IBM's Watson can sense sadness in your writing

The AI just got a big emotion-detecting upgrade.

Andrew Spear for The Washington Post via Getty Images

Artificial intelligence won't be truly convincing until it can understand emotions. What good is a robot that can't understand the nuances of what you're really saying? IBM thinks it can help, though. It just gave Watson an upgrade that includes a much-improved Tone Analyzer. The AI now detects a wide range of emotions in your writing, including joy or sadness. If you tell everyone that you're fine when you're really down in the dumps, Watson should pick up on that subtle melancholy. Watson is also better at spotting social tendencies like extroversion, and studies whole sentences (important for context) rather than looking at individual words.

It should be better at expressing emotion, to boot. The update brings broader availability of a new text-to-speech engine that offers responses in an appropriate tone. If you're angry, for instance, the computer won't answer back in an upbeat voice. That's not the same as taking the initiative on emotion, but this and the new Tone Analyzer point to a future where AI is sensitive to your mood, rather than cold and indifferent. Combine those improvements with an equally fresh, training-based image recognition engine and it's clear that Watson is becoming a full-fledged intelligent agent, not just a business tool or clever chef.