RoboCop, that tin-suited keeper of law, order and a heroic portmanteau, abides by three prime directives: 1) Serve the public trust, 2) Protect the innocent, 3) Uphold the law. He lives by these rules with algorithmic devotion. As well he must: each is written into his circuitry. Not only that, his existence is dependent upon the absence of error. A misstep by an American beat cop, as recent events have proved, may only result in gardening leave or suspension (at least, if the indiscretion was filmed by some dauntless passer-by). But should a cyborg officer so much as erroneously issue a speeding ticket, he and his electronic colleagues would surely be summarily melted down and their metal used to make candlesticks.
Likewise, there were 1,713 fatalities from reported road traffic accidents in Great Britain in 2013, most of which arose from human error. We view this as an unfortunate but perhaps inevitable tax on our freedom to drive. But will we be so lenient when news hits of the first fatality in a collision involving an automated car? No: there will be an apocalyptic debate on the idiocy of putting robots behind the wheel in the first place. Similarly, the first fully automated drone to fire on western civilians will be shot from the sky with the force of 10,000 remorseful politiciansâ€™ speeches. Humans and robots, in other words, are not created equal. When it comes to artificial intelligence, we are less lenient to it than we are towards holders of organic intelligence.
Our underlying distrust of intelligent machines has long been reflected in fiction. Thereâ€™s the treacherous HAL 9000 from 2001: A Space Odyssey, with his unblinking, devil-red eye, who tries to murder the crew of his spaceship and make it look like an accident. (Whatâ€™s worse than a killer? One who tries to slyly cover his tracks.) Thereâ€™s the blinking mainframe computer of 1970â€™s Colossus: The Forbin Project, which leads the world to the brink of nuclear destruction (see also: WarGames). There are the Stepford Wives.
Anxiety is even baked into the word â€œrobotâ€ itself â€“ which derives from science fiction, not science. Karel ÄŒapekâ€™s 1921 science-fiction play, R.U.R., which is credited with introducing the word to the English language, depicts a cyborg labour force that rebels against its human masters, leading to the extinction of the human race. The Czech origin word â€œrobotaâ€ means â€œforced labourâ€, and is derived from the word â€œrabâ€, meaning â€œslaveâ€. The slave-masterâ€™s fear of revolt is ancient, and that same angst lurks inside our every utterance of the word â€œrobotâ€.
In imagining what might go wrong if humanityâ€™s inventions were to go rogue, fiction performs an essential role, testing the implications of our creations, real and hypothetical. So it is no accident that today, as the wilder possibilities of AI begin to seem like a medium-term prospect rather than some comfortingly remote futurology, the number of fictional works exploring malevolent AIs is at an all-time high. Avengers: Age of Ultron (2015) tells the story of a sentient AI that wants to eradicate humanity in order to save Earth. Disneyâ€™s Big Hero 6 (2014) features a robot that turns murderous if its â€œhealthcare chipâ€ is removed. Chappie (2015) examines the question of whether robots should be hardcoded with morals or, like us, only given the capacity to learn them. Alex Garlandâ€™s Ex Machina (2015) sees a programming student fall in love with an AI and attempt to free her from a lab. In Her (2013) another man falls in love with an artificial intelligence; she eventually â€œleavesâ€ him, as sheâ€™s evolved beyond the relationship (a different kind of attack on humanity). Charlie Brookerâ€™s TV series Black Mirror tests various potential near-future outcomes of our relationship with technology, while Humans, which debuts on Channel 4 on Sunday night, examines the repercussions of artificial brains through the lens of domesticity.
This boom in AI-based plots is surely tied to the closing gap between this aspect of science fiction of the 20th century and science fact. Humans, tellingly, is set in a parallel present day â€“ not a move writers could credibly make in a story about, say, intergalactic travel. The public is keenly aware of the great strides that have been made in human-simulating artificial intelligence, if only through the quiet miracles of the anthropomorphic assistants who live in our mobile phones: Appleâ€™s Siri and Microsoftâ€™s Cortana (who took her name from fiction: the benevolent AI companion you play alongside in the Halo video game series). Then there are the revelations from industry: Amazonâ€™s claim (controversial as it is) that it has a fleet of delivery drones ready to fly, just as soon as the air regulators allow it; the news that the AI company DeepMind, which seeks to â€œsolve intelligenceâ€ by creating a digital human brain, has been bought by Google for a reported Â£400m; those distressing YouTube videos of military contractor Boston Dynamicsâ€™ giant, sprinting robot dogs, which steady themselves when given a sharp kick by some researcher.
