Living Mind: Why AI Without a Body Remains Blind express
The modern standard approach to AI training is like a child locked in a library: he remembers millions of books but doesn't know that fire burns and falling hurts. Our brain, built from carbon and other simple substances, works differently: first, the body explores the world through movements, touches, mistakes, and only then do words gain meaning.
Neuroscientists have identified five innate mechanisms of this 'grounded' cognition: spatial mapping, grasping, recognizing one's own actions — they are active even before speech. They are united by entropy, a measure of disorder that drives the brain to seek novelty, ordering the world. Infants distinguish their own actions from others' in the first months — this builds the sense of 'self' and empathy.
The takeaway for AI: it's pointless to stuff it with texts. A machine needs a body capable of falling and learning from its mistakes. Then physical intuition, imagination, social sense will emerge. This process, like the expansion of the universe, will open new horizons of understanding. The breakthrough will happen not in a data library, but on a robot playground where they'll get bumps and bruises.
Original abstract
as written by the authors arXiv:2607.13560
Recent advances in generative and embodied AI have been driven by large-scale predictive learning over multimodal data. However, the resulting systems remain largely based on passive training regimes where linguistic regularities create the scaffold onto which information from other modalities is attached. Conversely, neuroscience and cognitive science suggest that biological intelligence is organized in the opposite way, where grounded world models acquired through interaction with the environment provide the semantic scaffold to which language is attached. Here, we illustrate five examples of neural circuits supporting grounded world modelling, which underlie navigation in physical and conceptual spaces, affordance-based perception and interaction with objects, active perception and exploratory learning, allostatic control and emotion, and the distinction between self- and world-generated outcomes. These examples highlight several features largely missing from current embodied AI, including the role of intrinsic dynamics as a foundation for learning, the centrality of action in aligning these dynamics with the external world, the prominence of autonomous experience and open-ended learning over passive assimilation of externally provided data, and the fact that early predictive and control mechanisms scaffold higher cognitive abilities such as reasoning, conceptual navigation, planning, imagination, understanding others' minds, and communication. Finally, we discuss whether and how principles derived from biological systems may inform future embodied AI, including training regimes based on social interaction to construct world models that are not only grounded but also socially shared and aligned with human norms and values.
When you read the word 'kick,' the areas of your brain that control your legs subtly activate. Even in adults, language is 'grounded' in bodily experience.
Training AI through action echoes films like 'Chappie' and 'A.I. Artificial Intelligence,' where robots develop personality by bumping into reality.