Sim scaling laws praised in robot training: why pay millions for real-world data?
lukas_m_ziegler · x · 2026-10-11
David Hoeller of Flexion AI highlighted a robot training effort as one of the most impactful lines of work in recent years, noting similar scaling laws at FlexionAI and asking: why spend millions collecting real-world data when learning in sim is far cheaper? The comment points to the scaling potential of sim-to-real training for embodied AI.
More from Embodied
- euROBIN shows 15 AI robots from 31 partners, no European orders near State Grid's 8,500 — ingliguori · 2026-10-12
- XPeng's humanoid robot Iron: mass production late 2026, first deliveries H1 2027 — kimmonismus · 2026-10-12
- LILYGO BB keyboard device impresses in hands-on with Meshtastic mesh networking — glenbeer · 2026-10-12
- Paradromics BCI lets clinical participant communicate via imagined speech — PeterDiamandis · 2026-10-12
- AHA warns wearable blood pressure devices are inaccurate and may pose patient safety risks — EricTopol · 2026-10-11
- clef-omni demos real-time audio-video decision model across 8 interactive use cases — Xianbao_QIAN · 2026-10-11