Not only that, but public figures have upped the ante in terms of existential dread surrounding the subject. Stephen Hawking recently warned that advanced AI â€œcould spell the end of the human raceâ€. Elon Musk, the entrepreneur, claimed that artificial intelligence is the greatest existential threat to mankind. (â€œHope weâ€™re not just the biological boot loader for digital superintelligence. Unfortunately, that is increasingly probableâ€, he tweeted, before donating $10m to the Future of Life Institute, which works â€œto mitigate existential risks facing humanityâ€). Bill Gates added weight to the claim: â€œI agree with Elon Musk and some others on this and donâ€™t understand why some people are not concerned,â€ he wrote during a Q&A session on Reddit. German philosopher Thomas Metzinger has argued that the prospect of increasing the amount of suffering in the world is so morally awful that we should cease building artificially intelligent robots immediately.
Recent books with titles such as Our Final Invention: Artificial Intelligence and the End of the Human Era only intensify the dread. Nick Bostrom, a philosopher who directs the Future of Humanity Institute at the University of Oxford, posited a particularly frightening endgame for the development of AI in his recent book Superintelligence. He imagines a machine programmed to make as many paperclips as possible, and describes the intelligent machine devising a plan to clone itself in order to become more efficient in its one goal. As it continues its noble work of turning everything into paperclips, it decides, quite logically, to eradicate anything that is not a paperclip, including humans, in order to create more space. The story is a metaphor, but Bostromâ€™s aim is to demonstrate how programming even simple values into intelligent machines could have catastrophic outcomes.
Like all good end-times prophecies, thereâ€™s even an approaching date for the robot apocalypse. Ray Kurzweil, the futurist and director of engineering at Google, has set the date at which machine intelligence will exceed human intelligence as 2045. This point, known as the singularity (a term coined by science fiction writer Vernor Vinge), is for some a goal to be aimed at (the 2045 Initiative, founded by Russian entrepreneur Dmitry Itskov in February 2011, aims to be able to transfer a human mind to a â€œnon-biological carrierâ€ by the date), and for others an outcome to be avoided at any costs.
And for many others, the concerns in both fiction and philosophy about the end of our species at the hands of one of our creations are entirely overstated. Artificial intelligence tends to be developed along highly specialised lines. IBM may have created a computer program that can beat a grandmaster in a game of chess, but ask Deep Blue to play a game of noughts and crosses and it will be at a loss. Most AIs are able to do one thing incredibly well, and nothing else. It is intelligence along a single axis. Likewise, we have not yet produced machines with vision, natural language processing or common sense. Ask Siri a question and she will, most of the time, simply type the question into Google for you. Weâ€™re a long way from the realm of teary break-ups.
And yet, AI along a single axis still raises troubling and pressing questions. Potentially autonomous weapons already exist in the world (though most current models have safeguards that require a human to grant permission to fire). This military hardware is capable of identifying, tracking and firing upon a moving target from a great distance, theoretically without human intervention. The call from Human Rights Watch for an outright ban on â€œthe development, production, and use of fully autonomous weaponsâ€ has come too late. They already exist, and our thinking and laws havenâ€™t caught up. Even fiction has been slow to ask what that might mean for us. What international laws govern the deployment of fully autonomous weapons? What recourse will there be for victims of a deadly error? These questions of how to govern robot morality (Serve the public trust? Protect the innocent? Uphold the law?) no longer belong in the halls of futurology. Tellingly, though, fiction has failed to interrogate this kind of existential threat, presumably because a resolutely machine intelligence lacks dramatic potential. We can impress motive, character, even wit on to robot servants who look like us. An autonomous gun is less relatable.
The rise of the robots might not only present a physical threat to humans. As more of our work becomes automated, there is the threat to labour. In the 1950s the sci-fi dreamers first began to envision a future in which machines fully replaced manâ€™s work. Automated flying cars would deliver 21st-century humans to their pristine destinations, where robot chefs would prepare our meals then sweep up after us. We would be made redundant in the best possible sense: able to enjoy a life without toil. A recent paper from the Oxford Martin Programme on the Impacts of Future Technology claims that a less inviting version of this future is close: within 20 years, computers will be able to replace humans in 47% of current US jobs. This would be catastrophic for the labour force; Ukip will have a new immigrant force to round upon. Humans embraces the vision: its robots, who are indistinguishable in appearance from humans, are used as factory workers, cleaners, carers and ticket inspectors.
But science fiction has mostly chosen to focus on the possibility of a robot uprising rather than a robot supplanting. The enduring potency of this plot is rooted in the human anxiety about being fooled by our own technology, of losing control of our creations. The use of tools (including intelligent machines) defines our species, and separates us from most other living things. We know on an instinctual, elemental level that tools are something of which to be proud. And yet this pride is now coupled with fear: of being duped or let down by our technology, or in the worst case scenario, of being made to serve a version of that which, for millennia, has served us.
It is disturbing that most of our thinking about the implications of machine intelligence has, to date, been done in the realm of fiction: thought-experiments carried out in the name of entertainment. The writers and storytellers have done their job for close to century, posing the looming questions. Now, where are the philosophers, the engineers, the sociologists and the economists who might provide some answers?
Humans is on Channel 4, Sunday, 9pm