Figure claims 81 min of robot data per second; critic calls it low-value slop
MarcJSchmidt · x · 2026-10-23
Brett Adcock says Figure now "prints" 81 minutes of robot training data every second — a full human lifetime of waking hours every five days — spending an estimated $8.1M on data last week.
Marc JSchmidt pushes back: such datasets will be marketed by size and minutes (pointless metrics) rather than value/gradients, since their useful information density is tiny and mining it burns huge cash and GPUs. He also argues video is the wrong modality (vs. sensor data), so it's not even useful for world models — "all sizzle and no steak."
More from Embodied
- FSD 14 lite driver review: near-perfect performance, HW3 is all you need — haydendevs · 2026-10-23
- First real-world sighting of a Tesla Cybercab on public roads — MatthewChang · 2026-10-23
- ETH Zurich's ACP teaches Unitree G1 whole-body ladder climbing with 13 minutes of demos — CyberRobooo · 2026-10-23
- NVIDIA GTC robotics roundup: Agile ONE SN humanoid, Isaac Lab 3.0, and Jensen's 'ChatGPT moment' — lukas_m_ziegler · 2026-10-23
- Tesla patent reveals flat-printing and thermoforming process for mass-producing robot fingertip touch sensors — CyberRobooo · 2026-10-23
- Inside a Shenzhen robot-arm factory: 2,000 arms/month at ~$15k, AGIBot bought ~20,000 — DominiqueCAPaul · 2026-10-23