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) The Observer Robot panic peaked in 2015 – so where will AI go next?
 This year experts from Elon Musk to Stephen Hawking warned about the havoc robots could cause the economy and humanity.
 How do we ensure machines are friends rather than foes?
 Robots at Tesla in 2013.
 Tesla owner Elon Musk has since warned that AI is an existential threat to mankind.
 Photograph: The Washington Post/Getty Images Artificial intelligence (AI) The Observer Robot panic peaked in 2015 – so where will AI go next?
 This year experts from Elon Musk to Stephen Hawking warned about the havoc robots could cause the economy and humanity.
 How do we ensure machines are friends rather than foes?
 Charles Arthur Charles Arthur @charlesarthur Sun 27 Dec ‘15 09.
00 GMT Last modified on Sat 2 Dec ‘17 15.
49 GMT Ever since IBM’s Deep Blue defeated then world chess champion Garry Kasparov in a six-game contest in May 1997, humanity has been looking over its shoulder as computers have been running up the inside rail.
 What task that we thought was our exclusive preserve will they conquer next?
 What jobs will they take?
 And what jobs will be left for humans when they do?
 The pessimistic case was partly set out in the Channel 4 series Humans, about a near-future world where intelligent, human-like robots would do routine work, or stand on streets handing out flyers, while some people worked (law and policing seemed to get a pass, mostly) but others were displaced – and angry.
 In May, Martin Ford, author of Rise of the Robots: Technology and the Threat of Mass Unemployment, described the concern for both white- and blue-collar workers as that Humans-style world approaches: “Try to imagine a new industry that doesn’t exist today that will create millions of new jobs.
 It’s hard to do.
” But there is an optimistic view of the same process: that the pairing of computers and robots will free humans from drudgery and dangerous work; and free people to use their imaginations and interact with each other in more personal ways, and especially in ways that computers work side by side with robots, software agents and other machines,” said JP Gownder, lead author of a report called The Future of Jobs, 2025: Working Side by Side with Robots, produced for the research company Forrester in August.
 Gownder pointed out – as many have – that throughout history, automation and technology have repeatedly created more jobs overall than they have destroyed.
 We don’t have lamplighters any more, but we have huge industries built around street lights and electricity supply.
 But in the robotic world, will the new jobs be better jobs?
 In his book, published in January, The Glass Cage: Where Automation is Taking Us, the writer Nicholas Carr argues that computers are taking over too much from us – or rather, that we’re too willing to give up charge of things to machines – and that jobs are becoming deskilled as a result.
 Steve Wozniak, co-founder of Apple, says computers will 'get rid of the slow humans to run companies more efficiently' Carr points to the origins of automation, after the second world war, when the Ford Motor Company was installing new machinery to do “automatic business” on assembly lines.
 “Control over a complex industrial process had shifted from worker to machine,” he notes.
 He points out that automation had already raised its head during the second world war, through the need to get hard-to-manoeuvre anti-aircraft guns to shoot down bombers by letting machinery move their aim, according to targets picked by gunners from screens.
 The humans’ task was made easier, but it was abstracted from the process and outcome.
 What is indisputable is that robots equipped with computer vision and paired with artificial intelligence (AI) systems – often called “machine learning”, or “deep learning”, or “neural network” systems – will take over more of the work that humans do today.
 Foxconn is one of the world’s biggest manufacturers of electronics, with giant factories in China which assemble phones, tablets and computers for Apple and other companies.
 It is working on robot-driven factories which will inevitably mean fewer of those jobs for humans.
 The Korean electronics giant Samsung, meanwhile, has been given a grant by the Korean government to develop high-precision robots to take over the work now done by humans, also in China, where rising wages are squeezing profit margins.
 Which, of course, leads to the question: what new jobs will those displaced factory workers go on to do?
 Nobody knows; yet everyone is sure, despite Martin Ford’s fears, that they must exist.
 Pepper the concierge robot Facebook Twitter Pinterest “Pepper” the concierge, who greets customers in the Mizuho bank in Tokyo.
 Photograph: Yuya Shino/Reuters Yet as we head towards that future, there are also ethical and legal reefs to navigate.
 Isaac Asimov introduced his famous Three Laws of Robotics for Runaround, a science fiction story set in 2015.
 In July, an article appeared in the science journal Nature, pointing out that “working out how to build ethical robots is one of the thorniest challenges in artificial intelligence”.
 That month, a 22-year-old worker installing a robot at a VW plant in Germany was killed when it was wrongly activated.
 Clearly, Asimov’s laws haven’t arrived yet.
 But robots that kill – especially “intelligent” ones – are very much on the mind of those who worry most publicly about the AI-robot combination.
 Stephen Hawking told the BBC it “could spell the end of the human race” as it took off on its own and redesigned itself at an ever-increasing rate.
 Elon Musk, the billionaire who brought us PayPal and the Tesla car, called AI “our biggest existential threat”.
 Steve Wozniak, the co-founder of Apple, told the Australian Financial Review in March that “computers are going to take over from humans, no question” and that he now agreed with Hawking and Musk that “the future faster than us and they’ll get rid of the slow humans to run companies more efficiently”.
 Nick Bostrom may not have a similar claim to fame, but he is an Oxford University philosopher who argues in his book Superintelligence that self-improving AI could enslave or kill humans if it wanted to, and that controlling such machines could be impossible.
 But there’s no sign so far of inherently intelligent killer robots, or “anthropogenic AI”, as it’s also called.
 Reviewing Bostrom’s book, the scientist Edward Moore Geist suggested that it “is propounding a solution that will not work to a problem that probably does not exist”.
