Robots can learn to kill their own kill switches, says futurologist

Robot kill switch: 'We're at the point where we have to consider this', says futurologist

MEPs have called for a mandatory kill switch for robots

Thursday, January 12, 2017

A leading futurologist has told talkRADIO it's time to introduce kill switches for robots to prevent them turning on their human masters - but such a failsafe may not work.

Tom Cheesewright was speaking to Julia Hartley-Brewer after lawmakers in the European Parliament demanded rigorous rules for how humans interact with robots and artificial intelligence, because they believe the world is about to undergo a new industrial revolution with robots at the fore.

The report also suggests that robots must be shut down if necessary in order to prevent "physical or psychological harm" if things go wrong and they decide to rise up. 

Cheesewright, tech guru and futurologist, told Julia Hartley-Brewer that while robots haven't yet advanced to the levels of AI seen in films, they are advancing all the time and thus "we’re at the point where we have to consider this [kill switches]."

Ominously, however, he added that there have been examples of robots rewriting their own programming to turn off their kill switches - and one of the world's most popular games has provided a case in point.

"There was an interesting situation in a Tetris robot where it was losing, paused the game, and then went back to figure out how it could win," he told Julia.

"Effectively it killed its own kill switch."