Facebook builds AI that can read and understands posts and messages like a human

The site says that it will use the robot to find out more things that people want to see, or to remove things that they don’t like spam

Facebook CEO Mark Zuckerberg speaks at the F8 summit in San Francisco, California, on March 25, 2015. Zuckerberg introduced a new messenger platform at the event (Josh Edelson/AFP/Getty Images)

Facebook has built an AI system that can read text as well as a human.

The new computer can read through several thousand posts per second, in more than 20 languages, and understand what they are all saying.

The new tool will be used to let people get more out of the site, it said. “Understanding the various ways text is used on Facebook can help us improve people's experiences with our products, whether we're surfacing more of the content that people want to see or filtering out undesirable content like spam,” it wrote in a blogpost announcing the change.

The tool is initially being tested in Facebook Messenger, and is being used to help the site understand certain searches.

On Messenger, for instance, the site is using the tool to scan through messages and tell what people are wanting to do. It needs to be able to tell what the difference is between “I just came out of the taxi” and “I need a ride”, Facebook said in its blog post, so that it can help – or potentially in the future advertise to – them with more relevance and accuracy.

The tool, called Deep Text, has been developed by engineers working on the site’s bot platform. It uses deep learning, which means that it can understand things like a human and come to learn more about a language without people telling it what to think.

That will mean that it will be able to scan through conversations and pick up on slang or other differences in language automatically, Facebook said.

It hopes to go on to make its AI even smarter so that it can do other things like come to understand the context of a discussion, such as what page it is happening on, as well as making it better at understanding how text and pictures can interact to make meaning.