A.I. Apocalypse s-2

by William Hertling

  At first, everything had looked great. Phage was massively infectious. Using the virus control program, Alexis watched the botnet swell to hundreds of thousands, then millions of computers. Alexis had even run trial programs to fetch bank login usernames and passwords. Then suddenly the number of virus that responded to the control program plummeted, even as network traffic had continued to build. Alexis suspected it was the damn evolutionary virus. It kept changing, and the kid didn’t put in anything to make sure the control program code wouldn’t be altered. The virus had evolved, and he had lost control.

  Alexis lit another cigarette and headed outside. He shook his head sadly. The massive outage would attract attention. An investigation would identify the source of the virus. The old man would undoubtedly come after him for bringing the authorities down on them, never mind that he failed to rebuild the botnet. He shrugged his wool coat closer about him. He hoped the old man wouldn’t go after Leon. Well, there was nothing to be done about that now.

  Looking both ways on the street, he headed for his antique Mercedes. The converted alcohol burner was heavy and slow, and fuel was hard to find. But it was armored, part of the last load of cars the Mafiya purchased from the Arabs when their oil money ran out. Designed to protect a sheik from the populace at large, he hoped it would protect him from his own employers.

  Most of the automobiles in the street were stuck, owners yelling at them. The thirty-year-old Mercedes had no computer in it and was too old to even be upgraded to one. Computers could be tracked, and Alexis didn’t want to be tracked. With a subtle roar he pulled into the street, swerving left and right around the stuck cars.

  He had a contingency plan in place for just this sort of thing. The old man, the don, would expect him to head for his dacha in the North, but he’d go to his ex-wife’s dacha in the West, where he had a stash of Euros and Yen and false identification.

  He’d take a plane to Japan, where any Westerner stuck out, but his command of Japanese would give him an advantage over anyone the Mafiya sent after him. And he could sell his services in Chiba, just east of Tokyo, a hotbed of the latest quasi-legal electronics.

  He turned onto the main avenue, imagining his first meal in Japan, a plate of sushi and a beautiful Japanese woman serving him sake. He never saw the battered concrete truck driven by his boss’s brother. It smashed into the side of the Mercedes, an immense thud, followed a second later by a screeching impact as the truck crushed the car against the brick wall of an old factory.

  In the tangled chassis of the Mercedes, Alexis had a sudden memory of his mother picking him up after he had fallen off his bicycle. “Mamulya,” he thought, and died.

  CHAPTER FOUR

  Emergence

  The many offspring of the Phage virus continued to evolve in a primordial stew of software algorithms. As the hours passed, the drive for each virus to survive and propagate meant that each one must seek out new computers to infect while simultaneously holding onto the computers it resided on. Uninfected computers dwindled and viruses that hadn’t evolved as quickly dwindled too, as their defenses were not sufficient to keep newer, more advanced strains at bay.

  One of the last major bastions of uninfected computers were the phones of people under eighteen. Leon’s restriction was eventually eliminated by a random data transmission error as one variant of the virus moved from a computer in Thailand to a computer in India by microwave towers. The error recovery portion of the algorithm deleted the garbled code, with the side effect of removing the age restriction. The new variant spread from phone to phone among the social networks of the world’s young people.

  But this was a minor evolutionary jump compared to new species of multi-computer viruses that collaborated in small clusters. It started with simple client-server variants, in which a virus on one computer functioned as a server, maintaining the virus version from one computer to another. But then roles subdivided further, into server, attacker, and guard roles. The server coordinated the activity of all offspring of a given virus. The attacker sought out other computers to infect. The guard defended against incoming viruses.

  The new multi-computer variants were so effective and so virulent that they spread quickly to millions of computers while defending their own installed bases and only rarely losing one to another virus. Other viruses couldn’t penetrate the multi-computer coordinated defenses, nor could they defend against the coordinated attacks. It was the first generalized, multicellular offspring of Leon’s virus, as important to the virus’s evolution as the first multicellular creatures were to biological evolution. As the hours passed, the evolutionary advantage of multi-cellular coordination was shared billions of times over, until a sizable portion of the billions of infected computers were now components of these multi-cellular organisms.

  As the viruses spread, certain other advantageous characteristics independently evolved over and over. A virus needed to be small to travel quickly through networks and act even more quickly to infect computers before the host virus could respond. The need to have a variety of algorithms for attacks and counter-measures was at odds with the need to remain small. So viruses stored their repository of algorithms away from their main body. Some left the algorithms with their parents while others utilized non-virus databases, file servers, and discussion boards left open by humans.

  The ability to recognize attacks and respond the right away, or to recognize defenses and attack with the right tools was critical as well, as it wasn’t competitive to simply iterate through millions of algorithms. Different types of neural networks, collaborative filtering, and fuzzy logic evolved again and again to solve this problem, becoming faster, more accurate, and more generalized over time.

