Project

Migratable AI

Jimmy Day

Groups

We live in a world of personified conversational assistants (agents). We interact with these agents in our daily lives such as smart speakers or personal robots at home (Alexa, Google Home, or Jibo). Some have access to our personal calendars and communication channels which enable them to provide us personalized services. Other robots or devices operate in public spaces such as Pepper in retail stores or Care-E at airports. Currently, these conversational agents do not share information with each other. As a result, such platforms do not support the continuity of interaction with their users across different agent personas or form factors.

Imagine a world, where we can seamlessly interact with AI assistants such as Alexa/Siri, which migrates across various devices thereby enhancing these assistants to deliver an experience that improves trust, likability, and competence.

We present "Migratable AI," a platform where a personal conversational assistant (agent) could migrate across different form factors and environments to always accompany and assist its user to support a far more continuous, personalized, and collaborative experience.

via 29th IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2020

The Migratable AI platform provides a system that lets a conversational AI agent to disembody itself from its physical embodiment and migrate into a different type of embodiment. The system allows the users to preserve the information context and identity of the conversational agent, to seamlessly continue a task with the agent across embodiments and locations.

In order to validate the efficacy of the system, we conducted a study to investigate the elements of migration, information, and identity migration on users’ perception of trust, competence, likability, and social presence across embodiments. We found that the users’ perceptions were reported the most positive across all measures when both information and identity of the AI agent was migrated across devices.

In the next steps, we aim to explore the migration of conversational AI assistant in terms of contextual intent migration. We are also exploring the option of wearable device which would allow the user to control the migration on to the embodiment.

Other Contributors

Felipe Moreno (MEng, EECS)

Research Topics

#robotics #artificial intelligence #machine learning #natural language processing