Online robots are creating porn – with disturbing results

There’s a woman on a bed being asked to look inside a box by a man holding the camera. Dipping her hand inside, she pulls out what looks like a fleshlight. In the corner, a logo shows the words “Stepsiblings caught”. She smiles, looking with a hint of incredulity at the camera – but there’s something a little wrong with her face. It doesn’t quite sync, like you’re watching a Gal Gadot lookalike on a bad Facetime connection. Then you realise: it is the Wonder Woman actor’s face disjointedly matched onto an adult actor’s body.

Photoshopping faces onto the bodies of adult actors isn’t a new phenomenon, but the difference in this case is that a Redditer has used open source machine-learning software to create the high-tech face swap, morphing images to the original actor’s face frame by frame. The outcome is a ménage a trois of Reddit’s favourite subs: porn, practical jokes and tech.

This may be one of AI’s clumsy first forays into porn, but the industry has already been revolutionised by technology. Jon Ronon’s new podcast The Butterfly Effect maps the ripple effect search engine optimisation (SEO) has had. The launch of Pornhub by Fabian Thylmann created a pool of user-uploaded adult videos; to find the one that worked for them, users searched the database using the key words that matched their desires.

This has resulted in porn being made to fit the specifications of popular search terms such as “Stepdaughter cheerleader orgy”. Gone are the ridiculous, if retrospectively innocent, VHS porn titles from the nineties like Route 69: Sandy Rides Again, with sex scenes embedded into awkwardly acted storylines. Now videos are a salad of search terms, one of the weirder effects being that female actors between 25 and 30 years old aren’t getting much work anymore – there’s simply no search term for them.

It’s no coincidence that the Gal Gadot video and the example search term previously mentioned are incest-related – in 2016, “stepmom” was the second most search term on Pornhub, “MILF” the third, “teen” the fourth, then “stepsister”, then “mom”. Incest porn, especially that which includes young women, isn’t something any society should be encouraging. But algorithms don’t have an editorial code – they’re just mathematical probability systems giving people what they think they might want.

And this is the problem. One tapped-in desire leads to another popular one being viewed, and searched for again, creating a feedback loop. Palettes become limited as algorithms shape desires. Most of us aren’t have sex with robots, but the type of sex we are having is being directed by them, and it’s going in the wrong direction.

It won’t be long until the AI that created the Gal Godot porn and popular search terms are synthesised into a Frakenporn monster, derelict of any moral code.

On the other side of the internet, on the platform PornHub was based upon, YouTube’s kids’ videos are an indicator of this dystopian future. Cheap video content is being designed around key words such as “Finger Family Song”, “Learn Colours”, “Superhero”, “Animal Sounds”, and “Elsa”. The result is a nightmarish mishmash of cheap-looking automated 3D animations created from character models and motion-capture libraries; other times they involve humans acting out weird scenarios to fulfil titles like “Batman Finger Family Song — Superheroes and Villains! Batman, Joker, Riddler, Catwoman”. Again, there’s no editorial control here, just a mathematical probability expressed via an algorithm being acted out and autoplayed over and over.

It’s easy to see how porn could be made along the same lines. This may have hilarious results – an explosion of “Darth Vader does erotic massage”, for instance, every time there is a new instalment of Star Wars – but given the popularity of some of those other aforementioned search terms on Pornhub, I shouldn’t have to spell out why this is worrying. It is also concerning because none of the actors whose names might be dredged up by an algorithm have consented to taking part in an adult film, and it is additionally degrading for the adult actors whose heads are removed to make way for the transposition of a famous person. They are, after all, actors too, and didn’t perform in order to be suddenly reduced to random body parts.

But algorithms don’t care about consent, privacy or legality. Tech companies let algorithms do their thing, and then rely on communities to police and flag content that clearly doesn’t work – as has been the case with YouTube and the bizarre and sometimes worrying children’s videos which have made an appearance on the site in the last year.

PornHub has first and foremost sold itself as a tech company, boasting that its employees have never been on an adult film set. All the more reason why we should see it as a company with ethical standards to consider. Not every search term needs to be catered for. Sometimes there’s a reason humans need to intervene.

Not all porn is created equally, and there is thoughtful, fun, and boundary-pushing work out there by directors like Erika Lust which challenges common perception of what viewing adult content has to mean. But Lust’s work is so complex that it can’t be reduced down to a simple search term. The nuanced and sometimes quite bizarre fantasies she directs will never be stumbled across by a teenager in their bedroom, opening up a variety of consensual kinks for them to explore – instead they’ll find themselves faced with pages of thumbnails, their titles incomprehensible phrases at best, toxic ones at worst.

The creation of desires is an ineffable and complex process, not one to be whittled down to four to eight-letter words. Letting algorithms create human desire and dictate what gets us off is a future none of us want – so we better start preparing for when it becomes a serious possibility.