As Google’s AlphaGo super program takes out the world’s leading Go players, tech innovators like Elon Musk are growing more and more weary of a robot takeover. In response, Google is working on a robotic kill switch to ensure humans maintain the regime.
Google’s Artificial Intelligence (AI) supercomputer DeepMind’s AlphaGo program is now a force to be reckoned with in the Chinese game of Go. After beating one of the best players in the world, humans everywhere are seeing the realization of a multi-generational fear: robot takeover. Thankfully, Google is also developing a kill switch to control robots in case AI gets too crazy, or gets too good at Go.
DeepMind made headlines in March for beating Korean grandmaster Lee Sedol. Sedol ranks as a ninth-dan Go player and is recognized as one of the best players in the world at the game, which is considered one of the most challenging strategy games in existence. After winning 4-1 in a best-out-of-five match, AlphaGo earned an honorary ninth-dan ranking as well, and has raised some serious questions about the potential for AI.
At the MIT’s Aeronautics and Astronautics 2014 Centennial Symposium, business magnate Elon Musk said the development of artificial intelligence is the single most dangerous extraterrestrial technology. He urged tech leaders to better regulate the advancement of AI, and for good reason. If the algorithms and functions that govern robots stop listening to humans, AI has the potential to outthink us all. That’s why Google is developing a robotic kill switch, just in case a robot uprising is a little more serious than science fiction.
The concept is called safe interruptability and functions as an algorithm embedded into AI and robotics programming. These functions would allow new human directives to become part of a robot’s normal tasks to essentially force robots to listen to directives than are in the benefit of the robot, such as a directive not to walk outside during a storm, even if the machine’s daily tasks require such movement. But it has limitations.
The safety net might not work on machine learning programs, as these programs are meant to continue to develop and respond based on experiences (much like humans). In essence, AlphaGo might not readily accept a directive to stop being so good at Go. In this case, it would be worthwhile to create a physical kill switch that ensures man remains on top, at least in some realms.
AlphaGo has a tentative match scheduled with the number one Go player in the world: 18-year-old Chinese World Champaign Ke Jie. Jie expressed some anxiety about AlphaGo being able to learn from his tactics, but Google intends to make him stand behind the trash talking he’s done during the super program’s matches earlier this year.
Our advice to Jie: if AlphaGo gets too good, just kill it.
Have a story tip? Message me at:
