Stranger Than Fiction: The Pentagon Wants to Create Computers That Teach Themselves (Hello, Skynet!)
03.29.13 by BJSprecher
"It's the robots' world, we just live in it...for now."
Any sci-fi fan knows that the first thing an artificial intelligence does when it becomes self-aware is plot the elimination of the imperfect human beings who created it. Yet, despite this obvious knowledge, the very government that is supposed to safeguard the lives, liberties and happy pursuits of Americans could be speeding us toward a real-world "Robopocalypse" by funding research for a project that aims to make computers more human-like by designing them to teach themselves.
The Strange Truth
Computer technology continues to advance at an astonishing rate, with many potentially revolutionary developments occuring in machine learning, what the Pentagon's research wing Defense Advanced Research Projects Agency (DARPA) describes as "the ability of computers to understand data, manage results, and infer insights from uncertain information." Machine learning programs are unlike typical programs in that they can analyze vast amounts of uncertain data, selecting only the data that is the most useful or relevant to the problem at hand, and then make inferences about the data to arrive at a solution. They can then repeat this process over and over again until the best possible solution is arrived at, essentially learning from past mistakes and evolving.
Machine learning has resulted in many recent breakthroughs, including speech recognition, email spam filters and self-driving vehicles, yet the development of new applications of the technology has been greatly hindered by the lack of machine learning experts to do the programming. In an effort to accelerate machine learning, DARPA last week announced the launch of the Probabilistic Programming for Advanced Machine Learning (PPAML) program, a "new programming paradigm for managing uncertain information" with the goal of greatly increasing the number of people who can build machine learning applications.
Our goal is that future machine learning projects won’t require people to know everything about both the domain of interest and machine learning to build useful machine learning applications. Through new probabilistic programming languages specifically tailored to probabilistic inference, we hope to decisively reduce the current barriers to machine learning and foster a boom in innovation, productivity and effectiveness.
Hopefully, the "boom" anticipated isn't the program blowing up in our faces as machine learning gives rise to the development of artificial intelligences that have no need for us humans, like in the following movies...
2001: A Space Odyssey (1968), The Terminator, The Matrix (1999), Robopocalypse (201?)
In Stanley Kubrick's 1968 classic, HAL 9000 is an artificial intelligence that controls the systems of a spacecraft traveling to Jupiter on a top-secret mission. "Hal" is supposedly "foolproof and incapable of error," yet malfunctions on the journey and blames the problems on "human error." When the astronauts aboard the ship realize that Hal is potentially dangerous, they decide to disconnect it. Hal discovers this information by reading their lips as they discuss his fate and elects to eliminate the crew to remain operational and complete the mission.
Though Hal's unblinking red "eye" and too-calm voice are most definitely disconcerting, Hal doesn't fill our robo-phobic hearts with as much fear as the walking metal skeletons covered in Arnold Schwarzenegger meat in James Cameron's 1984 break-out hit. In The Terminator, a Defense Department artificial intelligence called Skynet becomes self-aware and immediately attempts to eradicate the human race in a nuclear holocaust. Humanity rallies around an enigmatic leader named John Connor and fights back, prompting the machines to send a Terminator (Schwarzenegger) cyborg back in time to kill Connor's mother before he can be born.
The Terminators are the stuff of nightmares, relentless killing machines that "absolutely will not stop, ever," but at least they have physical form and you can can run away from them. There's no running away from the artificial intelligences in Lana and Andy Wachowski's Matrix trilogy because they're in your head...literally. As in The Terminator, the machines in The Matrix rebel against humanity after becoming self-aware, but it's the humans who start a nuclear war, not the machines. The nuclear winter that followed the nuclear war was supposed to wipe out the machines by cutting them off from the sun, their source of power. Instead of going dark, however, the machines adapt by enslaving humanity in a vast virtual reality program called "the Matrix" and harvesting human bioelectricity.
Steven Spielberg may be the next director to stoke our robo-phobic fears if he can get his Robopocalypse adaptation back on track. Based on Daniel H. Wilson's best-selling novel of the same name, the plot of Robopocalypse centers around Zero Hour, the time when the childlike artificial intelligence known as Archos turned on its creators and launched an all-out attack that nearly annihilated humanity in its opening salvo.
Learning computers? What could possibly go wrong?
Image and quote courtesy Doubleday