The Quanta Podcast Why Do Humanoid Robots Still Struggle With the Small Stuff?
13 snips
Mar 31, 2026 John Pavlus, a science writer who covers robotics and tech, explores why humanoid robots still stumble on everyday tasks. He discusses the importance of human-like form, contrasts practical designs with flashy demos, and highlights gaps in force sensing, physical intelligence, and training data. The conversation examines breakthroughs, realistic timelines, and where true dexterity might come from.
AI Snips
Chapters
Books
Transcript
Episode notes
2015 Robots Were Clumsy And Brittle
- A decade ago humanoid robots were heavy, top-heavy, and brittle, often falling in DARPA demos.
- They relied on older sensing like LiDAR and handcrafted control algorithms that couldn't handle perturbations well.
Three Advances Drove Recent Robot Gains
- Three technological shifts improved humanoids: reinforcement learning, electric actuators, and language models.
- These combined advances enabled smoother walking, complex actions, and multi-step planning in recent demos.
Dishwasher Demo Shows Progress But Limits
- John Pavlus highlights Figure's humanoid emptying a dishwasher as a recent impressive demo.
- He notes other demos like Atlas picking irregular auto parts and Digit filling shopping bags to show varied capabilities.
