Have a chat with artificial intelligence

From Dec. 3-8, the 32nd Conference on Neural Information Processing Systems will be held in Montreal, Canada. The conference, which is the main event in the world of machine learning and neuroscience, will feature the Conversational Intelligence Challenge 2. This is a contest jointly held by the Neural Networks and Deep Learning Lab of the Moscow Institute of Physics and Technology (MIPT) and Facebook AI Research. The contest involves training bots on the new PersonaChat data set. Each party to a dialogue receives a profile that gives a brief description of their personality. Instead of just creating a "chatter," the participants need to train systems that can accept such profiles as an input and communicate accordingly.

Aside from developing chatbots, the organizers aim to compile a corpus of dialogues between humans and the computer model for use in further AI research. The MIPT laboratory is implementing Project iPavlov under Russia's National Technology Initiative. The goal of the project is to develop conversational AI, capable of meaningful communication with a human. To collect dialogue data, the researchers use their @ConvaiBot on Telegramand the Facebook Messenger. If you want to aid this research, all you need to do is engage in a conversation with the AI model (figure 1) and rate its quality afterwards. The bot speaks English.

Mikhail Burtsev, the head of MIPT's Neural Networks and Deep Learning Lab, told us about the contest objectives: "Our goal is to take the quality of the human-machine communication to a new level. IPavlov is a three-year project aimed to further the ability of the machines to carry on meaningful conversations with an actual person. To do this, we join the forces of researchers from Russia and elsewhere. Our partners include the University of Montreal, McGill University, Carnegie Mellon University, and many others."

The Conference on Neural Information Processing Systems — variously known as NIPS and, more recently, NeurIPS — is held every December and spans the fields of machine learning and computational neurobiology. The dialogue systems developed by those who take part in the Conversational Challenge 2 could mean an important step toward the creation of new human-machine communication technology.

More information:
convai.io/

Provided by Moscow Institute of Physics and Technology