Pornhub is improving search with an AI porn addict

The system uses machine learning to catalog and tag every clip in Pornhub's vast database.

NSFW Warning: This story may contain links to and descriptions or images of explicit sexual acts.

One of the (many) problems that porn websites that rely upon user-generated content have, is that its one-handed users often don't respect proper database use. As a consequence, you may have thousands of clips, all described with the same five words and two tags. That makes discovery and cataloging a problem, especially if you're looking for videos to cater to your very specific niche or favorite performer.

It's an issue that Pornhub is looking to remedy by implementing an artificial intelligence that will scrub through every frame of every video in its catalog. The system has been fed thousands of images of specific models and acts to create a database of names, faces and positions. Then, that data will be compared with the clips on Pornhub's system to automatically tag and catalog all of its vast library of content.

The system will then attempt to use computer vision to identify the act being carried out on screen at any given time. In (very, very unpublishable) clips shared with Engadget, the system was able to identify both the names of the performers in a scene, and what they were doing. Tags such as "blowjob," "doggy," "cowgirl," and "missionary" floated on screen with the corresponding action. The system is also capable of, for instance, identifying blonde performers and adding the requisite tags.

Currently, Pornhub groups videos by what other people watched on their research trail, combined with the aforementioned tags. In the future, it's hoped that visitors will be able to find clips according to Pornhub's AI smarts, with a confidence rating attached. If you're looking for Model X, you'll be shown all of the clips that the system believes they've performed in, alongside a percentage rating. Users will be asked to help train the machine further by up-voting correct guesses and down-voting errors.

The system has run through around 500,000 featured videos as of today, but the company is hoping to have cataloged its entire library by the start of 2018. And as dystopian as this all sounds, Pornhub has pledged that it will only tag professional, "known" models rather than amateurs. Not to mention that the system can't identify performers with masks or obscured faces. As a consequence, perma-masked BDSM / fetish performers like Emma Lee, and those looking for anonymity should be able to keep their identity off the site.

As creepy as this could sound (and, it does), the news also highlights AI's power to highlight and correct for human error. Pornhub could use the system to find low-quality duplicate clips and cut-downs of material and eliminate them. Not to mention that it can utilize machine vision and learning to effectively automate tagging and identification, plus description-writing. The end result should be better discoverability, higher quality videos and fewer duplicates. Should, at least.