Google Honestly Reports Robot Bottlenecks: Low Success in Dexterous Manipulation
chris_j_paxton · x · 2026-07-30
Google recently published a report honestly showcasing low success rates for seemingly simple robotic tasks. While the robot maintains medium-to-high success on whole-body and gripper tasks, multi-finger dexterity remains a significant challenge.
Specifically, on Apollo + Sharpa hands, unscrewing a bulb achieved a 92% success rate, but tying a trash bag was only 44%, using a ziplock 40%, and a dustpan dropped to 32%. This indicates that dexterous hands are still a challenging frontier in embodied intelligence.
More from Embodied
- New MT3 Paradigm Enables Robots to Learn 1000 Tasks in Under 24 Hours — chris_j_paxton · 2026-07-31
- Simulated LiDAR Needs Real-World Noise for Better Robustness — Sentdex · 2026-07-31
- Handroid: A Reconfigurable Robot Switching Between Humanoid and Dexterous Hand — SongShuran · 2026-07-31
- Sharpa Robot Hand Ties a Trash Bag, Demonstrating Dexterous Manipulation — chris_j_paxton · 2026-07-30
- New Version of AI Pendant Friend Launches at Higher Price with Locked Personality — Wired AI · 2026-07-30
- Viam Robotics Hackathon: Chess-Playing and Lab Robots Built in a Day — ditzikow · 2026-07-30