Electric Dreams

Chapter 1: Sex robots as science fiction

In the late 1990s, real robots and their human creators burst into the British public's consciousness, thanks to a show filmed in a Docklands warehouse on a budget of about three pounds. Robot Wars, BBC 2's cult classic 'robot combat competition' pitted amateur robotics engineers-or rather, their remote-controlled inventions-against each other in battle. The theatre of conflict was a flame-laden, obstacle-heavy arena policed by the 'House Robots'; professionally-designed machines with names like Sir Killalot, Mr. Psycho and Sergeant Bash. Watching the footage now, it's impossible not to develop extreme fondness for the amateur roboticists, usually men between 18 and 50 with names like Pete and Steve and Alan, sometimes with their partners, sometimes, their kids. The contestants have little to no stage presence and take their hobby extremely seriously; very often Craig Charles, a presenter of enormous charisma, has a hard time pulling more than single word answers out of them. But they are highly affable and thrilled to be there, and all of them have spent months designing and building robots with names like Behemoth and Napalm and Angel of Death, only to smash them to bits or see them set on fire within a matter of minutes, sometimes seconds.

There are precious few 'walkerbots'-robots with legs-on any series. In fact, the most indomitable robots were ones like Razer, a wedge with a crushing scorpion's tail; Chaos 2, a dual wedge with a powerful pneumatic flipping arm; and Hypno-Disc, an armoured rectangle with a wildly destructive horizontal flywheel on the front. Even the House Robots, built by professionals, relied on speed, weight, crushing ability or some sort of weapon; pincers, hammers or pneumatic tusks. They were hefty.

Not a single one of the robots on Robot Wars had anything humanoid about them; twenty-five years later, the same is true of the robots we now live and work with on a regular basis. Robots are everywhere; they are in factories, shifting heavy pallets; they are huge mechanical arms with, at most, six degrees of articulation, painting cars; they are featureless disks hoovering your home. As the University of Edinburgh's Professor Adam Stokes puts it, 'robots are good at dangerous, dull and dirty jobs.'3 And precisely none of these jobs currently require them to look human.

It's not just that humanoid appearance isn't necessary; it's that the human shape, broadly speaking, is completely inefficient. The amateur engineers of Robot Wars knew it, and those working in the field decades later know it too. Humans can be easily knocked over; we often trip and fall. We require an enormous amount of energy to do anything. All of our vital, fleshy, vulnerable major organs are right in our highly stabbable torsos, with only a few layers of skin, fat, muscle and sometimes bone between them and a deadly implement. The mechanism of the human body isn't a sensible one; as Simon Watson, Lecturer in Robotics at the University of Manchester puts it,

We like to think that we're the dominant creatures on the planet, so mobile robots should look like us. But the fact is, they shouldn't. We can't fly, we're not very good swimmers, we can't live in a vacuum and if we want to travel more than a mile, most of us will get on some type of wheeled vehicles. Bipedal locomotion has served us well but it is limited and requires a huge amount of brain power and years of learning to perfect.. After nearly 100 years of development, our most advanced humanoid robots can only just open a door without falling over.4

The last sentence here is not an exaggeration; in 2022 the news of a robot opening a door was splashed across media outlets with great fanfare. Despite our inefficiency, vulnerability and illogical design, the human body is a wildly, nonsensically capable object. When you break down what is required of a machine to reproduce even the most basic human tasks-like getting up off the floor, picking something off a high shelf, or, indeed, opening a door-you start to realise why people believe in a God. Not to be arrogant, but it is simply mindboggling what evolution has managed to create. We float about the world doing our jobs and hugging our loved ones and operating heavy machinery with barely a thought given to the complex, intricate and numerous systems at work. Take the human hand. Our hands have twenty-six degrees of freedom; to recreate this mechanically, every one degree would require individual control. We do it without even noticing. Neuroscientists estimate that humans have between twenty-one and thirty-three distinct senses, many working in tandem to ensure we can move about the world as we currently do. Without even one or two of them, we could be completely incapacitated in ways we can barely comprehend.

A favourite example is that of proprioception, which allows us to instantaneously sense and understand our body's movement, location and actions. This might not sound that momentous, but that's because we can't really conceive of not having that sense; indeed, there are only a handful of people in the world who don't have any proprioception. The most famous is probably Ian Waterman, who lost this sense at 19, when a viral illness caused the production of an antibody which destroyed the nerves that told his brain what his body was doing. Though there was nothing wrong with the physical mechanisms of his limbs, he woke up after three days apparently unable to move. It was discovered that he actually could move his body, but not coordinate its movements; he had no feedback as to what his limbs' locations or actions were. Ian learned to sit up, feed himself and, amazingly, walk again after laboriously retraining his mind and body, relying totally on visual feedback to know where his limbs were and what they were doing. Since losing his proprioception in 1971, he has lived a life of painstaking choreography, and still his stability is fragile; if an object is heavier than anticipated, it throws off his balance completely. If the lights go off in a room while he's standing, he will collapse.

Proprioception is just one sense that humanoid robots would have to recreate to move about the world. Vision is another, and though this might seem to be as 'simple' as installing a camera, the reality is very different. It takes the human brain just 13 milliseconds to process an image captured by the eye; a five-megapixel camera will give a robot 5 million numbers to interpret per single image, with the camera sending between twenty and fifty images per second. All this data needs to be processed and stored. This is before we consider the requirements of recognising anything within those images.

It is not just the complexity of building humanoid robots that works against their potential existence; it is the running of them too. As anyone with an iPhone will know, our ability to create exciting tech far outstrips our ability to make batteries that can power it for any decent amount of time. The processing of one hundred million numbers per second is incredibly energy intensive, and that is just for one second of vision for a humanoid robot. When you add in the energy requirements of bipedal motion, we are talking about truly enormous amounts of power; even the most efficient biped robots are three times less efficient than humans at converting energy to movement. In addition, there is simply so much going on inside robots that needs to be powered; it's estimated that over fifty percent of a moving robot's power goes towards its myriad sensors, controllers and computers. Spot, the 'digidog' from Boston Dynamics-the much-hyped yellow quadroped robot often seen scuttling over terrain and being assaulted by hockey sticks in the company's promotional videos-has a best run time of only 90 mins. This didn't stop the NYPD from buying two for $750,000 each in 2023, but it'll be interesting to see how that potential run time translates to real world usage. The company's humanoid demonstrator Atlas weighs as much as Spot; though they don't specify, we can estimate its run time at likely 30 mins or less. When you compare this to how easily a person can power themselves for sixteen hours on about 2000 calories, all of which can be gained via eating a relatively small amount of food (supplemented by a couple of cups of coffee), you do sort of ask yourself what the hell we're doing trying to recreate something even vaguely similar when we already exist, and when our existence can only be as a result of some grand cosmic miracle that we do not yet fully understand.

The resounding message from robotics scholars is that it makes no sense to make a robot like a human, and even if we did want to do it, we can't. As Professor Ruth Aylett told an audience at Heriot-Watt University's Robotarium, on the topic of making human-like robots,

Let me let you into a secret, not much of a secret amongst people who do robotics, I will tell you: We cannot do that. We have no prospect of doing that in the immediate future. The more you do robotics, the more wonderful you see living things are, even quite simple ones.5

This is not an unusual claim amongst roboticists and computer scientists: currently, the concept of a functional human-like robot is nothing but science fiction. Despite this, the common narrative of public robotics discourse is that humanoid robots are imminent. There are countless movies, books and other fictions written about the seemingly...