12

The control room was cold, and Alex was sweating. In one hour, he and Kate — which in practice probably meant he alone — had to give the most important briefing of his life. It was also likely to be the most difficult. He had insisted that the computing and data resources of the Seine were all that he needed to make his model into a practical prediction tool. Kate had told him that was bullshit, because the model was producing nonsense. He wasn’t sure he believed her. He was sure that he had no idea what might be going wrong.

Calm, stay calm.

First, run the model for the decades preceding the Great War. As before, it predicted the occurrence of the war to the year and to the month. Beyond the war that run of the model offered no prediction, but that was as it should be. Such a traumatic event was a singularity of the timeline, beyond which prediction was impossible.

So what about the runs that Kate had completed while he was, in her words, “diddling little Lucy”? He wasn’t sure it had been Lucy-Maria, but it wouldn’t help with Kate to explain that uncertainty.

The model automatically stored every parameter of every run. Alex called on the Seine to perform in parallel all the runs that Kate had tried, one after another, the previous night. It would take months to track every variable, so for the moment he wouldn’t try. He settled for gross aggregates. The crucial number for the moment was total solar system population. He asked for that value, averaged over all the runs that he and Kate had performed, to be displayed as a function of time.

And here it came, the number of humans in the whole solar system, for every year in the coming century and a half.

The starting value, for the year 2097, was today’s actual count of 5.2 billion. The number was rounded, to two significant figures, but Alex had demonstrated, over and over, that his results were not sensitive to small errors in inputs or minor changes in initial conditions. The value for 2098, 5.3 billion, came five seconds later than he expected. The amount of computation that Alex’s model required was enormous, but nowhere near enough to tax the capacity of the Seine. However, he did not command the system’s highest priority. That went to emergency real-time missions, and to the often-meaningless (in Alex’s humble opinion) computational demands of other government programs.

2099: the rounded average over all runs remained at 5.3 billion. Alex spot-checked the exact value, which showed an increase. 2100: sure enough, the number was up slightly again, to 5.4. Alex was aware of Kate at his shoulder, watching not the display but Alex himself. The years moved on steadily, the population count crept higher.

Maybe nothing was wrong. Maybe Kate had screwed up. Alex’s models were different from all earlier predictive models. He was still struggling to find a way to explain those differences in a way that Kate’s chosen test subject, moronic Macanelly, could understand, but every conceivable model had certain things in common. It had endogenous variables, computed within the model and used to produce future values of those same variables; and it had exogenous variables, values that must be fed into the model from some external source.

Any model needed both.

2105: 5.6 billion; 2106: 5.7 billion; 2107: 5.7 billion…

Endogenous variables were easy, you simply provided their present-day values and the model ran with them to estimate their future values. The big question was always, where do you get values for exogenous variables? In a model designed to predict the development of the solar system for a century or more, the one thing you knew for certain was that there would be surprises.

2110: 5.9 billion; 2111: 6.0 billion; 2112: 6.1 billion…

The increase in solar system population, averaged over all model runs, was accelerating.

By definition, a surprise was something that no one could hope to predict. And since surprises were inevitable, in this sense Alex’s model runs were all bound to be wrong. Cheap faster-than-light travel, true immortality as opposed to today’s longevity treatments, the arrival of aliens in the solar system; any of these might happen, in any future year. Alex’s model could assume any one of them, and still make predictions. However, none of the runs that Kate had made assumed any such thing.

2117: 6.5 billion; 2118: 6.6 billion; 2119: 6.7 billion…

What Alex had learned, over years of experience, was that futures in which no surprise exogenous variables were introduced tended to be conservative futures. They had slower growth, and better stability.

But what about war, all-out combat like the Great War that had torn the solar system apart thirty years ago? That produced huge effects, but it was not an exogenous variable, introduced from outside. It arose directly from a steady series of changes in human activities. It was a prediction of the model — the main prediction, in fact, that motivated all of Alex’s work. If your model could predict that a war was on the way, you had a chance to explore changes in exogenous variables to make the war disappear.

