Skip to main content current edition: International edition The Guardian - Back to home Become a supporter Subscribe Find a job Jobs Sign in Search Show More Close with google sign in become a supporter subscribe search find a job dating more from the guardian: change edition: edition International edition The Guardian - Back to home browse all sections close Artificial intelligence (AI) Opposition to autonomous warfare swells to 16,000 signatories Artificial intelligence community comes together in unprecedented numbers to call for a ban on AI-controlled weaponry GLaDOS, the malicious AI from the Portal series of video games.
 Image: Valve Photograph: Valve Artificial intelligence (AI) Opposition to autonomous warfare swells to 16,000 signatories Artificial intelligence community comes together in unprecedented numbers to call for a ban on AI-controlled weaponry Alex Hern @alexhern Thu 30 Jul ‘15 13.
59 BST Last modified on Wed 22 Feb ‘17 18.
08 GMT This article is 2 years old An open letter from AI researchers urging a ban on offensive autonomous weapons has now reached 16,000 signatories, after being signed by more than 15,000 people in the three days since it was released.
 The letter says “AI technology has reached a point where the deployment within years” and was initially signed by Tesla’s Elon Musk, Apple co-founder Steve Wozniak, Google DeepMind chief executive Demis Hassabis and professor Stephen Hawking.
 It has now been signed by over 2,000 experts, as well as another 14,000 individuals from outside the AI community.
 It warns: “The stakes are high: autonomous weapons have been described as the third revolution in warfare, after gunpowder and nuclear arms … The endpoint of this technological trajectory is obvious: autonomous weapons will become the Kalashnikovs of tomorrow.
 The key question for humanity today is whether to start a global AI arms race or to prevent it from starting.
” The letter was launched in Buenos Aires earlier this week, and Toby Walsh, one of the researchers involved in putting it together, says that the momentum behind it has gathered entirely through word of mouth.
 But despite the huge support, Walsh emphasised that the position is by no means universal amongst the AI community.
 “These issues are not black and white.
 There are arguments for both sides here.
 It’s just my opinion, and that of thousands of my colleagues, that the balance of arguments favours a ban before we end up in this next arms race.
” In fact, Walsh concedes that “there are several arguments against a ban”.
 He says: “For instance, such robots will be more precise and able to minimise civilian casualties.
 Another is that we should not risk human lives when robots can take their place.
 However, I find these carry less weight than the fear of an arms race and of these weapons falling into the hands of terrorist and other organisations.
” But he argues that the impetus for a ban is not merely hypothetical, and that there are robots available today “that have the capabilities needed to build such autonomous weapons”.
 “One such example (there are others) is the Google autonomous car: it can be given a high level goal (‘take me to the office’), make a plan for how to achieve that goal, sense where it is using computer vision and radar, start executing that plan, identify when the plan breaks (perhaps there are roadworks), find a new plan and start executing that, take evasive action when a car cuts them in, and finally identify when it has reached some goal.
 “These are essentially the capabilities you need to build an autonomous weapon system that can find, locate, track and target.
 Of course, terrorists don’t have the means or capabilities today to build a Google-like car.
 But it won’t be long before the technology is smaller, cheaper and better.
 That’s always the case with computing.
 the killer robots that I fear most are small ones.
 Imagine swarms of tens or hundreds of robots.
 It would be hard to defend against such an opponent.
 And small robots will be cheap and easy to replicate.
” Of course, even if AIs don’t get the explicit go-ahead to murder, that doesn’t mean that there won’t be ethical challenges ahead.
 “We must not forget the many other challenging and important ethical decisions in AI,” says Walsh.
 “To go back to the autonomous cars, what happens when such a car needs to make a life or death decision when facing a car coming towards it on the wrong side of the road?
” At a UN conference in Geneva in April discussing the future of weaponry, including so-called killer robots, the UK opposed a ban on the development of autonomous weapons, despite calls from various pressure groups, including the Campaign to Stop Killer Robots.
 Topics Loading comments… Trouble loading?
 most viewed The Guardian back to top all sections close back to top All rights reserved.
 cience%2CDeepMind]