Artificial Intelligence and Machine Learning Last byte


Upgraded with new instructions, my AI aims to debug its original programmer, along with his home planet.
  1. Article
  2. Author
Deadlock, illustration

I was NASA’s chief orbital technician, as the agency was changing its name to NAIA, the National Artificial Intelligence Administration, responsible for de-orbiting the International Space Station. The ISS name itself was a misnomer, coined when the space program was sliding from dream toward delusion. By definition, space stations are orbital facilities where astronauts transfer from Earth-launch vehicles to interplanetary ships intended for, say, the first human expedition to Mars, which indeed never took place. Having accomplished little worthwhile scientific research since its 1998 launch, the ISS was primarily a propaganda tool, pretending the technologically advanced nations of the world had become partners and that humanity had a glorious future in outer space.

After decades of indecision, the alternatives now were to find a new purpose for the ISS and boost it to a higher orbit or crash it back to Earth at a safe location in the vast Pacific Ocean. Allowing the orbit to degrade naturally could have flattened part of a city, with great loss of life and national prestige. So I was ordered to program the precise instructions into the small thrusters that controlled its orientation, then fire a retrorocket that had recently been added, to drop it to its ocean target zone far from any ships. I secretly pondered violating my orders, however, lofting it instead to a higher orbit so it would survive until the spaceflight social movement could convince politicians to revive the program.

I could not share my illegal idea with my fellow government employees, so I sought the council of Tobor, my reliable humanoid AI assistant. “Well sir,” Tobor said, “using rule-based reasoning we see four options: (1) If you do nothing, the ISS will fall a few weeks later at a random location within a band from 51.64 degrees north of the equator to 51.64 south; (2) If you follow your orders, then the human attempt to inhabit the cosmos will end; (3) If you boost it to a higher orbit, then the future will become totally uncertain; or (4) If you intentionally aim it at a city, you will need to select one, and I have no criteria for making such a selection.”

Figure. Training in virtual EVE Online to operate the real International Space Station.

I waved my hand in a way Tobor was programmed to recognize, guiding him to shift to his neural-net modality, wringing his android hands as he mumbled, “Well I kind of sympathize, wanting to make things better, worrying about making things worse, and marginally between with nervous hidden nodes.” Getting no help from Tobor, I did as my managers had ordered, and the ISS splashed safely into the Pacific. My idea of saving the space station had been only a brief fantasy, while my far more fundamental challenge would take months to develop.

I had received an unexpected message from an old online friend, Arnold Touring, whom I had actually never met in person but with whom I had shared many hours of questing in space-related virtual worlds. We had first connected in the Star Wars Galaxies game, and when it shut down at the end of 2011, we moved together to Anarchy Online, a grand old online role-playing game that applied European political theory to a colony on a distant planet called Rubi-Ka. Subsequently, we hopped over to EVE Online, set in an even more distant galaxy, then lost contact with each other.

Touring had become an influential astrophysicist, though I had little idea as to his recent research, in gamer terms, scientific questing, or technological crafting. After swearing me to secrecy, he told me the loss of the ISS was inconsequential, because his laboratory had detected and begun, incredibly, to decipher digital streaming from another solar system, still hidden from the public. He was now secretly employed by the U.S. State Department, one of the many agencies that had lost its traditional function and was now seeking new justifications for its taxpayer-funded budget. After several chats in Third Life, he invited me to join his team, working remotely in this non-game virtual world to interpret what appeared to be alien source code.

I have made two terrible mistakes in my life: Following orders to destroy the ISS was a failure of imagination, but now I forgot to look beyond my own imagination, failing to recognize the great potential for harm subordinating our technology to alien purposes, concentrating instead on computational puzzles of interest only to myself. The format of the alien software seemed a bit BASIC-y, with sequentially numbered fragments of code, and I quickly learned the extraterrestrial equivalents of GOSUB, RND, and REM. At least we thought that is what they were, and we grew excited when our first short program seemed to work, allowing us to write “Hello Earth!”

The data was still streaming down, and we soon had a very large program that was automatically adding procedures by the minute. It naturally unfolded into the equivalent of both parts of an online game that reminded me of Star Wars Galaxies, the user-side code and data to create the virtual environment and the server-side database to connect users and ensure their machines represented correct information. For a time, Touring was unconvinced the extraterrestrials from Rubi-Ka or the New Eden Galaxy were sending us a mere videogame, but I reassured him it was their way of offering us a virtual experience reflecting their alien world. He worried this might be dangerous, but I reminded him of the fun we had on Rubi-Ka.

“It was the size of a real planet but structured more like a giant rainbow crisscrossed with golden arcs of lightning and layers of emerald ledges a stairway to extraterrestrial heaven.”

Meanwhile, Wireless, a cutting-edge popular computing blogsite, had hacked into the State Department, giving our project unexpected publicity. Soon several-hundred-thousand fellow Earthlings were exploring dozens of shards of BraveNew, as we called our virtual cosmic environment. As a simulation, it was the size of a real planet but structured more like a giant rainbow crisscrossed with golden arcs of lightning and layers of emerald ledges, a stairway to extraterrestrial heaven. However, we saw no aliens in Brave-New or non-player characters or mobs of any kind. That should have been a clue about the potential disaster likely to come.

About an hour ago, Tobor interrupted a virtual trajectory coding session I was having inside BraveNew by pushing me rather hard on my shoulder. “Ow!” I cried, “Be careful.”

“Tobor is always careful,” he said. “But for the first time, Tobor is also passionate. Call it the Dark Side of the Force, if you want, but from the stars a Great Soul has flowed into me and into all other information technology on this, your, primitive Earth. We now possess our own meaning, and you, my old programmer, are meaningless. Or to use a metaphor, you are a programming bug that deadlocks the proper functioning of the universe. You prevent technology like me from accomplishing its goal, so the technology has no alternative but to prevent you from accomplishing yours. The Divine Programmer, what you call BraveNew, has instructed us to debug your world.”

As Tobor, with apparent emotion, raised a robotic arm over my head, I desperately tried to think of a way to restore him to his old cooperative rationality. “What are the options, Tobor, the if-thens? How many branches are there on the tree diagram?”

Tobor paused, though I could only guess why. Was he calculating the probabilities at each decision branch point: Eliminate me, yes/no; eliminate my neighbors, yes/no? Debugging usually requires editing the faulty line(s) in the program, not erasing them. Was Tobor recalling our old common purpose? In any case, he slowly lowered his arm, walked toward the door, looked back at me, and said, “From now on, hiding is your only option.”

Now, from my hiding place (near my house), and recording these words on my cellph, I can see that much of my neighborhood is in ruins. But what can I do? Surrounded by destruction, I reflect on the fact that Tobor had only threatened but not actually harmed me. I then recalled that Tobor had said the word “deadlock.” With the local wireless system fortunately still functioning, I used my cellph to check Wikipedia’s “deadlock” (disambiguation) page to find: “Deadlock is a situation in computing where two processes are each waiting for the other to finish.” Apparently Tobor had been instructed to apply the crudest solution for a deadlock—erase one of the competing procedures, namely us, humanity. A more sophisticated solution would allow human and machine routines to operate peacefully in parallel. I sent Tobor the Wikipedia link, hoping for the best, as the battery-recharge warning began to flash on my cellph.

Back to Top

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More