It’s called Deep Text. Facebook engineer Hussein Mehanna explained two weeks ago it can process 10,000 posts every second in 20 different languages. For the the applied machine learning team, created only last fall, that’s significant progress.

Teaching a computer to read, understand and react like a human isn’t quick project. But that’s what’s being built right now. If and when Facebook does get there, the company is going to have to prove the conversations won’t be used for anything beyond simple suggestions–and that humans aren’t reading them.

Humans can read more in the article from Recode here.