91av

Science fiction primes us to expect robots to run amok. But as they go mainstream, it’s still us that need policing

IT IS the year 2015. Engineers Gregory Powell and Mike Donovan, plus a robot nicknamed Speedy, are on Mercury to restart operations at an abandoned mining station. The men are field specialists for the United States Robots and Mechanical Men Corporation, employed mainly to test new or experimental robots.

Unluckily, their base’s photo-cell banks are low on selenium and the nearest source is 27 kilometres away. Since Speedy is the only one of the three that can withstand Mercury’s temperatures, Donovan sends it out to get some. But when Speedy has not returned after 5 hours, Powell and Donovan become uneasy. They finally locate the robot moving in a circle around a selenium pool, with a peculiar stagger. Asked to return, Speedy responds: “Hot dog, let’s play games. You catch me and I catch you…” and starts quoting Gilbert and Sullivan. In short, Speedy is acting like a human drunk.

Eventually, Powell realises that chemicals near the pool are destabilising the robot. Normally, Speedy would follow the rule that a robot must obey orders, but because it was expensive to manufacture, its level of adherence to another rule, that a robot must protect its own existence, was strengthened.

As the order to retrieve the selenium was casually worded, with no emphasis on returning, Speedy cannot decide whether to obey the order or protect itself from danger. Further away from the selenium, the “obey” order outweighs the need for self-preservation, while closer in the self-preservation rule is stronger. The clash causes a feedback loop which confuses Speedy to the point that it starts acting drunk.

Every attempt to get Speedy out of the loop fails as its conflicted positronic brain cannot accept new orders. All that happens is that Speedy changes its route until it finds a new place where there is an equilibrium between avoiding danger and following orders, ending up just as confused. The only thing that might trump both rules is a third rule: that a robot may not allow a human being to come to harm by its inaction. Powell risks his life by going out in the heat, hoping the no-harm-to-humans rule will overcome Speedy’s cognitive dissonance and force the robot to rescue him. Luckily, the plan works.

That is the plot of the 1942 story Runaround, in which Isaac Asimov explicitly states his Three Laws of Robotics for the first time. These are: First Law, a robot may not injure a human being or, through inaction, allow a human being to come to harm; Second Law, a robot must obey any orders given to it by humans, except where such orders would conflict with the first law; Third Law, a robot must protect its own existence as long as such protection does not conflict with the first or second law.

Real robotics is a science born out of fiction. For roboticists this is both a blessing and a curse. It is a blessing because science fiction provides inspiration, motivation and thought experiments; a curse because most people’s expectations of robots owe much more to fiction than reality. And because the reality is so prosaic, we roboticists often find ourselves having to address the question of why robotics has failed to deliver when it hasn’t, especially since it is being judged against expectations drawn from fiction.

It’s hardly surprising that the public debate on robot ethics is also confused, ranging from fantasy questions about when robots should have “rights”, to fears of robot takeovers bringing about the end of humankind, to well-founded concerns about cyborgs becoming more like robots, and serious questions about the moral dimensions of using robots to care for children or old people, and unease over military robotics.

Making sense of all this is not easy. But if real-world robots are not much smarter than washing machines, why worry? The answer is precisely because today’s robots are so dumb – more artificially stupid than artificially intelligent. But something big is happening in robotics. For the past 30 years, relatively unintelligent robots served industry well. The second wave of robotics involves mobile robots and, importantly, ones that have to work alongside humans. This next generation must be able to interact safely and reliably with people in unstructured and unpredictable environments like homes, hospitals and offices.

This was why the UK’s Engineering and Physical Sciences Research Council and Arts and Humanities Research Council convened a “summit” last year to discuss social, legal and ethical issues in robotics. The idea was to produce a draft document to stimulate debate among roboticists and the public.

In drafting it, we took Asimov’s Three Laws of Robotics as our starting point because they have cultural significance. Our big departure was in deciding that it was the ethics of the roboticists, not the robots, that we needed to worry about. After all, robots are not people: there’s little point worrying about how a robot can be ethical, moral or responsible, no matter how interesting this might be. We also decided that we would only concern ourselves with existing robots and those five to 10 years away from being built, not the super-smart robots of some imagined future.

So how do Asimov’s laws look, updated for 2011? Instead of Three Laws of Robotics, we have devised a Five Ethics for Roboticists, which are aimed at researchers, designers, manufacturers, suppliers and maintainers of robots. These are: First Ethic, robots are multi-use tools, and should not be designed solely or primarily to kill or harm humans, except in the interests of national security; Second Ethic, humans, not robots, are responsible agents, so robots should be designed and operated as far as is practicable to comply with existing laws and fundamental human rights and freedoms, including privacy; Third Ethic, robots are products which should be designed using processes which assure their safety and security; Fourth Ethic, robots are manufactured artefacts, so they should not be designed in a deceptive way to exploit vulnerable users (their machine nature should be transparent); Fifth Ethic, it should always be possible to find out who is legally responsibility for a robot.

“Robots aren’t people: there’s little point worrying about how a robot can be ethical”

Importantly, these “ethics” downplay the specialness of robots, treating them as tools and products to be designed and operated within legal and technical standards. The fourth ethic is the only one to explicitly address a quality of some robots: that of inducing emotional bonds or dependencies in humans. We think that designers or manufacturers should not exploit such dependencies, and that it should always be possible, as Toto did in The Wizard of Oz, to pull the curtain aside to expose the robot behind.

But despite our attempts to be practical, our ethical principles continue to be the object of controversy and dissent, both among summit delegates and those who have read the draft document. For example, should the first ethic be qualified by “except in the interests of national security”? I would say no, but excluding such a rider would instantly make these ethics irrelevant to the huge military robotics industry.

There are also criticisms that we have oversimplified a complex field, reducing it to sound bites, and some researchers point out that the Five Ethics raise more questions than they answer. They are right, but then, this is exactly the kind of debate we want to foster.

While our initiative is new, serious work on roboethics – a term coined in 2004 by Gianmarco Veruggio, lead author of the European Robotics Research Network Roboethics Roadmap – is not. Several countries have drafted ethical or standards frameworks for robotics, notably South Korea and Japan, and there is a growing literature on autonomous robot weapons. However, much of it focuses on the safety, accountability and even morality of robots’ behaviour. Our initiative focuses more on the here-and-now responsibilities of roboticists, and we believe that is what makes it important.