I know you're thinking, "WTF?!" But hear me out.
‘AI will always seek to avoid human intervention and create a situation where it can’t be stopped,’ says Rachid Guerraoui, a professor at EPFL's Distributed Programming Laboratory and the new study's co-author.
Of course, this is something almost everybody is afraid about: AIs improving so much that they become difficult to control and just circumvent humans' commands. So, researchers are finding ways to prevent that.
So, they've developed a system similar to the memory-erasing ‘neuralyzer’ from the Men in Black films. They aim to delete certain parts of AI's memory without affecting the way it learns.
‘The challenge isn’t to stop the robot, but rather to program it so that the interruption doesn’t change its learning process – and doesn’t induce it to optimize its behaviour in such a way as to avoid being stopped,’ said Guerraoui.
They've developed a system called ‘safe interruptibility’, which, as said, allows us to certain learning process of the AI when necessary without affecting the entire learning process itself.
That is, they alter the machine’s learning and reward system but doesn't affect its other processes.
‘We worked on existing algorithms and showed that safe interruptibility can work no matter how complicated the AI system is, the number of robots involved, or the type of interruption,’ Maurer said. ‘We could use it with the Terminator and still have the same results.’
Get weekly science updates in your inbox!