guessed I might contact him as soon as I saw that Time article.

  “Let us check in on the patient,” he said.

  “Great,” I said. Come down to the lab, I thought. And see what’s on the suh-luh-AAAB!

  We went through another two sets of doors and into a keyed elevator. Brrr. Freezing in here. We went down three floors to the sub-subbasement. Taro’s cold room was at the end of a long hall. I got the sense the complex was mainly industrial R&D. There were doors with lab names like HAPTIC feedBACK and LOW-FRICTION MATERIALS. Taro held his hand over a scanner and a door hissed open.

  The room looked like it was about forty-one and a half feet square with an eighteen-foot ceiling, all done in your basic morgue white with bone accents and a hundred thousand lumens of shadowless fluorescent lighting. Its only remarkable feature was the computer in the center of the room: a clear Lucite tank about the size of a Ford Explorer van stood on its end. LEON, which they said stood for Learning Engine 1.9, was suspended inside the tank, a black thing like a big grandfather clock. Skeins and ratkings of cords and hoses curled out from the bottom of the tank and stretched across the white epoxied floor to a mound of chillers and Eheim pumps and Acer 6000 storage drives, all pushed against one of the windowless cinder-block walls. Four genderless perpetual graduate students huddled at workstations in the corners of the room, tapping and mumbling to themselves in HLASM.

  “We replaced most of the silicon chips with doped germanium,” Taro was saying. “But the thermal dissipation is still nearly three hundred watts. So for now we are refrigerating him like an antique Cray. The coolant is the same type of plasma they use for synthetic blood transfusions.”

  He led me over to the tank like I was a tourist at the Big Chunk of Rock National Monument. I squinted into it. Up close you could see that the black thing wasn’t solid, but rather a tall stack of paper-thin black circuit boards, each about three feet square and a quarter-inch apart. Whorls of heat distortion spewed out of different layers through the clear liquid like diffraction waves over a summer highway.

  “Huh. Nifty,” I said. Demonio, this really was a cold room. About sixty f-ing degrees. I’m going to need a damn baby blanket, I thought. Or like, two shots of Tres Ańos. Brrrrrrrrrrrrrrrrrrr.

  “Of course this is just the CPU,” he said. “The drives are in another building. And the storage is … well, I do not know where all the storage is. Much of it is in Korea.”

  “How fast is it?” I asked.

  “Right now he is close to six petaflops.”

  “Wow.” Sounds expensive, I thought.

  “At the moment LEON is running two hundred and fifty-six simulated worlds about ten minutes ahead of the real one. And for each of those he is playing through more than five million branches of the Sacrifice Game tree simultaneously. Each one is a three-stone game.”

  “How many simulated trades do you run?” I asked.

  “Around twenty thousand a day,” he said. “I don’t know about the actual trades.”

  “Huh,” I said. That’s one of the great things about Taro, I thought. Most other people would have gotten all cagey and said something like “Where’d you hear we were running any trades?” But he just didn’t have it in him.

  “Would you like to play a game against him?” Taro asked.

  I said I’d love to.

  “Have you played with three stones?”

  I said I had. As I think I said, this means that you use three runners, that is, the stone that represents what actually happens and that runs away from the hunting stones, which represent different potentials. The thing is, it wasn’t three times harder than playing with one stone. It was 33 times, that is, twenty-seven times harder. It’s sort of like how a mate-in-three chess problem is many, many times harder than a mate-in-two. So anyway, usually I used two stones. But I’d been working on playing with three. I figured I could handle it, against a machine, anyway. Really, computers still can’t play the Game for shit.

  Taro got a wobbly shop stool and I sat down in front of an old NEC 3-D monitor. He hoisted himself onto the Formica work surface and started tapping on a touchpad.

  “You know how the average human brain runs about two billion operations per second?” he said, over the tappitty-tap-tap-taps.

  “Well, it takes a lot of work to be average,” I said.

  “And then after that we should budget in another six or eight billion of our own operations just to compile out the parallelism.” I nodded, as though I could easily have worked that out myself. “Then we must double that for record keeping and fail-safe. And then we have got about twenty billion ops per second. So as long as it goes through at normal speed and we do not have to store anything in LEON himself, that should be just enough.”

