ParlAI Is Facebook’s Newest Research Tool Into Conversational AI

Facebook Artificial Intelligence Research, or FAIR, has recently unveiled plans for a new research tool that they are calling ParlAI, pronounced like the word “parley.” Facebook hopes that ParlAI will be useful in helping to further develop AI technology to make conversational AI, such as chatbots, more realistic to interact with.

Much of today’s AI technology is being used for chatbots, such as customer service chatbots or virtual personal assistants, such as Siri or Alexa. They can be very useful tools at times. However, their responses, for the most part, are pre-recorded. While trying to communicate with these chatbots and virtual assistants, as if they were real people, can lead to humorous results, it is ultimately unsatisfying. The AI in these tools is severely limited. They are not yet capable of replicating the various complexities of human communication. That is what Facebook is trying to change with ParlAI.

“What we’re after with ParlAI, is more about having a machine where you can have multi-turn dialogue; where you can build up a dialogue and exchange ideas,” said Antoine Bordes, a researcher at Facebook’s AI research lab, as reported by The Verge. “ParlAI is trying to develop the capacity for chatbots to enter long-term conversation.”

ParlAI is described as a “One-stop shop for dialog research.” It is a tool that Facebook developed to help other AI researchers and programmers so that the limited AI that is available now can improve to the point of holding near intelligent conversations with human beings. ParlAI is connected to Amazon’s Mechanical Turk service, a crowdsourcing site, so that programmers can easily hire humans to interact with their AI in order to test the AI and help to correct it.

Bordes mentioned that there were two different kinds of dialogue systems currently available when it comes to chatbots, those that serve a purpose, such as virtual assistants, and those that are just for fun and serve no real purpose, such as Microsoft’s Twitter AI, Tay, which failed after repeating the offensive things it was taught by human beings.

According to an article in Tech Crunch, conversation for machines is not a single task the way it is for humans. Research can be done independently for different aspects of communication for a machine, such as answering questions or making small talk, but Facebook wants ParlAI to be used to research those different aspects simultaneously.

“Researchers often focus on one of these things alone, and that could be a fundamental mistake,” said Facebook research scientist, Jason Weston, in an article by Venture Beat. “We need to look at dialogue as a whole, and so what we’re trying to do in this new software platform, ParlAI, is to put all these things together and unify this research.”

For this reason, researchers at Facebook’s AI research labs want ParlAI to connect to several different dataset sources so that programmers and researchers can access them more easily to help AI appear more human. Facebook’s AI research team wants chatbots and other conversational AI to communicate as smoothly as possible, with natural sounding conversation skills. While there are other companies out there that are researching AI technology for educational purposes, Facebook says that ParlAI is simply for dialogue.

Who knows where this technology will lead, or what it means for the future of humanity when it is perfected? Society could be mere decades away from resembling a science fiction film. How long will it be before every house has a personal AI robotic assistant? In the mean-time, people can be entertained by asking silly questions to Siri and talking to the limited AI technology that exists in other, entertainment based chatbots.

[Featured Image by Just_Super/iStock]