Creating Conversational AI That Learns From Talking With The User
Currently, AI assistants are developed and deployed in separate phases. First, designers, engineers and researchers create an agent using the latest tools and frameworks, and then the agent is deployed to chat with actual users. This approach creates inflexible agents that are restricted to skills that were encoded by the developers or inferred from the training data. Moreover, the agents repeat their mistakes, which can lead to a frustrating user experience. In this talk, Pararth will describe alternative approaches developed at Google Research, based on recent advances in deep learning and reinforcement learning, which enable conversational agents to incorporate new skills and avoid mistakes by learning directly from interactions with the user. Such agents improve with more usage and lead to a more engaging user experience with fewer struggles for developers.
Pararth is a Research Engineer at Google Research, working on deep learning techniques for conversational AI. He develops new technology for conversational agents that learn from both offline data and online interactions. Previously, he was an early employee at a startup building the next generation software development platform. Pararth has degrees in computer science from Stanford and IIT Bombay.