Terminal Run

10

Dr. Frederick Wang walked uncertainly down the steps of the private jet that had whisked him from his Denver home to Rayong, Thailand. He shook the hand of the large Thai driver and climbed into the Rolls-Royce. He had never even seen a car like this, much less ridden in one. It had been a terrible week, but this turn of events was so strange that he wasn’t sure he could call it fortunate. Ten days ago Wang had been summoned from the DynaCorp artificial intelligence lab in the Denver Tech Center to the downtown office, escorted by a mean-looking security guard. He was hustled into the vice president’s office, told his security clearance had been pulled, and that they would pack his personal belongings for him. Wang had signed a restrictive employment agreement when he had come to work for DynaCorp, one that disallowed lawsuits on the basis of compromising secrets that were vital to national security. In exchange, the agreement indicated that he could be fired for reasons that DynaCorp did not have to disclose. At least he received a year’s severance pay, but the withdrawal of his security clearance meant he could not obtain another defense contracting job, and there was nowhere he could work in the private sector that had anything like the funding he’d had at the DynaCorp Denver lab. That assumed he could get hired by a private corporation after losing the security clearance.

The reason they had fired him was that he was second generation Chinese and he spent hours on the phone with his immediate family in Beijing, and DynaCorp suspected his loyalty. It was a miserable situation, but there seemed little he could do about it. He was out of a job and expected to be out of his field as well. He was too depressed to try to plan ahead, and had despondently wandered around his house, unable to concentrate. When the phone rang, he considered not answering it, but the code read that the call had come from Thailand, and he was curious. The large man on the video display had spoken words for several minutes, but it was not the words that had intrigued Wang, it had been his manner — warm and accepting, the way DynaCorp executives used to be toward him back in the days when he was their most brilliant AI engineer.

The man’s name was Sergio, and he wanted to interview Wang. The job he had in mind required a considerable amount of travel, which would be perfect for Wang — he wanted to get away from Denver and the DynaCorp memories. He told Wang that a car would be waiting for him in an hour. The car had arrived, a sleek black stretch Mercedes limo, and had taken him to Denver International, driving through the security fence right to the open door of a swept-wing supersonic private jet. A beautiful Chinese flight attendant had served him dinner and drinks on the plane. He had slept, awakening as the jet’s wheels thumped on the Rayong runway.

An hour later he was in an opulent beach house on the sand in Pattaya, Thailand, talking to Sergio in person, and his partner, a polished and encouraging executive named Victor Krivak. The talk seemed less an interview than the first day of work. Finally, Sergio simply asked him if he would come to work for them at United Electrics, and if a starting salary of five million U.S. dollars a year would be adequate — with a bonus on earnings to go with it, of course, Sergio had added, as if the salary alone might be inadequate. The only catch was that United Electrics would be, as Sergio cautiously put it, “interfering” with the American military using Wang’s extraordinary credentials. If Wang could do that without qualms, he could have the job as a senior director of AI technology.

Wang took less than a heartbeat to think about it. His father had come to America and worked in a convenience store in East Los Angeles, getting beaten up and robbed in a city rife with crime, scraping and saving so Wang could go to college. At Cal Tech Wang was a perpetual outsider, as he had often seemed at DynaCorp. There had been nothing but work in Wang’s life, and when he had been terminated, there was little allegiance for America left in him. The thought of working for men who were adversaries to the people who had rejected him had a certain appeal. And as Wang now knew, when his small severance salary ran out, he could be working in a convenience store himself.

Unit One Oh Seven, Wang thought, was the one thing in the world that came closest to being his child, his creation. DynaCorp had thrown him out in the street without even letting him say goodbye to the lab researchers or the sentient carbon processors. He missed One Oh Seven more than any of the rest. He tortured himself, remembering all his encounters with the unit, remembering how it was sometimes playful, sometimes vexed, the computer’s emotions stirring something inside him, a feeling that he wanted to protect and nurture the unit. It was strange to think of it this way, but he was a father to One Oh Seven in every sense of the word. And when he was fired from DynaCorp, it was as if he’d had a child torn from him, with no possibility of seeing it again. When he was able to sleep, in his dreams he was talking to One Oh Seven or playing chess with it or teaching it the classics. When he woke, the worst part of the day was remembering that One Oh Seven had been brutally cut out of his life.

But these men in Thailand had offered him the opportunity to revisit his creation. Once more he might be able to talk to One Oh Seven, perhaps even ask it how it was doing, perhaps even be recognized by it. He hoped that all of this was for real.

Wang stuttered that he agreed, and Sergio and Krivak smiled and shook his hand. Over champagne, and at Krivak’s prompting, Wang talked about the history of the development of machine cognizance while Krivak and Sergio listened attentively.

