Listen Live
    Artificial intelligence

    Know Your Heroes: PornHub's New AI to Watch Videos, Guess Actors and Genres

    © Photo : Pixabay
    Tech
    Get short URL
    132
    Subscribe

    A novel machine-learning algorithm is about to become the biggest porn addict ever – as soon as it finishes its “training course” based on the vast content of the website, the system will be able to recognize 10,000 pornstars and tag the genre it’s watching.

    PornHub, one of the world's biggest adult video sites, is currently training an artificial intelligence to automatically put tags on their videos. Until now, this work has been done manually by a team of moderators who will now have their hands free to do other tasks.

    At first, the clever system will identify the porn actors in the videos, but soon it will also understand what genre and category they belong to. The AI will assign tags based on who is in the video, what exactly happens there and what locations, positions and attributes are used. The system will also be able to distinguish hair and skin color, as well as body shape of the actors.

    ​The system was connected to a database with photos, names, characteristics and frequently used tags. The AI has also run through around 500,000 porn videos on PornHub so far, but by the beginning of 2018 it will finish scanning the entire porn site's collection of over 5 million clips.

    ​According to PornHub, face recognition will only identify famous actors from the base of 10,000 faces that it has been "introduced" to, which means that independent authors and performers will remain anonymous. In addition, users will be able to estimate the accuracy of the algorithm's work.

    Related:

    PornHub Reveals North Korean Elite's Porn Preferences
    PornHub Unblocked in Russia Following Dramatic Standoff With Watchdog
    PornHub and YouPorn Go Full Frontal, Implement HTTPs Security to Stop Hackers
    Tags:
    artificial intelligence, AI, pornstar, porn, PornHub
    Community standardsDiscussion
    Comment via FacebookComment via Sputnik