Musings On Robot Sex Dolls and Companions

The currents of the internet work in odd ways; this past week the theme seems to be robot sex. Since I have had it on the brain, I figure I will contribute to the trendiness and throw my own 2c in. (Just as a note, I will indicate any link that is explicitly Not Safe For Work).  I am going to blur the line a bit between just discussing robot sex and discussing robot companionship, a somewhat more involved relationship than the purely physical.

It seems to me there are essentially three main questions when it comes to human-robot sex. First, can we build a machine that anyone would want to have sex with? Second, how “intelligent” should that machine be? Third, is this just a fetish for weirdoes?

Technical Feasibility:

Not only can we build robots that people want to have sex with; we already have.

Certainly, there are all manner of devices people use for sexual pleasure, but I want to focus on machines more sophisticated than your average vibrator.

The aptly titled fuckingmachines.com (NSFW) is a pornographic site founded in 2000 that features videos and pictures of women having sex with robots that are not particularly technically advanced, and certainly not on the level of a sophisticated android sex-bot. Think battle bots for the bedroom. Despite the lack of sophistication, these are industrial pieces of hardware. For the home user, somewhat tamed versions of machines built for pleasure are available from mainstream websites like this “Love Glider Sex Machine” from Amazon.com (NSFW).

Andydroids.com (NSFW) has a number of both male and female android dolls for purchase. Although the website is not well constructed, this page (NSFW) seems to show various servos, circuit boards, and otherwise fairly advanced robotics working together to create a somewhat lifelike robot. Less sophisticated, but perhaps more lifelike, are Real Dolls (NSFW), in production since 1996. Real Dolls are as close as I have seen to human-looking sex bots, but are still a long way from indistinguishable from human.

The most realistic robot that I have yet seen (though it is not designed specifically for sex) is Geminoid F from Osaka University’s Professor Hiroshi Ishigurou. This robot can smile, talk, move, and appears very lifelike. According to this video, she even has “basic emotions and behaviors” programmed in. The biggest problems that I can see from the demonstration videos are that (1) the robot might be firmly entrenched in the “uncanny valley” (2) her movements are still a little jerky, and (3) her software is highly advanced, but hardly lifelike.

The uncanny valley is a hypothesis that argues that as robots become more human-like a human observer’s emotional response becomes more positive and empathetic. However, at some point, the robot is –too- lifelike, and a feeling of revulsion quickly replaces the positive and empathetic emotional response. If the robot becomes yet more lifelike, to the point of being indistinguishable from a human, the human observer’s emotional response will again become positive and empathic. Thus, to have a sex bot that anyone would actually want to have sex with, the robot is going to have to be on one side or the other of the uncanny valley; either not particularly lifelike, or extremely lifelike. For a robot that is expected to be more than a sex toy (say, for someone that a human might want to be partnered with) the robot would have to be extremely advanced and nearly indistinguishable from a human being.

Jerky movements can be compensated for by ever-better servos and other methods of movement. Popular Science, for instance, recently reported on Nobuhiro Takahashi and the University of Electro-Communications’ new robotic butt that responds to “slaps, caresses, and finger pokes.”

The video is a little creepy, but shows the sort of fine ‘muscle’ movement that Geminoid F lacks; movement that could be very useful in other parts of the robot as well.

ExtremeTech posted an article about Kissenger, a telepresence robot designed to allow two humans to kiss across great distances through a robot. Although this is hardly more advanced than previous robots, it does suggest that humans are willing to at least attempt to transmit an emotional connection through a robot. In addition, as ET points out, how much of a stretch is it from kissing a robot with another human on the other side to kissing a robot controlled by an A.I.?

This ScienceDaily article highlights synthetic skin that could, one day, allow a robot to feel. Even if we assume that there is no qualia (roughly: experiential consciousness) behind a robot feeling, all the data streams involved in transmitting some kind of feeling could be very useful for triggering micro-movements in various parts of the skin, perhaps even including subtle changes like goose bumps, etc.

Technically, I think we are about there. Some more materials development (in particular a temperature regulation system and a lubrication system would be two huge upgrades that I have not seen) some finer muscle control, and some more realistic design and robots might just climb out of the uncanny valley. However, what about the software side of the robot?

A.I. and Sex-Bots:

The next question is how much artificial intelligence a robot companion ought to have.

On one end of the scale, we have Real Dolls – essentially human-looking mannequins without any sort of robotics or artificial intelligence. These sorts of sex-bots are fine as far as they go for purely physical entertainment, but most people probably will not develop any emotional connection to their toys (especially if they hang their Real Doll by the “removable neck bolt” as their FAQ suggests.)

Towards the middle of the scale, and likely right at the edge of our current capability, we have Geminoid F; a robot with basic emotional scales programmed in that can spontaneously create new reactions to situations. The jerky physical movement is mimicked by the jerky emotional reactions; they are broadly appropriate, but are not exactly finely tuned enough to seem human.