2124: 7.6 billion; 2125: 7.8 billion; 2126: 8.0 billion…

But suppose that, without future surprises or another great war, humanity over the span of centuries dwindled and faded and vanished? What did the modeler do then? If Kate were correct, and the model runs all went that way, Alex had better have some kind of an answer ready for the meeting with Mischa Glaub. Kate thought Glaub might have a couple of other people with him, members of the project review committee.

2134: 9.2 billion; 2135: 95 billion, 2136: 99 billion…

They had come to the year of the model run where Alex had left last night — the place where he had dragged himself away from the displays and headed off with his mother to meet Cyrus and Lucy-Maria Mobarak. At this point Kate had taken over. Now he had to pay extra attention to other variables, while continuing to monitor population growth.

2137: 10.0 billion. Running along smoothly, except that the rate of population growth was suddenly down. Now he recognized another complication. The model was set up to accept inputs, where appropriate, from other sources. Before the Seine came into operation those other sources were limited and well-defined. Now, suddenly, a million new data sources could feed the model. They included other predictive models whose outputs Alex did not trust.

“How did you limit exogenous variable inputs?” He snapped the question to Kate, without taking his eyes off the display.

“I cut them out.” She was standing very close to him, where she could see everything he saw. Her breath on his cheek was as warm as her voice was cold. “You left without telling me how to pick them, or what values to use. I didn’t take any that you hadn’t included.”

Alex nodded. New exogenous variables were a source of possible instability. Kate had made the conservative choice, by prohibiting new ones. All the macroscopic measures looked good to him. He could see no sign of the precursors of war, the ominous indicators that had popped up all over when he did simulations of System activity for forty years ago.

But something odd was happening. The model was now forty-five years out, and although the population count continued to creep up, two other variables had reversed their trend. The index for Outer System activity was down, with the cancellation of three development projects among the moons of Neptune. Just as disturbing, no new extra-solar probes had been launched for the past seven years in any of the model’s predicted futures.

“Can you see it?” Kate didn’t sound angry anymore. She was just tense.

“Yes.”

“Do you know what’s causing it?”

“No idea.”

“What are we going to tell Mischa Glaub and the review committee?”

A good question, but not one to concern Alex at the moment. He had too much else on his mind. Sixty-five years out, with no hint of coming war; but transportation cargo volume and inter-world passenger traffic were down. So were terra-forming activities, Oort Cloud exploration, free-space research stations, and what Alex thought of as “inverse terraforming” — the genetic modification of Earth’s plants and animals to match the geography and physiography of other worlds.

And now, in 2140, the population curve was totally flat and he thought the fitted curve showed the first hint of a downward slope. Why, when the solar system was peaceful and stable?

“Mineral deficiencies?” He hit the sequence to provide figures on population breakdown. “Maybe reduced fertility?”

“You think?—” Kate was crowding him, almost sitting in his seat.

But Alex had an answer before she could complete a question. The available minerals and trace elements needed for human existence were on the increase. The fertility indices were fine, general health was better than ever, longevity was increasing — and still the figures for total System population were declining. As the model moved forward, ticking ahead another five years, the gentle decrease was turning into a nosedive.

“What’s happening, Alex? What’s causing it?”

“I don’t know.” He wanted to say, this is impossible, it can’t be happening. Either you have a steady human expansion, or you have a war. Humans just don’t die out, with no reason. That had never happened before in any of the models he had seen — his own, or other people’s.

2152: 7.1 billion; 2153: 6.4 billion; 2154: 5.7 billion; 2155: 5.0 billion; 2156: 4.3 billion…

The population prediction wasn’t just decreasing, it was plummeting. Alex waited and watched, but in his mind he had already extrapolated the curve. They were losing seven hundred million people a year. Unless the curve flattened, in a little more than six years the projected human population of the solar system would be zero.

A plague, a major universal plague which left no survivors? That was the only thing he could think of. Such a disaster could certainly occur, as one of the surprises which any real future might contain. But from the point of view of the model, the plague would have to be fed in as a new exogenous variable. Neither he nor Kate had introduced any such event.

“Alex…” Kate said.

