Where Have All The Robots Gone?

16th April 2010

Watch or read any science fiction from almost any era and robots feature strongly. Films and TV from 40 or more years ago assumed that by the year 2000 there would be robots everywhere, indistinguishable from humans in many cases and doing everything for us – or trying to kill us. Unfortunately for those making predictions, but perhaps fortunately for the survival of the human race, it has proved incredibly difficult to make robots either look or behave like humans. So why is this? Why is it taking so long for reality to catch up with 1950s science fiction?

Phil Wall reports…

We’ve had robots of sorts working in factories for decades – by the end of the 1970s Fiat was claiming that its cars were ‘built by robots’ – but these are static machines programmed to do relatively simple repetitive tasks, albeit more effectively than humans. Meanwhile computers – which are robot brains – can beat Grand Masters at chess. However, chess-playing robots are really just as limited as production line machines: their sole task is to scan through millions of permutations and choose the one that their program has registered as the best option given the current state of a chess board.

It was in 1950 that the British mathematician Alan Turing published a paper that considered the question ‘Can machines think?’ This led to what has become known as the Turing Test. There are several variations on the form, but, essentially, success in a Turing Test means a machine conversing in a way that makes it indistinguishable from a human. So far no machine has got near a pass – and, of course, this covers the thinking part of the equation only; physical appearance and movement are not even addressed.

Literature and popular culture are littered with tales of machines that can think, from the ancient Greek mythology of Hephaestus making slave girls from gold, through robots in Fritz Lang’s Metropolis, the Tin Man in The Wizard of Oz, HAL the computer in 2001, right up to the ‘boy’ in Steven Spielberg’s film AI. Actually making a machine that thinks has been more of a problem. Cinema depictions of robots have perhaps fooled us into believing that the real technology is just around the corner but, in fact, robots built so far have trouble doing things that even insects manage comfortably. Although machines can easily see and hear much better than humans, for example, they have trouble interpreting the information. Navigating a room or identifying an object and picking it up are much more difficult to programme than was originally thought. And every time the robot moves, and changes the angle at which it’s viewing the world, it has to recalculate everything. Meanwhile ants can avoid obstacles and find what they want easily enough, while fruit flies, whose brains contain only about 250,000 neurons (humans have over 100 billion), can navigate easily in three dimensions, while looking for food and avoiding fast-moving rolled-up newspapers at the same time.

The recognition that insects have solved these problems has led to a ‘bottom-up’ approach from scientists, concentrating on building machines that learn from the experience of bumping into things without needing to know what they’ve hit. This research has produced household robots that will vacuum or cut grass, and can learn about obstacles, manoeuvre around them and avoid them next time. Despite sales of over three million Roomba vacuum robots, these machines are very far from any form of consciousness. Slightly bizarrely, though, owners are known to sit for hours watching their Roomba perform, in the same way that people sat and watched the first automatic washing machines go through their cycle.

Similar in form to Roomba is Rovio. This wi-fi enabled robot camera with an extending arm travels around on wheels and can be remotely operated via the internet. Its purpose is to patrol the house while the owner is out or away. It also features two-way communication, so when you’re on another continent you can still shout at the dog, or, if you’re a really bad parent, at the kids.

On a darker note, the US military spends hundreds of millions of dollars each year developing new ways of killing people, and would love to replace soldiers with machines. Already they have unmanned vehicles programmed to make choices in a limited field; soon they may be identifying targets and deciding whether to attack. So far though, these neither look like a human nor think like one. Dr Neil Smith of the Open University’s Technology faculty believes there’s little incentive to produce multi-purpose robots yet, as they perform much better in a narrow range.

The problem with reproducing human-like thought is that so much depends on ‘common sense’. We know that water is wet, that fathers are older than sons and that time doesn’t run backwards, but those things can’t be expressed mathematically and programmed. We also have a subconscious mind that takes care of the vast bulk of thinking without us even realising it. When we walk into a room, see a glass of water and have a drink we don’t have to think about the billions of calculations required to identify our target, avoid any obstacles and perform all of the movements successfully. Our conscious thought on the matter is extremely limited – “There’s my water, I’ll drink some” – and we might well be using more effort consciously thinking about an entirely different subject. Machines simply can’t compete.

In fiction, even when robots can think and move like humans they still lack something: emotion. The Tin Man was alive but he wasn’t happy just thinking; he wanted a heart – literally and metaphorically.

Can a machine ever have emotions? You may think not, but putting electrodes into someone’s head enables doctors to trigger different emotional states – which begs the question: are these emotions real? Emotions, like thought and consciousness, may exist on a sliding scale, and some scientists believe they are just an evolutionary response. Many animals kept as pets clearly have emotions of sorts, but who can say whether an iguana or a bee has enough of a conscious mind to feel emotional? When we feel loneliness, jealousy, sorrow, love or anger these might just be ways of keeping us involved in the tribe, and ultimately increasing our chances of reproduction. Artificial intelligence expert Hans Moravec believes that it may be possible to programme a robot to ‘feel’ the same kind of responses to situations, for example, to feel fear when it knows that its power is low.

RoboCup © DKriesel/dkriesel.com

Emotions or not, machines that think like humans are still a long way off. Machines indistinguishable from us – Terminator-type cyborgs – are even further away. In the meantime the most lifelike robots are along the lines of Honda’s Asimo, star of TV adverts, who can greet guests, serve drinks and play musical instruments. There is also an annual ‘RoboCup’ football tournament, the overall aim of which is to produce a team to beat the human world champions by 2050. David Kriesel, a University of Bonn graduate researcher who has taken part, says that progress is steady but slow, with expert opinions differing on whether the target for 2050 is realistic. So far, he says, “What I can see is not really human-like, that's the negative. But every year, things get a lot better, so maybe . . .”

Even if the robots do win at soccer in 2050 they’ll need a wider repertoire of thought to be worth having a drink with afterwards.

Nonetheless, machines are replacing humans in an ever-increasing number and range of jobs, even in surgical operations, where they can be more precise than any human surgeon. On a bigger scale, autonomous cars are being developed, with the ability to avoid other vehicles and decide when to brake, and unmanned spacecraft that dock themselves are regularly used to resupply the international space station. At present robots are stuck with these narrowly-defined tasks, and Bill Gates has likened robotics to the PC industry of the 1970s: it needs a breakthrough such as Gates helped to provide in computing in order to achieve a critical mass. One day humans will cease to be the most intelligent beings on earth. For all the wonders of the modern world, it hasn’t happened yet.

Find Your Local