MaxInsights pushes physical AI data to 2M hours of real-world experience, 450K hours produced monthly
CyberRobooo · x · 2026-09-23
Following Dyna-2's finding that 1M+ hours of egocentric human video can scale robot learning without robot data in pretraining, MaxInsights is pushing to 2M hours of real-world experience.
- Frames physical AI data scaling around two dimensions: Experience Scale (hours) and Experience Density (learnable physical info per hour) — Effective Experience ≈ Hours × Information per Hour; ordinary video and footage capturing hands, contact, force, and temporal changes differ vastly in value
- Dataset combines egocentric video, hand/body motion capture, tactile/force signals, 3D information, and data evaluation for VLA and World Action Models
- Claims 2M hours accumulated, up to 450K hours of video production per month, 50K+ scene types, and 9 categories of physical signals
More from Embodied
- Homebrew robot arm build log Day 7: geartrain CAD, wiring and gripper sizing — _Stocko_ · 2026-11-11
- Perceptron AI launches EgoHome, an open benchmark for sub-second egocentric hand understanding — rohanpaul_ai · 2026-11-11
- Runway open-sources Praxis-1, turning video pretraining into real robot control — runwayml · 2026-11-11
- Unitree G1 humanoid wearing VR goggles pilots an A1 drone in land-air robot demo — CyberRobooo · 2026-11-11
- CoRL Spotlight Meta-Ctrl predicts robot plan failures, Llama 3 8B beats GPT-5.5 on VirtualHome — guyvdb · 2026-11-11
- Luna, a jewelry-style AI wearable that speaks before you ask, launches on Kickstarter — nikola_mr64990 · 2026-11-11