ParlAI built by the Facebook AI Research team is a new open source platform to train and test dialog models across multiple tasks at once
Facebook has gone a step further in exploring its AI capabilities. The company has released a one-stop-shop platform that can utilize advanced artificial intelligence to make machines act smarter. The framework named ParlAI (pronounced as par-lay) will allow researchers to build conversational AI systems and to combine varied machine dialog approaches. This will encourage dialog research where new tasks and training algorithms can be submitted by researchers in a single, unified and shared repository. The ParlAI framework will allow developers to build smarter and more articulate Chatbots that would not be easily confused by an unexpected question.
Better conversational systems and dialog platforms could mean numerous benefits for commercial applications. Facebook is known for allowing external developers to build Chatbots on its platform to complete important tasks like placing a product order or searching for information. For instance, ‘M’ is its conversational assistant that is being tested by a group of select users.
The futuristic vision for ParlAI includes its active participation in advancing the process of natural language research by decreasing the amount of work around developing different approaches.
The framework has in-built data sets with twenty different natural languages, including question-answer examples from Microsoft, Stanford and Facebook. It also gives access to several popular ML libraries. AI research has always found it challenging to create algorithms capable of learning how to answer multiple types of questions at once. ParlAI is integrated with Amazon’s Mechanical Turk platform that outsources small tasks. This enables researchers to take human help in training their dialog systems, which remains a significant hurdle in creating smarter conversational agents. The biggest hurdle that defines the limitations of AI is language complexity and its relation with learning and common-sense.
Yann LeCun, Facebook’s director of AI and a stalwart in this field told to MIT Technology Review, “This is a problem that goes beyond simply getting machines to understand language or generate speech.”
According to Richard Socher, chief scientist at Salesforce with expertise on machine learning and computer dialogue, “A complete question-answering system requires a lot of different components, which this framework looks to provide. The community will benefit immensely from a larger data set testing platform like this.”
Head of the Allen Institute for AI in Seattle, Oren Etzioni, says, “ParlAI should be welcomed by anyone working on natural language understanding. It isn’t a breakthrough, but it should be a helpful enabling technology.
There has been relatively little focus on dialogue systems in recent years.”
Etzioni added, “New machine approaches such as reinforcement learning, which mimics the way animals learn through positive feedback, could help make conversational machines way smarter. I think you will see real progress in the next five years.”
Also Read: DiscoverOrg Releases its SMB Dataset