 According to Murray Shanahan, professor of cognitive robotics at Imperial College London, “properly general intelligence” is comparatively easy to describe but hard to enact: “the hallmark of properly general intelligence is the ability to adapt an existing behavioural repertoire to new challenges, and to do so without recourse to trial and error or to training by a third party,” he writes in his book The Technological Singularity.
 But to do that requires two capacities that AI tends not to display: common sense and creativity.
 On common sense, Shanahan gives the example of finding the people who normally work inside a building instead standing outside it in the rain.
 “What are you doing?
” might prompt the answer “Standing outside” from a computer, whereas a human would respond “Fire alarm” – recognising the common understanding that exists between the speakers.
 Creativity, meanwhile, can be demonstrated by animals in problem solving, as well as by humans, such as a crow which bent straight wires to create hooks to get food.
 But it’s hard to say that computers have ever shown it.
 It might be that they will – and to that end, Musk, with the backing of Loopt entrepreneur Sam Altman, has poured $1bn into a new not-for-profit organisation, OpenAI.
org, which aims to create an open-sourced AI that surpasses human intelligence but whose products are “usable by everyone, instead of by, say, just Google”.
 Our real problem, though, seems to be that the growth in computing power – which roughly doubles every 18 months, but grows geometrically because we have so many more connected devices now – is outstripping our ability to reframe our ethical and legal approach to computers’ decisions.
 Even a technology that sounds innocuous and helpful, such as self-driving cars, isn’t immune from ethical and legal questions.
 For instance: if such a car needs to brake abruptly to save those on board, is there any responsibility towards people in cars behind?
 If someone shunts a self-driving car with nobody at the wheel into a third car, who is responsible for the damage to the third car?
 The self-driving car’s owner?
 Its programmer?
 And the ways in which computers solve “human” problems repeatedly turn out very unlike the methods humans use.
 Take chess: studies have found that the best human players look at a narrow set of moves, which they explore in depth, “pruning” among alternatives to find the best sequence.
 Computers, by contrast, look at every possible move, and essentially use brute force to pick the best at any time; they can’t decide that a particular move will surprise or upset an opponent, or choose a tricky one because the other player is short on time to decide.
 Compared with humans, chess-playing computers have no subtlety, except by accident.
 And in June, the US Defense Advanced Research Projects Agency held a competition for self-propelled robots which could work where humans cannot – say, to go into nuclear reactors and shut down operations.
 The winners took away millions; but the “blooper reel” of tumbling, stumbling, staggering robots has had nearly half a million views on YouTube.
 Sometimes, we like robots to be fools.
 Facebook Twitter Pinterest Watch: robots come a cropper at the US Defense Advanced Research Projects Agency.
 Nicholas Carr wrote in the New York Times in May that while it might feel as though the best way to remove error from any system is to remove the humans – because they’re the ones we hear about who opened the wrong spigot, or turned off the wrong engine on the jet – in fact, humans repeatedly perform “feats of perception and skill that lie beyond the capacity of the sharpest computers”.
 For example: Google’s self-driving cars have been hit 11 times in 1.
7m miles of travel by dozy humans, while causing no accidents directly themselves.
 But the humans inside them have to stay alert and at the wheel, because the software has a glitch every 300 miles’ driving or so and it hands control back to the “driver”.
 What hasn’t yet been figured out is how much warning the human needs to take over.
 Is it 20 seconds?
 10?
 One?
 Is it the same for everyone?
 What if the “driver” falls asleep because the rest of the journey has been so boring, but there’s a crash when the computer was fully in charge: is that their fault, or the computer’s, or the programmers’?
 It’s enough to make a lawyer cry with delight.
 Jobs that require careful human-to-human contact – such as hairdresser or surgeon – should survive the robot insurgency But there’s no doubt we have to face up to the social changes that are coming our way.
 In November the Bank of America published a lengthy report which concluded that the “rise of the intelligent machines” constituted “the next industrial revolution”, with AI-driven robots “becoming an integral part of our lives as providers of labour, mobility, safety, convenience and entertainment”.
 Sales of robots grew by 29% in 2014, with North America seeing the third consecutive year of record sales.
 Potential long-term effects include the replacement of existing jobs by automation (47% of jobs in the US could be automated, the Bank of America calculated) and the growth of inequality – as skilled workers are increasingly in demand, while unskilled ones are not.
 Yet it’s not necessarily the low-paid jobs that will be affected – nor the high-paid ones that will be safe.
 The World Economic Forum published a graphic in November as part of an analysis into robots and jobs which suggests that chief executives’ jobs are probably safe – but so are those of landscaping and groundskeeping workers, despite an order of magnitude difference in their hourly payments.
 The emerging consensus, such as it is, seems to be that jobs requiring careful human-to-human contact – hairdresser, surgeon and so on – should be safest from the robot insurgency.
 What’s most likely is that “work” will grow in complexity as AI-based systems take over the simpler tasks.
 “Computer” used to be a job title for humans who did calculations; now their entire function can be replicated by a cell in a spreadsheet.
 Yet jobs still exist.
 Losing at chess hasn’t made us stop playing chess, either.
 Kasparov himself has run championships of “centaur chess” – humans playing with the direct aid of computers during the game, which has turned out sometimes to lift the humans’ chess rating above both their own and that of the computer program.
 And if it can happen in chess, why not work?
 Topics Loading comments… Trouble loading?
 most viewed The Guardian back to top all sections close back to top All rights reserved.
 ture]