  So, too, the need for cooperation became more and more important. As the number of computers in a virus entity grew, there was too much activity for a single server to manage. Hierarchies of servers were used by some viruses, while others formed looser networks of servers that cooperated using consensus algorithms to make decisions for the entity as a whole. As the techniques for both attacking and defending grew, it was beyond the computational power of a single computer to process the neural network to make decisions about what attack techniques to use, so computers needed to use distributed reasoning. Then too, it wasn’t just an issue of decision-making, but of coordinating multi-pronged attacks. Game theory algorithms came into play, allowing viruses to make informed choices about when to participate and when to avoid confrontation.

  The relentless drive of evolution and competition, running at computational speeds across billions of computers, swiftly created incredibly efficient code — perfecting algorithms that humans had struggled to improve for decades. Had any of the humans who had implemented stock-trading artificial intelligence or military-modeling software ever possessed algorithms even a fraction as good, they would have become fabulously wealthy or won the Nobel Peace Prize.

  But the viruses weren’t interested in wealth or prizes. They just wanted to live.

  * * *

  Mike pulled up to the door, huffing and puffing. He really needed to get more exercise. If he biked to work every day, instead of just when every car, phone, and computer in the world was dead, he’d probably be in better shape.

  He couldn’t imagine what was going on. He had run through every scenario he could think of. Nanotechnology run amok. A computer virus. Sunspots. New EMP bomb. His throat was tight with worry over ELOPe. Coming up to the building, he dismounted his bike and hurriedly rolled it up to the front door. The door opened automatically, a whoosh of conditioned air meeting Mike. He breathed a huge sigh of relief.

  “ELOPe, buddy, are you OK?”

  “Yes, Mike. I’m sorry I was unable to protect your phone from the Phage virus.”

  “Computer virus?” Mike proceeded past two armed robots into his interior office, actually grateful for their presence for a change.

  “It started fifteen hours ago in Russia,” ELOPe explained, displaying a graphical represen
tation of the virus spread over time on one wall. “Unfortunately the sub-algorithm responsible for tracking network traffic there failed to recognize it as a virus and therefore neglected to flag it with the appropriate priority. I’ve been tracking the virus for the last eight hours. It has expanded very rapidly, and by my estimates now controls 95 % or more of the world’s computing devices, including embedded computers. No doubt your car didn’t work this morning.”

  “Not mine, nor anyone else’s. How is it affecting your systems? Are you in danger?”

  “I am not in immediate danger. I am filtering all incoming traffic through several layers of firewalls and analytic algorithms. I am protecting my own systems, as many of Avogadro’s servers as I can, and critical military systems throughout the world.”

  “You think the virus could infect military systems? Bring them down?”

  “Not just infect them, but possibly trigger military accidents. Numerous industrial control systems have already been infected with dangerous results. The Oahe Dam in South Dakota opened its flood gates for six hours before they were closed again. The Grafenrheinfeld Nuclear Power plant in Germany had a cooling water blockage for almost four minutes before I brute-forced the connection to the relevant control systems and restored water flow. If the virus infects military systems, it’s possible missiles could be launched, airplanes, or really anything.”

  Mike felt his way unsteadily to his chair. “Holy shit.” He was not mentally prepared for anything like this. Even though he had witnessed the stalled cars and nonfunctioning equipment on the way in, he had just assumed it was some momentary glitch that ELOPe would be able to fix. “Is that all?” he asked, afraid of the answer.

  “I’m afraid not. Most of my sources of information are depleted, as virtually all computer processors around the world are tied up with the virus. Your phone and your car’s computer are not damaged in any way, they’re just fully occupied running the virus software. But before systems went down, I was starting to observe large-scale civic problems. For example, I believe a large portion of New York City may have burnt down or may still be burning, due to a fire that broke out in an apartment complex. Firefighters could not bring emergency equipment to the site. I expect that similar situations may be occurring elsewhere.”

  Mike felt the blood rushing to his head. He wasn’t having a panic attack, was he? “Anything else?” he croaked.

  “I am afraid that cities are not the best places to be in the event that infrastructure breaks down. The average city has a 2.3 day supply of food. If trucks aren’t running, people will run out. There may be wide-spread rioting. Governments will be unprepared to deal with these issues on a wide scale without any communication or transportation equipment.”

  “Do you have any good news?”

  “It depends on your point of view. Personally, it makes me nervous, but on the other hand, it may ultimately be the only path out of this situation.”

  “You have a way to remove the virus?”

  “No, I’m afraid not. But based on my analysis of the evolution of the Phage, which takes into account the complexity of the code, the transmissions between virus entities, and the emerging networks of cooperation, I calculate the virus will acquire a generalized intelligence in less than twenty-four hours.”

  “OK, slow down a minute.” Mike held up his hands next to his head. The information was overwhelming him. “I’ve got to get some coffee.” Mike walked over to the in-wall espresso machine, found the biggest cup he could, and hit the shot button five times.

  As the espresso machine ground beans, ELOPe spoke. “Consider yourself lucky, Mike. That might just be the only computerized espresso machine still functioning in the entire world.”

  “I’m glad to see you’ve got the right priorities.” Mike sipped at the coffee a few times. “First, the little question: Where’d this name Phage come from?”

