Deepak Pathak: There Is No Perfect Data Source for Robotics
chris_j_paxton · x · 2026-10-21
CMU professor and Skild AI co-founder Deepak Pathak argued at the Encord Long Horizon Summit that there is no "golden probe" — no perfect data source for robot learning.
- His slide compared four sources on scalability, diversity, and closeness to real robots: Simulation scales best but has low diversity; Video scores high on both scalability and diversity but is farthest from real robots; SUNI sits in the middle on all axes; Teleop is closest to real robots but lowest in scalability and diversity.
- Takeaway: every data source has clear weaknesses, so foundation models for robotics must combine sources rather than bet on one.
More from Embodied
- Stanford's FACT model uses human force signals to beat T-Rex by up to 60 points on dexterous tasks — stanfordnlp · 2026-10-21
- Stanford's T2-z lets robots probe hidden physics with a poke, lifting success from 34% to 73% — stanfordnlp · 2026-10-21
- Snap Hires Robert Downey Jr. as Chief Customer Experience Officer to Sell Specs AR Glasses Against Meta — adariostrange · 2026-10-21
- Upcoming paper claims big models can handle latency-critical real-time robotics — ManlingLi_ · 2026-10-21
- New project at the intersection of agents, embodied AI, and games gets a nod — ZeeshanZiaML · 2026-10-21
- NVIDIA open-sources Flora, a robotics data engine 125x cheaper than GPT-6 Astra — DrJimFan · 2026-10-21