We hear a lot about computer crime. Ingenious hackers develop more and more ways to fleece us. Turns out there is another side; the criminals can actually be the computers!
We all know how clever computers are. With their super speed they have all the time in the world, compared to our turtle-speed brains. Many years ago a computer program was used to design new radio antennas. It came up with some really weird configurations, things we would never have imagined. They all worked.
Computer Beats Champ
Recently, Alpha-Go soundly beat Lee Sedol, the world champion Go champion. It did this by combining a number of Artificial Intelligence methods, plus splitting the machine in two so it played against itself time after time after time. Alpha-Go developed some truly weird opening moves, never before seen.
In these examples the computers behaved exactly as they were supposed to. But it is also possible for computers to seriously misbehave. They will do exactly what they are told to do, but will sometimes develop their own bizarre ways to do it. Ways that could only be called criminal.
A long time ago a tic-tac-toe computer pitted against another computer figured out that making improbable moves caused its opponent to crash. It followed all the rules, with a totally unexpected and undesirable result.
When experimental computers go rogue (some would call it cheating) it is usually not reported. Swept under the rug, like a mentally ill relative.
However a few AI researchers study these bugs to find the roots of algorithmic impishness. “We don’t want to wait until these things start to appear in the real world,” says Victoria Krakovna, a research scientist at Google’s DeepMind. She keeps a list of AI bugs which includes more than three dozen incidents of miscreant algorithms. Here are some;
In a survival simulation, one AI species evolved to subsist on a diet of its own children.
Algorithms exploited flaws in the rules of the galactic video game Elite Dangerous to invent powerful new weapons.
A four-legged virtual robot was challenged to walk smoothly by balancing a ball on its back. Instead, it trapped the ball in a leg joint, then lurched along as before.
The specimens collected by Krakovna and fellow bug hunters point to a communication problem between humans and machines: Given a clear goal, an algorithm can master complex tasks, such as beating a world champion at Go. But even with built-in constraints, it turns out that mathematical optimization empowers programs to develop shortcuts humans didn’t think of. Teach a learning algorithm to fish, and it might just drain the lake.
Gaming simulations are fertile ground for bug hunting. Earlier this year, researchers at the University of Freiburg in Germany challenged a bot to score big in the Atari game Q*bert. Instead of playing through the levels like a sweaty-palmed human, it invented a complicated move to trigger a flaw in the game, unlocking a shower of ill-gotten points.
“Today’s algorithms do what you say, not what you meant,” says Catherine Olsson, a researcher at Google who has contributed to Krakovna’s list and keeps her own private zoo of AI bugs.
As AI systems become more powerful and pervasive, hacks could cause massive problems. If a neural network managing an electric grid were told to save energy—Google’s DeepMind has considered just such an idea—it could cause a blackout.
“These systems are creative and can do things you never thought of. It’s important to understand their power and danger,” says Jeff Clune, a researcher at Uber’s AI lab. A recent paper that Clune coauthored, which lists 27 examples of algorithms doing unintended things, suggests future engineers will have to collaborate with, not command, their creations. “Your job is to coach the system,” he says.
What are your ideas on this subject? Do think that we can outsmart our incredibly powerful creation, Dr. Frankenstein?
Paraphrased from Wired Magazine
Regardless of what’s going on, you have control. Much of the control is your focus, thinking and mindset. Discover how to control these things with my FREE audio meditation files. http://truefortunes.com/audios