Charles Sheffield: Dark as Day

“The run results can’t be any better than the models. The Seine won’t change them.”

“Runs also can’t be better than their inputs. The Seine will draw from every data bank in the System, no matter where it is. At the moment we’re starved for Belt data. Suppose that’s the missing ingredient?”

They had reached Kate’s office. It was twice the size of Alex’s, and as cluttered as his was empty. In pride of place on one wall, where Kate would see it whenever she looked up from her work, was a hand-embroidered cloth. Within an elaborate floral border were the words, “Prediction is difficult, especially of the future.”

Alex slumped into the chair opposite Kate, accepted a tumbler of her made-to-order carbonated drink, and said abruptly, “What do you want to talk about?”

“You. How are you feeling?”

“I’m fine.”

“Lie Number One. Every time you meet your mother or anybody in your immediate family, you can’t think straight for hours. No, make that days.”

“So why did you insist that I talk with her?”

“Suppose I’d put her off until later. Would you have been able to work, or would you have worried all the time until she did reach you?”

When Alex said nothing, Kate went on, “You know, your mother just offered you what most people who work here would die for.”

“You tapped in to a private conversation!”

“I might have. Most of it I knew already. Anyway, you were talking in working hours, so I could claim the right. But don’t let’s get sidetracked. Me, I need to earn a living. I have to work, and I have to put up with bureaucratic bullshit. I even generate some myself, though I try to keep it down. But you don’t. You could walk out tomorrow. You’d have the freedom to work on what you want, when you want, where you want. There’d be nobody like me to pester you for reports.”

“You don’t understand.”

“Probably not. But I really want to. I’m a relatively recent arrival, but you’ve been here for over three years. Why do you stay?”

“You have the reason sitting right there on your wall.” He pointed to the hand-embroidered sign. “I agree with Niels Bohr, prediction is difficult. What will happen in the next ten years, or the next fifty? We don’t know. I just happen to think that it’s the most important question in the solar system.”

“I’m with you. And maybe the hardest.”

Kate said nothing more, but sat waiting patiently until Alex at last took a huge gulp from the tumbler, swallowed hard, and burst out, “The models in use when I came here were useless. They couldn’t even predict the past. They’d been run over and over for the years leading up to the Great War, and they never saw it coming until the Armageddon Defense Line was gone and Oberth City was destroyed, and by then it was too late.”

“What about your models?”

“You saw today’s run. You said the right word: garbage.”

“But isn’t that a problem of inputs, and of computer limitations? You designed the models to run with more than ten billion Faxes. That should be enough to include a simulation of every individual in the System, even if you let the prediction run for a whole century. You’ve always been forced to aggregate to a million or less. What do you think of the models themselves?”

“They’re pretty good.”

“I think I ought to call that Lie Number Two. I’m not able to judge what you’re doing, but before I took this job I talked to people whose judgment I respect. I also love modest men, but tell me true. Don’t you have an entirely new theoretical basis for predictive modeling?”

“I believe I do.” Alex could feel the knot inside him starting to dissolve. Was it something in the drink, or something in Kate Lonaker? “At least, no one seems to have run across it before.”

“That’s what I’ve heard. Look, you must know by now that I’m not much of a techie. I’ve looked at your papers, and didn’t get diddly-squat out of them. Can you describe what your models do in words of one syllable, so I’ll understand?”

“I don’t think so. Not unless you have a few hours to spare.”

“I don’t. But your models did predict the Great War?”

“Sort of. When I ran from 2030 on, they reached a singularity in 2067. That was the correct year, but of course you can’t compute past a singularity of the time line. So there was no way of knowing the war’s outcome.”

“You predicted a cataclysm. That’s good enough for me. Let’s go on. I asked you to tell me true, now it’s my turn to do the same. My worry list has three items at the top of it. First, I’m worried that you’ll take your mother’s offer, leave, and set up your own research shop.”

“Not a chance.”

“Why not — and don’t tell me it’s because your mother makes you nervous.”

“She does, but that’s got nothing to do with it.” Alex paused. “You said you love modest men. This is going to sound anything but.”

“I didn’t say I didn’t like immodest men. I’ve certainly met enough of them. Go on.”

“All right. My models may be producing garbage, but every other long-range predictive model that I’ve ever seen, here or elsewhere, is garbage. My models have the potential to get it right. You say you don’t understand what I do, but in a way you don’t have to. Because if you approve my results, they go up the line, and with any luck they’ll keep on going up to the point where the results lead to action.”

“I hope so. Otherwise there’s no point in either of us working here.”

“Now suppose that I go off and do what my mother suggests. I’d have plenty of research funds — Ligon Industries is huge, and it’s all in the family. Vast available assets.”

“Richer than God, if you believe the media.”

“So I run my models, and suppose they produce surprising results. I come here, and say, look what I’ve discovered. What happens next?”

“We’d have to verify them before we could act.” Kate nodded. “Go on. I think I see where you’re leading.”

“You’d verify them. Of course. And verify with what? The other models you have floating around here, that I know are crap? No agreement, we can pretty much guarantee that. And it would be NIH for me — Not Invented Here. I could come in showing that the Sun would go supernova, and I wouldn’t be heard. I’m working on the most important question in the solar system, but what’s the point if I’m not taken seriously? And for that, I must be an insider. Does that take care of your first worry? I’m not going to leave, unless somebody higher up comes along and throws me out.”

“Which conveniently leads me to my second worry. You told me that you can’t easily describe your models in a way that I can understand.”

“It would take hours.”

“I believe you. But I can’t accept that answer. Because I have faith in you and your models, and I assume that soon — maybe starting tomorrow — they will start producing meaningful predictions, results that we really believe. So I take them up the line to Mischa Glaub. And the first thing I’m asked to do is explain what’s going on in a way that he can understand — and he has a lot less time to spend on this than I do. Then he has to brief his boss, Tomas De Mises. And he has to explain to anyone on the Council who shows interest.”

“You make it sound impossible.”

“If you talk about ‘iterated multiple convolution kernels,’ which is a snappy phrase I remember from your last paper, it is impossible. I want you do to something for me, and put it as high on your priority list as anything in your models. I want you to find a way to explain the models in a way that someone with no special training will understand.”