It is hard to make robots (including telerobots) that are useful, and harder to make autonomous robots that are robust and general. Current robots are created using manual programming, mathematical models, planning frameworks, and reinforcement learning. These methods do not lead to the leaps in performance and generality seen with deep learning, generative AI, and foundation models (FMs). Today’s robots do not learn to provide home care, to be nursing assistants, or to do household chores and other services reliably.
Addressing the aspirational opportunities of robot service applications requires improving the path to get there. The high cost of bipedal multi-sensory robots (“bodies”) is a significant obstacle for both research and deployment. A deeper issue is that mainstream FMs (“minds”) do not support sensing and acting in the world. They do not lead to robots that experiment, communicate, or collaborate. They do not lead to robots that learn from and with others. They do not lead to robots that know enough to be deployed in service applications. This paper focuses on what service robots need to know. It recommends developing experiential FMs for bootstrapping them.
Publications
Stefik, M. (2024) What AIs are not Learning (and Why): Bio-Inspired Foundation Models for Robots (14 pages) arXiv https://arxiv.org/abs/2404.04267