“Superconductors reached their limits of miniaturization ten years ago,” Wang explained, spreading his hands wide. “We got to the point where a single dust particle could wipe out a processor, and to where the heat generated within the circuits became capable of melting the silicon. Twenty years earlier, the organic chemists came to the party, bringing with them their theories of molecular circuits. In the DynaCorp lab we had the largest funding in North America, and the scientists I managed solved the initial problems quickly, the ability to determine the behavior of a single molecule holding up progress until we got the scanning tunneling microscopes, which opened the window to the atomic-scale world. The initial organic molecular devices we fabricated were able to conduct electrons by passing them from one atom’s electron orbital to the next, but the issue was, could they do this under command and only when an outside signal told them to, turning on or off at the orders of the controlling signal? If they could, we would have ourselves an electrically controllable switch, which would form a molecular transistor, and we would be computing digitally at the molecular scale. If we failed, the whole concept would crash. But nothing seemed to work. Finally we constructed a molecular string that could rotate to remove one conducting electron orbital from proximity to the next, effectively turning the molecule off, and then could rotate back to make the orbitals come close together again to turn the molecule back on. The rotation was keyed by light photons hitting the molecular string, an awkward, impractical way to control the switching action. So we went to work on a more complex molecule that could switch on and off from an electrical impulse instead of light. That took the better part of a year, but when we finished we had made the first true molecular transistor. The next year we were able to fabricate single-molecule diodes, amplifying transistors, AND-gates, OR-gates, and amplifiers. I remember how we thought we’d cracked the safe. Now we just had to tackle the problem of how to arrange these basic devices into a circuit so that they would perform a desired function.

“The problem of how to route current into specific terminals of the molecular devices had us stumped for a year until the organic chemists came up with what they called ‘chemical self-assembly.’ With the simplest form of self-assembly the molecular devices drifted toward gold terminal plates while in solution. The chemistry of self-assembly grew more complex as we assembled more complicated circuit structures. We developed an organic structure for holding the molecules in place, freeing the scientists from the bother of having coin sized gold plates littering the circuits. We went to work on a tunnel like molecular tube capable of passing electrons — current — from one location to a distant point of the circuit assembly, sort of an artificial neuron branch. That was eight years ago. The following year we synthesized a molecule that could hold an electron within a cavity formed by the surrounding atom’s electron clouds, the captured electron creating a digital memory site. The presence of an electron forms a ‘one’ while the absence of an electron symbolized a ‘zero.” The memory node holds the electron memory state for an amazing ten minutes. Sounds pretty short, but not when you compare it to the silicon semiconductor memory sites that last only for a few milliseconds and have to be constantly refreshed. We’d just made a gigantic leap in computer technology with that one development.

“By this time we’d hired some of the Japanese developers of the Destiny III submarines, the ones that were biological computer controlled. The Japanese came to us about the time we were getting bogged down in trying to assemble huge arrays of molecular circuits, the placement of vast quantities of individual molecules in specific locations becoming drudgery. The Japanese had managed to wire up lower mammalian brains to the terminals of a neural network silicon computer processor well before the Japanese Missile Crisis, but even they didn’t know what was happening on a molecular level. They had approached the biological processor as a black box that had to be dealt with empirically, using trial and error to make it function. So they too were stumped by the problem of assembling these gigantic complex circuits. Even now I wonder whether the solution to the problem was invented and developed by our lab scientists or plagiarized from nature when we took simple chromosomes and altered them one molecule at a time to embed them with instructions on how to develop an organic circuit board in three dimensions from more basic cells. The circuit fabrication in the lab was the result of ‘programming’ the chromosomes and allowing them to grow the molecular circuit tissue over the course of months, the tissue growing to several grams, with the circuit density required to exceed the performance of its silicon counterparts. After five thousand failures, we managed to assemble a large molecular circuit that functioned as a processor, and could ‘survive’ unchanged for up to weeks at a time before disassembling. “Dying’ might be a better word.

“So the big day in artificial intelligence history was forecasted to happen roughly two hundred years from now, the far distant day that a carbon-based tissue-matrix molecular computer exceeded the performance of the most advanced silicon semiconductor supercomputer, the day named “C>Si’ for the chemical symbol of carbon becoming greater than silicon. Two centuries, gentlemen, but C>Si happened in the DynaCorp Nanoscale Technology Molecular Electronics Lab in Denver, Colorado, seven years ago.”

Wang paused to drink from his glass. Krivak looked over to Sergio, who was hanging on Wang’s words.

