When you were a child, do you remember learning what it means to be kind?
What about reading sadness in someone’s face, understanding the anger you felt, or respecting personal space?
If you don’t remember learning about any of these human moments, then Moxie, a conversational AI robot built by Embodied, is the android friend you never knew your childhood self needed.
Moxie is designed to teach children social and emotional maturity, both through human-to-robot conversation as well as various missions that send the child out into the world to learn about kindness and friendship.
To learn more about what makes Moxie tick, we hosted Stefan Scherer, Chief Technology Officer at Embodied, on the latest episode of our How AI Happens podcast. Stefan explains the language processing happening within Moxie, and how the team was able to generate reliable conversational ability through lean few-shot learning.
Moxie’s ability to shift between dynamic responsive conversation and scripted programmatic content raises fascinating questions about conversational AI: exactly how much responsiveness is required to create the feeling of a 1:1 conversation? And how can technologists draw a circle around a concept such as “kindness” in such a way that enables it to be shared between a child and an AI?
Stefan explains all this and more in the full episode. And, of course, we hear from Moxie, too.