09/07/2025
Google DeepMind is developing a system that enables robots to generate internal monologues—a sort of “intra-agent speech” that narrates their observations and thoughts. This would allow a robot to process visual and contextual information by generating natural-language descriptors like “the person picks up the cup” to guide its actions and learning
DeepMind’s “inner speech” approach may unlock smarter, more adaptable robots by giving them a verbal thought process, improving learning, action planning, and transparency. While still in patent stages, it reflects a growing trend of integrating language models directly into robotic cognition.