Twitter has begun using deep learning, a trendy type of artificial intelligence (AI), to recommend certain tweets in users’ timeline, reported CNBC.
To server its 328 million monthly active users, Twitter is now using artificial intelligence on its micro-blogging website to rank up tweets in users’ timeline, quitting on a year-old algorithm it had been using until now.
The ranking algorithm is powered by deep neural networks, leveraging the modeling capabilities and AI platform built by Cortex, one of our in-house AI teams at Twitter, Twitter software engineers Nicolas Koumchatzky and Anton Andryeyev wrote in a blog post.
Originally, Twitter used other types of algorithm such as decision trees, logistic regression, feature crossing and discretization for ranking tweets.
Twitter had acquired machine-learning startup Magic Pony technology and open-sourced some of its deep learning software. However, the firm remained tight-lipped on its works on artificial technology into the platform.
Twitter is constantly evaluating and scoring thousands of tweets per second to analyse what’s worth recommending in timelines. The firm is also taking pictures, videos, likes and retweets into account, among a number of factors to determine which tweet should show up in timelines.
"Online experiments have also shown significant increases in metrics such as Tweet engagement, and time spent on the platform," Twitter engineers wrote.
