Facebook's 'near human' AI is analysing your posts to strengthen its search engine
Facebook has announced a new AI product called ‘DeepText’ to help it better understand the meaning and sentiment behind everything posted on the platform.
The world’s biggest social network announced the news in a blog post, saying that its new deep learning-based text engine can understand “with near human accuracy” the textual context of several thousand posts per second.
The technology, which will be used to analyse all text posted to the platform be it a status update or a private message, recognises over 20 languages, and has been trained to help surface content for users that they may be interested in while filtering out spam.
The machine learning element of DeepText, means that the more Facebook analyses users' posts, the more it will be able to identify what they may be interested in, and essentially gives the social company the ability to become a global search engine by absorbing how users make everyday conversation.
The social network says it’s already testing the AI on messenger, analysing interactions between users to detect intention. The move could give Facebook the option to push some of the service and product integrations it already offers to users – ie someone saying they ‘needed a ride’ or ‘wanted a taxi’ the tech could let Facebook ask if you want to book an Uber or a Lyft without leaving Messenger.
Facebook has made several forays in to AI this year already, announcing its separate Bots on Messenger offering at its F8 developer conference a few months ago.
The branded-chatbots, which have already courted interested from the likes of KLM, will let users review orders, give feedback to advertisers and more.