MIT: AI Robotics Hype Outpaces Physical Reality

An analysis by MIT Technology Review and Aventine challenges the timeline for widespread humanoid robot adoption, citing skepticism from leading researchers like Yann LeCun. While companies like Tesla and Google DeepMind showcase progress with models such as Gemini Robotics, these systems rely on vision-language-action (VLA) architectures that struggle with tasks outside their specific training sets. The article highlights that hard-coded policies are being replaced by AI-driven 'robot policies,' yet current VLAs lack the generalization required for true autonomy in unstructured environments. Consequently, experts argue that the physical variability of the real world remains a significant barrier to the 'superhuman dexterity' predicted by industry leaders.