It’s a fairly common problem on adult portals where visitors upload the content: almost nobody bothers to catalog properly what they’ve uploaded. This leads to a real nightmare of wrong categories, meaningless tags, and very poor search results. In an attempt to fix it, the people at Pornhub are throwing the power of artificial intelligence at their five million videos to analyze and reclassify each one.
Why the Catalog Is a Mess
Some actresses in the adult entertainment world sincerely don’t need any kind of introduction, and to be fair, that extends to many actors too. In fact, their names are usually the main search term, but more than once we’re surprised by the poor results. The reasons are obvious: mixed categories, confused titles, wrong tags… the list goes on. The simple mistake of tagging a professional video as “amateur” is enough for it to get lost in the search engine and appear only by chance.
How the AI Will Fix It
The people at Pornhub know this is a serious problem, and with five million videos on their platform, we can comfortably say that manual identification is a losing battle. That brings us to the launch of a new AI model based on facial recognition and computer vision. The model will be trained with thousands of videos and official photographs. Its first beta processed 50,000 videos in a month (it will need to be much faster than that), and once it has developed an advanced level of confidence, the AI will extend its analysis to the rest of the content on the site, identifying each star. Finally, its modifications will be validated by the community, reinforcing the successes and learning from every mistake made.
Pornhub’s goal is to scan the five million videos in its database before the end of 2018. It’s also expected that the AI will begin modifying categories, removing erroneous tags (another critical factor in improving search results), detecting spam entries, and optimizing personal recommendations, although I don’t think the latter is so necessary. After all, whoever enters Pornhub knows exactly what they want...