She didn’t need to say more. The year index read 2160. The population count was 1.5 billion. As they watched, the year advanced to 2161 and the count fell below a billion. 2162, 2163, 2164… The count slowed, steadied, hovered around the one hundred million mark. And then — 2165, 2166, 2167, 2168 — the number began a final and implacable downward run.

By 2170 it was over. In that year, and in every year beyond, the human population of the solar system was a steady, flat zero.

They stared at the display in silence. Finally, Kate said, “Well, it is only a model.”

Normally those would have been fighting words. To Alex’s mind, provided that you fed the model reasonable inputs and possessed enough computer power, the results you got back were a possible future. And more than possible; plausible. Not the only conceivable future, certainly, because of surprise factors that no model could include. A future, however, that was far more than an assembly of random predictions.

Now a hundred different runs, with a hundred different sets of initial conditions, pointed to the same melancholy conclusion: No humans by the year 2170. Alex was reluctant to believe the results, but he could see no basis for rejecting them.

Population zero; and two years before that, all transportation, development, and outward expansion had ended. He was still staring at the flat-lined results when Kate made the day’s discomfort complete.

“Come on,” she said. “We’re running out of time. In half an hour we have to brief Mischa Glaub. I hope you have something sensible to say. Because I sure as hell don’t.”


Kate had warned of a tough review session. Alex decided in the first thirty seconds that the real thing was going to be worse than her fears.

It began as soon as Kate led Alex into a small conference room flanked by modern display panels and old-fashioned pictures. The latter were of stern-faced individuals, presumably past members of review committees.

Seated at the table were four people — Kate had promised at most three — no less stem-faced than the images on the wall. Alex had met Mischa Glaub before. He was a short broad man with a shaved head, a sour expression, and a permanently angry disposition. Old hands in the Department of Planning made it a point to avoid meetings with Mischa in the morning. Food, it was said, softened his ire. Unfortunately this session was starting an hour before lunch, and it would run until the committee members were satisfied.

Alex had also met, at least briefly, two of the others. They were Glaub’s boss, Tomas de Mises, and Ole Pedersen, head of the Methods and Logistics Directorate which sat at the same organizational level as Mischa Glaub’s empire. It was no surprise to find them here, though Pedersen’s presence might be a problem. Kate had warned Alex that Ole Pedersen was wily and ambitious, always promoting his own group’s products and critical of anyone else’s. Tomas de Mises was less of a worry. He was older, close to retirement, and reluctant to say or do anything likely to cause trouble.

The final person at the table, however, was the real shocker. She was a middle-aged brunette, whose white skirt and floral blouse broke the rule for office uniform. Rules were for lower level staff. Alex recognized her as Magrit Knudsen, Tomas de Mises’s boss. She was a major force, already a Jovian Worlds’ cabinet member and tipped to become a leader in the Outer Planets’ Development Council.

Why was she here, for a routine progress review meeting?

The way that Kate had explained the procedure, Alex, or possibly Alex and Kate, would brief Glaub and a couple of staff assistants. They would brief de Mises, and de Mises would in turn provide summaries for people farther up the chain of command. Apparently normal procedure didn’t apply today. Alex was expected to brief the whole ladder at once, from top to bottom. Magrit Knudsen seemed to be studying him with special interest.

Kate’s raised eyebrow said, “Don’t blame me. I didn’t know she’d be here, either.” But there was no time for discussion, because Kate and Alex were barely in the room before Mischa Glaub snapped, “All right, let’s get on with it. And keep it short. We’ve all got other business to attend to.”

Kate glanced at Alex and nodded. He was on, with an instant decision to make. Either he described what they had as early results, the product of a still-evolving model and therefore not to be taken seriously, or he said what he really believed, which was that his model was right, that it was far superior to anything that had ever existed before, and that it predicted terrible danger in all plausible human futures.

The rational thing to do was to be modest about the model, dismiss this set of results, and promise better in the next review. There were two problems with that. First, given the promises that Alex had made for model performance once the Seine was in operation, there might be no next review. The whole project was likely to be scrapped. And second, Alex was a lousy liar. He couldn’t stand up and make statements that he didn’t believe. What he did believe was his model.

