Google and Oxford University scientists are developing an override switch for artificially intelligent machines.
(CCM) — Scientists from Google's artificial intelligence division have teamed up with Oxford University to create a kill switch for robots. According to a report by the BBC, the team outlined in an academic paper their plans to code future AIs to prevent the machines from developing the ability to override human input. With the introduction of a kill switch, the scientists hope to assuage worries that, at some point, robots will become so intelligent that their human supervisors will lose control of them. The scientists' plans rely on a system of teaching by reinforcement wherein the machines may have their actions corrected by their human overseers without the AIs ever learning how to avoid said intervention. A scientist at Google's DeepMind, Dr. Laurent Orseau, commented, "It is important to start working on AI safety before any problem arises. AI safety is about making sure learning algorithms work the way we want them to work."
Dr. Orseau also mentioned the desire to develop a system that can predict potential bad behavior in AIs in order to avoid it entirely. This would have come in handy, for example, in the case of Microsoft's Tay, the AI that began tweeting racist and bigoted messages after being provoked by internet trolls. The company introduced Tay to the public in late-May but had to destroy the AI immediately due to controversy.
Photo: © tanuha2001 - Shutterstock.com