Embodied AI faces a 100x data gap as industry races toward 10-million-hour collection
CyberRobooo · x · 2026-11-11
This thread combines a Memory AI team interview with Chinese official figures to sketch the humanoid field's desperate hunger for real-world data:
- A few months ago, industry egocentric data volumes were around 1 million hours; recently figures have reached 2 million hours, and by next year Physical AI data collection could enter the 10-million-hour range — which may still not be enough.
- At the World Internet Conference in Wuzhen, Li Shuai, deputy director of the AI Center at MIIT's Fifth Electronics Research Institute, put the core supply-demand gap in embodied AI data at roughly two orders of magnitude.
- The original post also highlights @memoriesai, which turns first-person video into searchable Physical AI training data, tracking each hand's actions — betting that high-quality curated data at scale will shape humanoid dexterity.
More from Embodied
- Simulated hand with 39 muscles learns 9/9 poses after 426M training steps — MaziyarPanahi · 2026-11-11
- Origami Robotics and PalatialSim unveil real2sim2real Dexterity Factory, 100 open datasets next week — ybisk · 2026-11-11
- Trainable JWST model boosts exoplanet sensitivity 2–15x in Nature paper — bravo_abad · 2026-11-11
- Clean point clouds: Tersus GNSS MVP S1 scans turned into high-quality splats in LichtFeld Studio — janusch_patas · 2026-11-11
- Flashing cheap ESP32 IoT devices with ESPHome for a smart-home upgrade — haydendevs · 2026-11-11
- NYU's Steertuning Lets a Robot Learn a Chore From 5 Minutes of Human Demo — CyberRobooo · 2026-11-11