Facebook's artificial intelligence lab has developed an animated bot capable of mimicking human responses to conversation by training it using footage from hours of Skype conversations, per New Scientist.
- One big problem: AI's biggest hurdle has been the uncanny valley, a state in which robots read to humans as almost realistic but still undeniably artificial.
- Facebook's researchers plotted 68 different facial reactions that take place during a conversation into the bot, which eventually learned the "correct" ways to respond to conversational cues. Volunteers judged the bot and a human to be equally natural during conversation.
- Why it matters: The bot is only still approximating — through repeated machine learning — what it thinks a conversational response should be, rather than holding a genuine conversation. Science still has a long way to go until such technologies can develop their own personalities or be tailored to response to specific situations.