Avoiding Kate’s eye, Alex described the runs of the past two days and displayed their results. At first, the four people across the table sat and listened, sometimes nodding approval. Then he came to the critical years and showed the trends flattening and turning down. The audience became restless.

The model reached 2154 and the population dipped below 6 billion. Mischa Glaub was the first to break. He exploded, “You know what you’re showing there? You have the whole bloody System in catastrophic decline. But I’ve seen six other projections in the past month, and not one shows anything but expansion.”

Alex drew a deep breath. “All the other models are no good.”

Pedersen, whose group had produced three of those other projections, said, “Now just a minute, if you’re going to accuse my people—” Mischa Glaub snorted and said, “Cut the crap, Ligon. Unless you got damned good reason—” At the same time, Kate said, “What I think he means is—”

“Why?” Magrit Knudsen spoke no louder than the others, but her one word cut them off in mid-sentence. She went on, “It’s not enough to tell people that your model is right, and all the others are wrong. You have to explain why your model is better.”

When Alex said nothing, she added, “Ligon — that’s your name, right, Alex Ligon? — there’s an old saying: a man who understands what he’s doing can give an explanation of his work that the average person can follow over drinks in a bar. I happen to believe that’s true.” She glanced at the clock. “We’re not in a bar, and I’m supposed to be somewhere else. But I count myself as an average person. I’ll give you half an hour. Tell me about your work. Tell me why I should keep funding it, rather than canceling you on the spot.”

She knew his first name, although no one in the room had used it. How come?

Alex postponed that question for later. She had put him on the spot. He had not had the time to fine-tune and polish a simplified explanation to the point where Pedersen’s man, Macanelly, would follow it. He must go ahead with what he had, and hope that Magrit Knudsen was three or four rungs higher up the monkey ladder than Loring Macanelly.

He began with a direct question. “Did you ever take physics courses?”

Knudsen looked puzzled, but she nodded. “Twenty years ago. Don’t assume I remember anything.”

“I’m sure you’ll remember all we need.” Alex in principle was briefing the whole group; in practice he was talking to Magrit Knudsen alone. “For instance, for hundreds of years the scientists who worked with a gas would describe it by some basic properties… Not just what sort of gas it was, but they would measure its pressure and temperature and volume. Later on they got more fancy in their descriptions, and added things like entropy and enthalpy, which we don’t need to bother with now. People used those basic variables to tell how a body of gas would behave under different circumstances. They called the branch of science that was developed to do this thermodynamics.”

He looked at the others. Magrit Knudsen nodded, tentatively and apparently a little puzzled. Mischa Glaub from his expression was ready to explode, but he and the others wouldn’t override their boss. Alex figured he had about five minutes.

He went on, “The important thing about thermodynamics is that you don’t need to know anything about the gas at a more fundamental level. You get valid results without knowing that a gas is actually made up of separate molecules. The thermodynamic variables you are dealing with actually represent averages over a huge number of individual particles, but your results are correct even if you have never heard the word ‘molecule’ or ‘atom.’

“But then people learned about molecules, and they had a mystery to solve. How did the overall general properties they’d been dealing with somehow emerge from the action of a whole lot of separate particles? It took a while, but eventually physicists like Maxwell, Boltzmann, and Gibbs developed a theory based on the molecules themselves. The theory was called statistical mechanics, and it showed how to relate the behavior of ensembles of tiny particles to the general thermodynamic properties that people were used to.”

His audience was becoming more restless. Mischa Glaub was squinting and glaring. Ole Pedersen muttered something to Tomas de Mises that sounded like “What the hell’s he going on about?” Even Kate, who knew where Alex was going, was biting her lip.

Magrit Knudsen nodded. “I follow you so far. But I hope this is leading somewhere.”

“It is. The other predictive models used in the departments are like thermodynamic theories. What I mean is, they work with general variables. A general variable can be anything you choose: economic production by industrial sector or by location; overall computer capacity; transportation supply and demand; population; commodities and services. The theories tie these things together, and model the way that they evolve over time.

“But something like transportation demand is a derived quantity. It arises because of the separate needs and actions of more than five billion people. You could say, it’s like a thermodynamic variable that arises from the combined activity of a huge number of small, separate units. That’s a true statement, but it doesn’t go far enough, because all molecules in a gas are essentially identical. Whereas every human being is essentially different.