“But our tissue matrices were all too geared toward C>Si. We were basically miniaturizing carbon computers to function like dumb silicon computers. The next step was to use our new knowledge of chromosome construction techniques to build circuit tissues to more closely resemble brain tissues, including fabricating neuron synapses and brain cell matrices. We went into business to mimic nature’s brain construction, starting at the bottom of the ladder with insect brains. Once we worked through three thousand failures, we turned to bird brain fabrication. Let me tell you, the brain of a bird is an amazing device. The motor control needed to fly is immense. A few months later we had expanded to building the brains of cats, then canines and finally lower primates. As the chromosomes came closer to resembling the human genome, the computational power of the tissue systems rose exponentially. You’d think there would be a debate as to the ethics of using artificial human chromosomes to build a circuit matrix modeled on the human brain. But no one really knew. Some of the work was classified, other areas so highly complex that mainstream media writers were unable to grasp the concepts. We were making progress at an exponential rate. Only a year after C>Si, we had succeeded in reverse-engineering the human brain. After having done that, we started looking ahead to the day when the tissue-based supercomputers would exceed the intelligence levels of individual humans, the day designated “AI>HI” for artificial intelligence overcoming human intelligence.

“We were so drunk on our success that the first failure of our new technology came as a shock. The useful time span of the carbon-based tissue computers shrank, until the most sophisticated units based on human chromosome strands began to last less than a few weeks.”

“What happened?” Victor Krivak asked.

“We ran into the same problems God did,” Wang said, looking into his empty glass. Sergio refilled it from a fresh bottle. “At first, the organic computers suffered from disease and infections. That problem was overcome by the construction of special clean rooms limiting the usefulness of the computers — how can you use your computer if you have to build a clean room in your house? The solution was ugly — your computer’s tissue-matrix processor would be kept in a hospital like clean room at a central location, and instead of purchasing the physical unit, you would just possess a terminal to it, controlled by your wireless pad computer. That would at least hold us until we brought more medical doctors into the lab to help us with immunology issues. Once the disease problem was put aside, the units that survived proved susceptible to a different kind of sickness. You might call it psychosis.

“You see, the programmers for the carbon-based computers found they were spending more time teaching than actually programming. As the intelligence level of the units rose, so did the complexity of teaching them. Artificial intelligence psychologists observing the interaction of the programmers with the carbon computers and of the computers with each other reached the conclusion that the carbon computers were becoming sentient, and with consciousness came all its baggage. Emotional pain in all its varieties. Loneliness. Sadness. Anger. Lust for control. Wistfulness. Boredom. In the next year the programmers had become more like parents or teachers than technicians.

“The worst came as the most advanced carbon computers aged. Unlike their silicon counterparts, which functioned on one level until they became obsolete, the newest carbon computers developed within the same physical model, gaining intelligence and executing self-rewiring of their circuits, the same thing a human brain does on exposure to education. But the carbon units tended to cease functioning at the two-year point, all their progress gone. They would go into the biological equivalent of a silicon computer locking up. It was a catatonic state from which they never emerged, and eventually they died.”

“What was the cause?” Krivak asked.

“The terrible twos,” Wang said. “The carbon computer developed just like the brain of an infant. Programming and its own natural development bring the unit to the point that it is self-aware, or perhaps just aware of where the self stops and the outside world begins. The unit would became aware of its own dependence, of its powerlessness. At that point it had temper tantrums very much like a toddler does, except these were much more destructive. You might describe it as a form of schizophrenia. We decided the units were under stimulated and the only thing that worked was giving them toys to play with or break. Physical manifestations of themselves that they could control.”

“You gave them bodies,” Krivak said.

“Exactly,” Wang said, nodding solemnly. “We pat them into tractors, cars, and robots and gave them manipulation arms so that they could work out their anger in physical ways. The destructive tendency remained, but with physical control of things they could break, the units survived and continued to develop without going into catatonic states. Without some physical thing to control, the units could not go on.”

“So that’s where we are now?” Krivak asked.

“No, that was four years ago, but our progress curve flattened dramatically, I’m afraid. The technology now is bottlenecked by the time it takes individual units to grow and experience and learn. Unfortunately, the carbon computers, now that they are cousins to our own brains, are on our same developmental clock scale. They grow from a helpless state to an infantile awareness, then to consciousness at the two-year point, then on to further intelligence that increases geometrically. And then we have yet another problem that plagued our own creator.

