Famed roboticist says humanoid robot bubble is doomed to burst (techcrunch.com)

Renowned roboticist Rodney Brooks warns that the current humanoid-robot boom—fueled by high-profile startups and massive VC rounds—is built on flawed assumptions and unlikely to deliver scalable products. Brooks argues that trying to teach dexterity by showing robots videos of humans is “pure fantasy”: human hands have roughly 17,000 specialized touch receptors and decades of bespoke sensing and data-collection infrastructure underlie past AI wins in speech and vision, neither of which exists for touch. He also flags fundamental safety and physics problems: full-size bipedal robots store large amounts of kinetic energy, so falls become exponentially more dangerous as size increases. Brooks predicts practical “humanoids” in 15 years will abandon the human form—opting for wheels, multiple arms and task-specific sensors—rather than being literal human replicas. For the AI/ML community this is a call to shift priorities from glamorized end-to-end training experiments to engineering realities: build richer tactile sensing, robust control, safety engineering, and manufacturable hardware-software stacks. The critique lands while companies like Figure and Apptronik attract billion-dollar commitments (Figure recently reported over $1B in committed capital at a ~$39B valuation), underscoring the risk that large investments are funding unscalable training runs instead of foundational tooling. Brooks’ view — echoed by empirical cautions about perceived AI productivity gains — signals a needed focus on embodied intelligence fundamentals rather than humanoid aesthetics.