‘Consistent Personality' for Facebook's chatbots

Facebook is still trying to make the best of its chatbots, even if personal AI assistant M failed miserably. Over the years, Facebook has slowly improved what its AI chatbots can do. However, here comes a difficult challenge. And that is making small and casual talk with a consistent personality.   

Researchers from Facebook’s FAIR lab explain that chatbots fail at a number of tasks and this includes small talk as well. The researchers explained that the chatbots don’t display a ‘consistent personality,’ and stick with the same facts about themselves throughout a conversation. Also, the chatbots can’t remember their conversations and neither can they understand questions.

Bots for Research

This is where Facebook is trying to improve the chatbots and make it more engaging. Considering the sample conversations from Facebook’s study, it was seen that the bots were more consistent and fluent than bots trained on movie phrases. However, these chatbots definitely wouldn't pass a Turing test.

With this research project, Facebook isn’t certain whether the lessons learned here will apply to real-world chatbots or other conversational AI systems. Facebook sees that many AI oriented helpers have no personality at all or are solely identified by their stock phrases. And so, Facebook wants to put in a personality so that the users can talk more with their bots than just about the weather!