  “Great,” I said. Enough for what? I wondered. To create a new master race of all-knowing nonorganic superbeings? Well, at least then I’ll have someone to talk to. Yep, in the final showdown between man and machine, I know which side I’ll be on—

  “But I do not think it will ever surpass a human player,” he said. “Even if LEON becomes as large, computationally, as a human brain—even if he becomes as smart as a human brain—this does not mean he will be as intuitive as a human brain.”

  The Sacrifice Game was like Go, and unlike chess, in that people could still play it much better than computers. A low-intermediate human player can still beat the world’s best Go program. And Go’s a very describable game, close to what programmers call a clean environment. The Sacrifice Game’s a lot more anecdotal, more connected to the world, so it’s at least a few million times messier.

  “Well, don’t sell yourself short,” I said, “at least, not to any grant committees—”

  “They guess that already,” he said. “That is why we have so … gone corporate. At any rate, at this stage LEON is primarily of value as an assistant.” He led me over to a bank of OLED monitors. “It helps improve the performa
nce of novice adders. Like advanced chess.” That is, what chess players call it when they play in consultation with two computers. I nodded.

  He sat. I sat.

  “We are working with five student players,” he said. “Two of them learned the Game in Maya communities, and the others have trained here. One of them is very promising. He was not an adder before, though.” I waited for him to say “Still, he cannot hold a candle to you, you ace,” but he didn’t. Instead he showed me some charts and pointed out where the spikes were on what, for want of an elegant term, we’d called the “worldwide event space.” Basically, it confirmed that the Game did best at guessing what groups of people will do in crisis situations. “This is, of course, still very useful,” Taro said, “and over time it could be immensely profitable.” But it wasn’t the sort of prediction his backers wanted. For instance, it didn’t do a good job of predicting the markets per se, but just what people will do in the markets. You’d think this would be the same thing, because markets depend on psychology. But in fact there are still all these nonhuman factors going on in market fluctuations, too, industrial lag, capital flow, weather, and on and on, and getting that stuff together with the psychology requires interpretation. It’s one of those things that is very hard, or maybe impossible, to teach a computer.

  So Taro was having roughly the same problems I was … but still, I thought … hmm. Suppose their simulated trades average, say, 0.02 percent over industry standard, then that’s still enough for a company that size to make a few million a minute. These days, even less of an edge could turn whoever had it into a market-devouring monster. The Warren Group could be on their way to being the richest company in the world. Although you’d think they’d be bigger already. Maybe they’re just spending a lot more than they’re reporting. Which might also explain why they’re being so secretive about the Game thing. They’d be bragging about their investment results all over the place unless there were some specific reason not to. People don’t dismiss game studies out of hand anymore. If anything, they’re all trying to get a piece of it. Everybody wants to hire the next Johnny von Neumann.

  Or maybe they don’t want to manage other people’s money, they want to grow their own. Maybe Lindsay Warren and some of the board members want to buy back the public shares before any word gets out. Or maybe they’re afraid that if the government finds out they have something militarily interesting, they’ll take them over. Maybe that’s even something to be a little uneasy about, isn’t it? Suppose Warren or somebody else does take the Game to the next level, then what? Maybe they’re going to end up owning everything and just take over the world? It’s like if Taro’d been running the Manhattan Project, except instead of working for the War Department he was sponsored by Marvel.

  Maybe I should just post whatever I know about the Game. Maybe even later today. I’d been thinking about doing it for a while and I had most of it written up. Then at least everybody’d have it. I’d kept putting it off because—well, a few reasons. I felt like I hadn’t figured it all out yet. It was still tough to learn and harder to master. Also, I had a few things I wanted to take care of with it myself before I attracted any attention. Well, frankly—I wasn’t going to mention this, but maybe I should level with you, now that we know each other a little better—the truth is I was saving up to sponsor a blind-contract hit on García-Torres. It’s not an easy thing to do these days, since the people you hire tend to turn you in even if they actually do the job. Still, I thought … but still, the other thing was, it wasn’t even clear that putting the Game out in the world would be the best thing. Maybe it would be like with nukes; it’s bad enough that some crook politicians have them, but it’s still better than giving one to every nut on the planet. Hmm, mmm, mmm …

  Except, the thing is, if Warren was trying to keep a lid on, why’d they let Taro talk to Time? If they couldn’t keep the Codex from getting published, because too many Mayanists knew about it, they might have asked him to say something lukewarm—

  “Would you like to see the current game board?” Taro asked.

