Computers are just horrible at recognizing sarcasm. No, seriously it's true. Edgy phrases that are clearly made in jest cannot be identified by artificial intelligence yet. A team headed by Silvio Amir at the University of Lisbon, Portugal by using a page's tweet history on Twitter, and marking which posts aren't made seriously. The algorithm by Amir's team can discern sarcasm correctly 87 percent of the time. That's only a B+; I'm not impressed. Actually, I am, because it's another large step for artificial intelligence and its interactions with humans.

Source: https://techxplore.com/news/2016-08-deep-neural-network-approach-sarcasm.html