Ideally, it seems like the perfect robot companion ought to have emotions that at least mimic human emotions very well; the ability to smile, wink, and bite their lip at just the right time and have something that at least seems plausibly like a twinkle in their eye. Perhaps complex human-based personality profiles could be uploaded that allow the robot to seem very much like a human being, albeit with customizable settings for each individual user to account for differing tastes. Maybe the robot could exhibit this personality outside of the bedroom as well; transforming a sex robot into something more like a personal companion or even a partner.

However, it seems important to limit both sex robots and companion robots to non-conscious levels of intelligence. Most importantly, because I think that cognitive criteria are the defining hallmarks of a “person,” and that a robot with actual consciousness ought to be considered a person. If we think it is wrong to keep people for sex toys (and we certainly do) then I cannot see the same behavior being justifiable for conscious robots.

However, even outside of the moral personhood angle, a conscious robot would have something like free will, or at least clearly articulable preferences. If the goal of a sex-robot or companion robot is to have the ideal partner, then we certainly don’t want our robot telling us ‘no’ or ‘I’m not in the mood’ (unless we program that in for some sort of more realistic behavior.) We want to be able to program in our individual desires and preferences which make the robot ideal for each of us, and a robot with free will would presumably be overwriting our preferences with their own fairly often. A robot with true artificial intelligence would not have many advantages over a human partner.

In short, much like the physical problem of the uncanny valley, we want a robot intelligent enough to seem human-like without actually being conscious enough to be a person.

Who Would Want A Sex Robot?

We can dispense with the obvious fairly quickly; probably people with intimacy issues, various kinks and fetishes, and those who just want sex without everything else that often comes with it would be first in line for a very realistic sex-bot. ExtremeTech recently wrote an article about robot prostitutes that argues that robots could take over the prostitution industry (wouldn’t a sex-bot be cheaper over the long run, after all?) in addition to lessening human trafficking, pedophilia, and other sex crimes.

I think, however, a compelling case can be made that more than just the socially awkward and sexually deviant (in the clinical sense) would appreciate a sex-robot. Dick Pelletier recently wrote a piece for IEET where he highlights a number of authors who have argued just that, including tech luminary Ray Kurzweil: “Author Ray Kurzweil says tomorrow’s ‘droids could quickly learn to flesh out our positive feelings, providing an addictive allure almost impossible for us to resist.” Indeed, with ruthless, cunning efficiency a robot with sophisticated enough software could read various biometric signals that humans give off, allowing him or her to customize their personality to the preferences of their human owners that the owner may not even know that they have. Moreover, like any good device, the robot would presumably become more accurate over time, and change as their owner does. This sort of adaptive learning is an ingenious solution to forcing the operator to think of all of their own preferences and program them into their robot companion; something humans have a difficult enough time expressing to each other.

The allure of the perfect seducer / seductress is vast, and not to be underestimated. No matter how fabulous your human partner is, there is bound to be –something- about him or her that is not 100% ideal. Maybe they snore. Maybe they like to cut you off while you are talking. Maybe they just forget to put the toilet seat down. Whatever it is, trivial or serious, there is some way (and, likely, a number of ways) that they are not ideal. Of course, humans overlook these qualities in other humans all the time during relationships; coping with each other’s idiosyncrasies and quirks (which might even become endearing after a while) is largely what human relationships are about, and provide an extra level of intimacy in a relationship. Nevertheless, even if your human partner –is- wonderful and you cannot think of a single thing you would change about them, they are still only one personality.

An interesting implication of robot-companions is that there is little reason why multiple personalities could not be installed within one physical frame, and those personalities could be changeable at will. Maybe you want a sultry professional for an office meeting, a wild party girl for a Halloween party, a tomboy for a Super Bowl party and a quiet intellectual for a lazy Sunday afternoon. Perhaps you want a nice gentleman for dinner, a jock for the pool, and a real alpha-male for bed later. A robot companion can switch effortlessly into different personalities, each tailored to your specific desires. These personalities could even be ported into different physical frames for those who desire a differing physical appearance every now and again.

Beyond the physical and personality advantages, there could be greater emotional security from a companion bot as well. From Dick’s IEET article: “A robot partner would be the perfect mate, never showing boredom or being inattentive, Levy says. You will always be the focus and centerpiece of their existence and you never need worry about their being unfaithful or going astray, because loyalty and being faithful are embedded in their programming.” With a divorce rate hovering somewhere around 50% in the United States, human relationships seem to be the emotional equivalent of a coin flip (and subsequent relationships fare even worse.) Never mind the cost of alimony and child support.

In short, I think that with advanced enough A.I. (but not too advanced, per the above) sex or companion robots could very well become the ideal mates for humans. Human-robot relationships could be purely sexual, or they could become more like true companions. Either way, such human-robot interactions do not necessarily mean the end of human-human interactions, or inevitable extinction for lack of reproduction. There are, after all, plenty of children to adopt, and there is little reason to think that the technology involved in creating children will fail to advance as rapidly as other technologies.

We are still a long way from this sort of interaction, but the upsides seem considerable.