Can this technology make your porn experience even better?
Porn platforms offer many enjoyable things, but an easy-to-navigate user experience isn't one of them. Now they’re hoping that artificial intelligence could change all that.
Source: Cassie Murdoch
In recent weeks, several major players in the adult entertainment space have announced plans to roll out AI-led features. Pornhub debuted theirs earlier this month, xHamster says it has a system in the works, and today YouPorn is launching their own version, along with a site redesign and a new logo.
The promise of AI is clear: to maximize personalization. Instead of relying on users alone to tag videos — which is haphazard and inconsistent — it will allow these sites to build a more comprehensive database of what they’re offering. And, in turn, it will allow viewers to find videos that are more closely aligned with their interests.
The promise of AI is clear: to maximize personalization.
There is real need for improvement, which you know if you’ve spent much time poking around these cluttered sites. YouPorn, for instance, reports that in a survey of their users, 30 percent of them said they “can’t find content they like.” And, certainly, AI promises a constantly improving way to help them.
The potential is wide-ranging, but for the moment Pornhub’s efforts are focused on recognizing specific performers, as are xHamster’s. While YouPorn is focused for now on finding common threads in their videos.
xHamster's forthcoming system at work
As Brad Burns, YouPorn’s vice president, told Mashable, "We are aiming to match genres and themes in videos." He added, "We are looking at viewing patterns across all the videos hosted on YouPorn so that we can better surface content that a particular user would be interested in. While only part of the equation, it might help to think about Netflix’s 'Because you Watched' feature or Amazon’s 'Customers that viewed this item also viewed.'"
While this does seem likely to improve your experience, this approach does have its limits. For one thing, there’s the issue of capacity. There is a huge amount of content on all of these platforms, and scanning it all is a massive undertaking.
SEE ALSO: Pornhub streams documentary on Russian corruption. Yes Pornhub.
In the case of Pornhub, the AI model views a few thousand videos and some photos of well-known performers. Then it applies that knowledge to a new set of videos, and identifies porn stars that appear in them. Pornhub users are then asked to either upvote or downvote the tags given to videos by the AI model.
It continues to improve based on the feedback, but in the month they were running it in beta, it scanned just 50,000 videos. That may sound like a lot, but the site is home to some five million videos, and they say about 10,000 new ones are added each day.
While the platform aims to have all of their videos scanned with in the year, it’s possible that it will take quite a while for the average user to really see any benefit in terms of what’s recommended. Meanwhile, xHamster, whose system is still being designed, is so confident that their system will be more efficient that they’ve issued a yet unanswered challenge to their competitor to see who can index the most videos the fastest.
SEE ALSO: Facebook has a plan to stop revenge porn—by collecting your nudes
The ultimate goal for Pornhub, at least, is to move beyond just sorting by performer. They’ll scan for every possible feature — sexual positions, locations, hair colors, etc. — in any given video. That should allow users to drill down and find things that really push the right buttons. It’s just a question of how long it takes them to get there.
But beyond simply the speed, there are ethical issues that using AI raises. Namely, could this be used as a way of making it even easier to find things like revenge porn or even just performers who wish to remain somewhat anonymously in these videos.
YouPorn says it won’t be an issue for them, because, as Burns told us, “In our case we aren’t doing computer vision or facial recognition so it’s not related to identifying the participants in the video.” Pornhub, meanwhile, has said they’re only planning to tag professionals who are already known in that ecosystem.
SEE ALSO: Australia launches a world-first national reporting tool for revenge porn
There’s also the question of whether or not YouPorn and Pornhub and anyone else who adds these capabilities will turn the power of AI on themselves. While a model can be used to scan for porn star’s faces, it can just as easily be trained to hunt for undesirable things — like pirated movies, spam, and even underage performers.
YouPorn's Burns told Mashable there was no plan to single out revenge porn yet. "As for expanding our machine learning efforts, we intend to fully leverage this technology in many aspects of our business, but nothing else specific at this point. That said, this type of revenge porn content is already strictly prohibited on our platform as it violates our Terms of Service.“
Given just how much content — both perfectly legal and deeply questionable— is running on these platforms on any given days, it will really come down to their willingness and ability to keep a very close eye on what’s being tagged.