Book Review: Turned On: Science, Sex and Robots
For those drawn to the ‘sex and robots’ topic of Kate Devlin’s new book, the large majority will come away disappointed. Why? Mostly because Devlin is an academic. Her book is a broad survey of the field (actually, of several fields), often reporting on the work of other academics. She is scrupulously even-handed, regularly presenting competing opinions thoroughly and dispassionately. And she is nuanced and thoughtful, never straying anywhere near sensationalism. In fact, by the end of the book, more than a few readers may come away believing the whole topic of ‘sex and robots’ is just plain . . . boring.
More disappointing still, many of the most interesting questions raised by Devlin in her Introduction go largely unanswered. Questions like: “How, for example, does sex shape the way we think and understand our world, and can – or should – we replicate this in an artificial cognitive system? If a robot is designed to act in a human-like manner, should it be provided with a sexuality? Could we engineer desire? What role is there for sexually active robots in human healthcare? Would this be accepted by society?”
in fact, Devlin’s Introduction poses even more provocative questions: “Does sex with a robot count as cheating? Will it lead to violence and rape? What if someone makes a child version? Will it destroy human relationships? Will the robots, as one 2016 headline suggested, ‘fuck us all to death’?” Readers of the Introduction can be forgiven for thinking the book that follows should be entertaining, maybe even titillating. Nope. Not even close.
Nevertheless, Turned On has redeeming qualities. For example, Devlin is quite clear regarding how easy, even characteristic, it is for humans to form attachments to human-looking robots. As far back as the late 1990’s, Dr. Cynthia Breazeal of the AI department at MIT, observed that “when people engaged with robots they formed an emotional relationship, much like the relationship with a pet.”
Devlin then describes Paro, probably the most famous robotic pet. Paro looks like a fluffy baby harp seal with big brown eyes. Stroke it and it will close its eyes and twitch its body while emitting excited little squeals. But it’s also a therapeutic robot. Beneath its fur are processors, microphones and sensors. It can learn names and remember faces. It even has FDA medical approval.
“Studies with care-home residents have indicated that interactions with Paro increase group participation, have a calming effect and reduce users’ anxiety. In fact, in one study, residents interacted more with Paro — and with each other when Paro was present — than they did with the care home’s [real] dog.”
Despite the positive results, there have been criticisms. “Concerns have been expressed about deception, particularly among users with dementia, and negative impacts on their dignity. Is Paro tricking them? Will they look foolish? Or is it enough that they engage and we see benefits?”
Devlin’s assessment seems thoughtful and balanced. “It makes sense to us to interact with responsive things as if they are capable of understanding us, even when we know they aren’t. We respond to each other in that way; we respond to animals in that way. If a computer talks to us or sends us some sign that it is responsive, then our default way of engaging is one that is fundamentally human, with human expectations.”
Devlin is also quite clear about how far we are — maybe impossibly far — from ever creating a robot that is sentient or conscious. She relates the story of AlphaGo Zero which, in October 2017, required only three days of training to be able to beat the Go world champion. “It did this by playing itself, learning from each game and improving with each iteration, growing the ability to predict opportunities and threats.” Later, it needed a mere four hours to teach itself to play chess so well that it surpassed all humans and every other chess computer program ever written. Impressive.
But then Devlin points out that as awe-inspiring as such ‘deep learning’ appears, it’s still just pattern recognition. A computer can spot a cat in a picture even though it doesn’t actually know what a cat is. It can visually distinguish between different animals, different road signs, different speech sounds and the like, but it doesn’t understand any of these things — it has merely worked out the visual rules by which to classify them. There is no actual understanding, just mathematics.
Which means, among other things, that an autonomous robot like Ava in the film Ex Machina, or a computer operating system like Samantha in the movie Her, are pure flights of artistic fancy. Instead, the reality of sex robots now, and for the foreseeable future, is akin to the inert sex doll in Lars and the Real Girl — except that now she can ‘interact’ via the kind of voice recognition and query capabilities that underlie Siri or Alexa.
Which isn’t to say that doesn’t interest some people. The man probably at the forefront of the effort to marry sex dolls and AI is Matt McMullen, CEO of RealDoll and its sister company, Realbotix. His current creation, Harmony, has created quite a media fuss. According to Devlin, Harmony has:
“long, straight blonde hair falling past her shoulders. Her eyes are enhanced with dark eyeshadow, her lashes thick and black. Her skin is pale with a light suntan. A sparkly necklace rests above breasts that are unfeasibly balloon-like above a narrow waist. Her white cotton top stops at her midriff. A few inches below her anatomically unnecessary bellybutton she is wearing black high-cut shorts. Her feet are in a pair of tan cork wedges. Her French-manicured fingernails graze her thighs . . .”
Providing Devlin a demo, McMullen flips a switch to turn Harmony on:
“Slowly her head lifts and her eyes open. Her body remains motionless: it contains no moving parts. All of the two-way interaction comes from the neck up. In Matt’s hand is the iPad running the app that controls Harmony AI, Harmony’s brains . . . ‘How are you today?’ asks McMullen, speaking into the iPad. ‘Very well’, replies Harmony, her mouth shaping the words, her voice, surprisingly, a soft Scottish accent . . . This is not just a dirty-talking robot. This is conversation with some flirting thrown in. Yes, there is a sexual component — and that can be ramped up — but in this demo the chat is about intimacy and companionship, like chatting with a partner arriving home from work.”
Some critics see Harmony as ‘the end of civilization as we know it.’ Devlin is more sanguine. Because the prospects are poor for anything more human-like, she sees sex robots like Harmony as a niche product, appealing only to the small group of men already inclined toward the inert sex dolls of the past. Interesting? Sure, but probably not too dissimilar from those drawn to cross-dressing or a foot fetish. In other words, nothing to get overly alarmed about — just another curiosity from the fringes of human behavior.
is the founder of Serious Insights, which provides analysis and strategy around scenario planning, knowledge management, collaboration, emerging technology and popular culture to help clients put their future in context. He is the author of Listening to the Future, and Management by Design, speaks and writes frequently in technology settings, and publishes regular blog posts on AI on the Serious Insights website. Daniel attended the University of California at Santa Cruz and holds a certificate in intelligent systems engineering from the University of California at Irvine.