  I said sure.

  “I should not be showing it to you because it is absolutely secret, but of course you helped develop it and I know we can trust you.”

  I said thanks. Damn, I thought. I’d been a real pisado. It was all choking me up a bit, actually.

  Taro clicked it up on the screen:

  Whoa, I thought. Simple. Elegant. Sometimes you look at something and it’s just obvious it’s right.

  Damn. Why hadn’t I thought of that?

  “Huh,” I said. “And, uh, this one’s based on the Codex, uh, Nymphenberg?”

  He said largely, yes.

  I spent about half an hour clicking around the board, trying different calendrical assignments and getting used to the interface. It wasn’t so hard to adapt to as I’d have thought. One tends to think of game boards as always being the same size, like 9 squares for tic-tac-toe or 64 squares for chess. But that’s not really true. Some people teach chess to newbies on a 36-square board. Shogi, or Japanese chess, has 81 squares. A standard Go board has 361 points, but even serious players sometimes play quick games on boards with only 81. Serious tic-tac-toe people play on larger or multidimensional boards. In feudal Japan, generals and courtiers used to play shogi on 625-square boards filled with all sorts of wacky pieces like blue dragons, evil wolves, and drunken elephants. And of course, in episodes 1-2, 1-3, 1-20, and 3-14, Kirk and Spock played chess on that trilevel board that you can now get a replica of from the Franklin Mint. So the Sacrifice Game is the same way; you can play on a larger or smaller board without changing the rules or even changing the strategy all that much. But it can take a long time to get really good at playing on a new scale. I could do nothing but kid around with this for ten years. Damn it, this is the right stuff, I thought. If I’d been using this version I’d have made billions by now, not just millions. Taro’s company must be picking trades with this thing. If they’re not, they’re insane. Well, don’t worry about it. Focus.

  “I think I’m ready to try it out,” I said.

  “All right,” Taro said, “I have the first question.”

  I fished my packet of chaw out of my watch pocket. “Ajpaayeen b’aje’ laj k’in ik’ … ,” I said. (“I’m borrowing the breath of today.”) I tapped the screen five times, cast some so-called virtual seeds over the board, and snuck a look behind me, at LEON. Ripples rolled up through the fluid as the thing started to really think. I nodded that I was ready.

  Taro threw me a few softballs and then started asking tougher questions. Suddenly the new board seemed bigger than I’d thought, as though my runners could get lost out in the wasteland and not get back until sometime after the Big Crunch. And LEON was a bear. Right away it felt like one of the best players I’d met. And it was certainly the fastest. But for a first shot I didn’t do as badly as I might have. The Game was the one area of life I wasn’t under-confident about.

  “Our good student is on his way over,” Taro said, after about two hours. “Would you like to play a real-time prediction?”

  I said sure. I got a little flutter, though. Competition was never my long suit.

  “The IRs are downstairs,” Taro said. He meant isolation rooms. I said great.

  What’s going on? I wondered.

  I took a break. I went upstairs, got four vending-machine espressi and a baglet of Jelly Bellies, and came back. Taro led me down to an even more sub-subbasement, through a cold hallway, and into a small conference room.

  “This feels kind of like a test,” I said.

  “Well, you know I like testing everything,” Taro said.

  “If I do really well, do I get to see the Codex?”

  “We would have to call Marena Park and ask her,” he said. At first I thought he was talking about a place. “She is the big boss.”

  I said okay.

  “Tony has been practicing with this layout for over a month,” Taro said. “So I do not expect you to outplay him.” I nodded like, Neither do I, since I am but a humble disciple. “But I know you do better under the stress of
competition.”

  “Right,” I said. Thanks, I thought. Yeah, I just need a little motivation. You know, I do even better with a xenon headlamp shining in my face and electrodes clipped to my scrotum. I was beginning to remember why I’d dropped out of the project.

 

‹ Prev