Facebook, Google and Microsoft have all previously jumped aboard the AI train and now Twitter is embracing so-called deep learning – an on-trend subsector of AI research involving the ‘training’ of software through constant repetition and use of different variables.
Up until now Twitter had relied on far less advanced methods for highlighting content but over the past year it has been quietly building up its capabilities in the burgeoning field through a series of specialist acquisitions.
Last year this saw it begin to provide a tailored service with the implementation of an algorithmic timeline which sought to calculate relevance rather than simply list tweets in reverse chronological order. The new system is far more computationally demanding however, assessing image and video content, retweets, likes and previous interactions to determine what is most likely to catch an individuals eye.
Twitter’s engineers hope that by tweaking algorithms behind the scenes they can boost engagement and time spent on the micro-blogging service by its 328m monthly active users.