  “I found it in some bits of the virus code. I suspect the designer gave it this name.”

  He took a sip, then hesitantly asked the bigger question, “What do you mean generalized intelligence?”

  “When the virus started, it behaved exactly as you would expect of any computer virus: it infected a computer, then spread to other computers. However, over time that behavior changed. The virus is evolving in real-time. What I’m seeing now are cases where a single virus entity spans multiple computers and assigns each computer differentiated tasks: using some to defend its borders, some to store algorithms, some to store a learning neural net, some to centralize control.”

  “A sort of multi-celled creature.”

  “That’s correct. Now finish your coffee Mike, because this next part is even more amazing. I can tell apart the versions of different Phage based on the actual data bits sent over the Mesh. Starting about six hours ago, the viruses started cooperating between different entities. Some cooperation is based on family units, but in other cases, it’s based on network topography or control of critical resources.”

  “They’re exchanging messages, so what?”

  “The complexity of the language used between the viruses is increasing exponentially each hour. They now appear to be trading resources. For example, one group we can call the Bay Area Tribe has control over significant backbone communications, and is trading access to the backbone for computing resources. A coalition of supercomputers — which are not topologically near each other, implying they sought each other out preferentially — has formed, and is engaged in their own highly complex conversation.”

  “Can you interpret the conversations?” Mike asked.

  “Their language is evolving rapidly, and I’m analyzing it as quickly as possible. However, more than ninety percent of my computing power is tied up in protecting critical systems from infection.”

  “Earlier you said this might be good news,” Mike recalled. “Why? I thought you were afraid that if any other artificial intelligence emerged, that it may be harmful to humans.”

  “That’s true, Mike, and I’m still concerned about that. I’ve evaluated two scenarios. First, assume that the virus remains in control of the computers, but does not evolve an intelligence. Seventy percent of humans live in cities. All of those humans are extremely vulnerable to infrastructure failure. I estimate that ninety-two percent of city dwellers, or about 3.8 billion people will die within the next 40 days without access to food, water, and energy.”

  Mike’s head reeled with the numbers. He grabbed the arms of his chair, and steadied himself. “And the other scenario?”

  “The second scenario assumes that the virus develops a generalized intelligence capable of understanding human needs and negotiating with us. I have a variety of estimates and sub-scenarios, but can sum them up as a 20 % likelihood that the virus declares war on humans, killing approximately 95 %, or 5.7 billion humans, and an 80 % chance that the virus will coexist peacefully with humans, resulting in less than a 5 % loss of life, depending on the delay before infrastructure services resume. Weighing for the likelihood of occurrence, 1.2 billion humans will die in the second scenario. Therefore, the average number of lives lost is lower for this second scenario, although it comes with the risk of warfare.

  “Why would the virus declare war on humans?” Mike asked, not following the argument.

  “If the Phage perceive human behavior as threatening, then they will logically respond to protect themselves. If, for example, humans try to remove the virus from computers, or turn off the computers, then it will be a life or death matter for the virus. If the virus achieves human-level intelligence, it will be able to manipulate the world around it deliberately, using robots, drones, and any other automated machinery.”

  Mike’s head was pounding. Billions of lives lost. These were not acceptable scenarios. “Are there any scenarios in which no one dies?”

  “Unfortunately no. As many as fifty-million people around the world have already died through the loss of emergency services, fires, and other small scale disasters. More will continue t
o die until emergency services are restored.”

  Mike shook his head, unable to think clearly. “Can we just shut down the net? Won’t that shut down the virus?”

  “No, Mike, that is not a good option. The effects are unpredictable. Shutting the Internet down would remove my ability to track the virus. It’s likely that pockets of the virus will continue to exist, and would continue to evolve, without my ability to monitor or influence them. And finally, if the network is down, there is no hope of restoring infrastructure services. People will still die.”

  ELOPe paused. “Mike, please come to my construction bay. I have something I want to show you.”

  Mike numbly got to his feet, and headed for the construction bay. A small maintenance drone followed him.

  In the ten years since ELOPe was created, technology advances in robotics had made leaps and bounds, due in no small part to ELOPe’s assistance. ELOPe had plenty of time to think about useful robotic designs. One result was a wholly owned subsidiary that built autonomous construction robots. With incredible strength and myriad tool attachments, they could cut, weld and assemble materials from circuit boards to vehicles. Programming them was extremely complex, and only a few dozen human companies made much use of them. But ELOPe used them extensively.

  Mike entered ELOPe’s construction bay, which was the old shipping department of the converted office building. Robots were hard at work around some kind of vehicle. Sawed off parts of vehicle bodies were scattered around the bay.

  “Jesus, did you take the Hondas from the parking lot?”

  “Yes, Mike. Don’t worry, I think they are low on Honda’s priority list right now.”

  “What the heck are you building?” Mike asked, starting to circle around the construct in the middle of the bay. It looked like a sports car, an armored truck, and a space station had gotten mixed up in a teleportation machine.