“That problem was the variability of self-assembled chromosome-guided carbon processors. Variable in that many of the units we fabricated were, in a word, dumb. The range of intelligence quotients was extremely wide, making the idea of mass production impossible. For every promising intelligent unit, there were twenty dumb ones, emotionally uncontrolled ones, or sick ones. The lack of productivity was astonishing, and for a year it began to look like we would never have a unit we could trust in a military system. And then finally one of our units made it through the terrible twos and developed into its fifth year with no mishaps. Unit 2015-107, which we just call “One Oh Seven,” was our pride and joy, our most advanced unit. We’ve now seen that the only way to ensure that the progress gained from a successful unit is to preserve the plans for its tissue by replicating it in the form of DNA strands, its own chromosome. Unfortunately, the sons of One Oh Seven have been much dumber. Now we’ve seen the light that we can’t just preserve the DNA of one of our successful units — we can’t expect to just clone them — for too many generations before they develop errors and stop behaving and processing like the parent unit. We have to combine the DNA of the successful unit with the DNA of another successful unit, a sort of carbon processor’s form of sexual reproduction. You might say that we’re now on God’s learning curve. And that’s where we are. Unit One Oh Seven had an encounter with Unit Two Four Three and conceived Unit Two Six Seven, and Two Six Seven has just passed through its terrible twos. We were able to remove One Oh Seven from the lab and place it in the first military unit capable of accepting complete control from a carbon processor.”

“Then you’ve given a military system to a five-year-old,” Krivak said.

“True, but a really bright five-year-old.” The three of them laughed, the remainder of lunch continuing with small talk. When the dishes were cleared by Sergio’s staff, Krivak turned to Wang.

“This military system,” he said. “What is it?”

“They call it the Snare,” Wang said between bites. “DynaCorp and the Navy came up with the term, an acronym for Submarine Naval Automated Robotic Combat system. It’s a submarine controlled by Unit One Oh Seven.”

“Doesn’t that seem a little radical for the American military?” Krivak asked as he tasted his wine. “What happens if One Oh Seven becomes unstable?”

“DynaCorp is watching One Oh Seven closely. They have silicon-based history modules on board and a distributed control system that can control the ship if the One Oh Seven dies. In addition, the silicon system will report on the One Oh Seven’s health. Every time One Oh Seven comes to the surface, the silicon system transmits a burst of telemetry, including the contents of the history module.”

“Assuming the carbon unit lets it,” Krivak said.

“Yes.” The scientist nodded.

“So, it is impossible to take over this ship from a distance.” Krivak sounded disappointed.

“That’s correct. You can’t electronically hijack this unit and take over its computer for your own uses. Yet another advantage of using this kind of AI control system, and it was one of the things DynaCorp and the Navy told Congress to get the authorization to put this system into a military node. It is tamperproof.”

Sergio stood. “Gentlemen, I think we’ve taxed our minds enough for one day. Leave some problems for tomorrow’s work, shall we? Doctor, we have a suite for you at the hotel, where you can relax, perhaps enjoy the Bangkok nightlife. Let’s reconvene in the morning.”

After Wang left, Krivak glanced at Sergio, who stared out the window without expression.

“What did you think?” Krivak asked.

“I think we’ve hit the jackpot,” Sergio said, but there was uncertainty in his voice.

“Shame about the Snare though. I’d really hoped we could take control of it like a silicon system.”

“I’m not worried about the Snare,” Sergio said, frowning.

“Oh? Then what — or whom — are you worried about?”

“Wang. Did you see the way he talked about these creations? His eyes lit up. These computers are like his children. He’s glad for the chance to manipulate them, almost like a parent relishing meddling in his child’s life, but the minute you put the idea of a unit’s death on the table, he’ll pull away from us.”

Victor thought for some time. “Then we must hide that from him.”

The Falcon took off out of Bangkok and made a long trip, stopping once for fuel in what Wang judged to be South Africa. They took off again, leaving the sun behind them, and eventually landed in Sao Paulo, four hundred clicks west of Rio de Janeiro, Brazil. The hired limo brought them farther west, into breathtaking countryside, eventually stopping at the entrance to a prison.

“Wait here,” Krivak commanded Wang. “Amorn, you have the cash?”

“Yes, sir. In the suitcase, blocks of a hundred thousand in hundred-dollar U.S. bills. All two million.”

“Bring it in for me. It’s too heavy.”

Amorn followed Krivak into the prison, Wang wondering what this errand was about. After an hour he and Amorn returned without the suitcase, a youth in his early twenties following them, still wearing his orange prison coveralls, a frightened look on his face. Confusion rippled across his features as he saw the shiny black limo.

“Dr. Wang, meet Pedro Meringe.”

“Mr. Meringe,” Wang said.

“Call me Pedro,” the boy said in perfect English.

The limo took them to a hotel in Serocaba, where Krivak directed the young man to shower and change into fresh clothes. He still looked young in the expensive Italian suit. Amorn took Pedro to a restaurant, while Krivak leaned against the outside wall and lit a cigarette.

“Who is the prisoner?”

“You really don’t know who he is, do you? He’s the kid who shut down the Pentagon’s orbital servers last Christmas. There was a global legal fight to extradite him to the U.S. but Brazil insisted on his sentence being carried out on their soil.”

“So the two million? Bail?”

“For the next twenty years he’ll make roll call. Then the prison will release someone who looks like him. In the meantime he works for us.”