“My predictive model recognizes that fact. It derives the general variables that other models take as basic. If you want to think of it this way, the model is a statistical mechanics for predictive modeling. It allows you to derive all the ‘thermodynamic’ general variables of the older and obsolete models.”

Alex saw Ole Pedersen’s head jerk. The word “obsolete” was a red flag, since Alex was describing what the models of Pedersen’s directorate still did today. Pedersen was bristling and seemed ready to interrupt. Alex hurried on.

“We can’t stop there, though. There is another necessary model innovation. If you try to deal with all humans as identical, the way that gas molecules are identical, you’ll get garbage for results. Human progress depends to a large extent on the differences between people. So the individual units in my predictive model are not simple equations or data items. They are programs. Each program is a Fax, a duplication at some level of an individual human. My code allows anything from a Level One to a Level Five Fax to be used.

“Before the Seine was up and running, I had to cut corners. It would have taken forever to make runs with five billion separate Faxes, even if I used their lowest levels. So I was obliged to work with aggregates. I knew that was oversimplifying reality, and my results proved it. They were unstable. They blew up, just the way that any predictive program becomes unstable if you make the time-step too large.

“With the Seine up and running, though, I can finally run my model the way it should be run. No aggregation, but with a representation of every individual as an individual. And I can use Level Five Faxes if I want them, with complex decision logic and interaction powers, rather than simpleminded Level Ones. So I’m running a real solar system, with real people. But using the full power of the Seine computer, my virtual solar system will evolve six million times as fast as the real one. A year of solar system development takes only five seconds on the computer.”

“Five seconds? You say only five seconds, but that’s a long time to produce nonsense.” Pedersen stood up. “I’m sorry, but I’ve had enough. All the meaningless analogy with thermodynamics and statistical mechanics, and all the talk of superior approaches. Then you show us that.” The sweep of his arm took in Alex’s final results, still frozen on the displays. “Population zero, humans extinct, solar system development dead. Is there anyone in this room who believes such a thing for half a second? All our other models show nothing remotely like that. I don’t know about the rest of you, but I’ve wasted too much time on this — this horseshit.”

“Now, Ole.” Tomas de Mises waved a hand at Pedersen, palm down. “Don’t let’s go to extremes. Though I must admit…” His voice trailed off. He stared at Alex’s final results, and shook his head.

“When did you perform the runs you just showed us?” Magrit Knudsen ignored the reactions of Ole Pedersen and Tomas de Mises and addressed Alex directly.

“Last night.” Alex didn’t want to look at Kate. If he had been present when the runs were first performed, they would have had time for a more detailed evaluation. “We repeated them this morning.”

“Then everything is less than one day old. Bugs in new models are the rule rather than the exception. I, too, have trouble believing what you have shown us. However.” Magrit Knudsen stared right at Ole Pedersen. “Regardless of anyone’s skepticism, these runs suggest problems in solar system development so grave that we must take them seriously. I do so, even if there is only one chance in a thousand that they are correct.” She turned to Mischa Glaub. “I want this work to continue on a high-priority basis. If you require additional resources, of humans or equipment, do not hesitate to ask for them. That’s it for this meeting.”

She stood up. “If you have the time, Ole, I’d like to spend a few minutes with you in my office, discussing your directorate’s models. You, too, Tomas, unless there’s something more urgent on your calendar.”

Magrit’s tone suggested that was unlikely. When the two men had trailed out after her, Mischa Glaub turned to Alex.

“After what Knudsen said I guess I can’t fire you on the spot, which is what you deserve. I should have known not to hire somebody with more money than sense. Don’t you ever come into one of these meetings again and go off half-cocked with results that you haven’t run by me and checked ten times over. If there’s one thing I hate, it’s surprises. So get the hell out of here, both of you, and work on that goddamm model.”

His expression changed from irritation to poorly-suppressed glee. “But did you see Pedersen’s face when you talked about ‘obsolete’ models? He looked like he was crapping barbecued rivets.”

Загрузка...