Extending robot minds through collective learning.

Prorok, Amanda · Sci Robot · 2025

Where this comes from

Abstract

The current trend toward generalist robot behaviors with monolithic artificial intelligence (AI) models is unsustainable. I advocate for a paradigm shift that embraces distributed architectures for collective robotic intelligence. A modular "mixture-of-robots" approach with specialized interdependent components can achieve superlinear gains, offering benefits in scalability, adaptability, and learning complex interactive skills.