How robot startups scrape for data: free cleanings, exoskeletons, sim limits
binarybits · x · 2026-09-04
A deep Understanding AI feature on the robot data scramble: startup Shift offered free NYC apartment cleanings while cleaners wore camera hats to sell manipulation data to robotics firms. The largest open robot dataset (ABC-130K) holds just 3,500 hours of demonstrations — nowhere near internet scale. The author interviewed dozens of founders and researchers, visited UPenn's lab, and watched VR-puppeteered humanoid data collection in China. The post's clip also illustrates why training manipulation in simulation still doesn't work well.
Related event: Why Simulation Falls Short for Training Robot Manipulation(2 posts)→
More from Embodied
- Sentdex trains robots to go from lying down to a half kneel, eyeing a full stand — Sentdex · 2026-10-21
- NVIDIA open-sources MotionBricks: 350K motion skills at 15,000 FPS, 2ms latency — aftahi_ai · 2026-10-21
- NVIDIA Cosmos3-H-Dreams: Generative Simulation Brings Clinical Surgery to Life — Hugging Face Blog · 2026-10-21
- FLoTS-SLAM uses optical-flow residuals to hit 0.73 DM-IoU in dynamic monocular SLAM — zhenjun_zhao · 2026-10-21
- Plato ring launches at $199 to turn near-silent speech into AI commands — TinfoilTricorn · 2026-10-21
- Dream-1 controlled 31 unseen robot types across 53 countries within 24 hours — gharik · 2026-10-21