LLMs' best robotics role: generating controllers to pretrain robot models
igilitschenski · x · 2026-09-09
Igor Gilitschenski argues that if LLMs excel at generating code to control robots, their best use is producing controllers and, downstream, training data for pretraining robot models — mirroring how LLMs already interface with game/physics engines to generate training data for video models. The framing positions LLMs as a data flywheel for embodied AI rather than direct policy networks.
Related event: MIT Researcher: LLMs May Be Best at Generating Robot Pretraining Data(2 posts)→
More from Embodied
- Bedrock Robotics raises $600M as its fully autonomous excavators do paid work on live job sites — Sethwinterroth · 2026-10-27
- AW-0 hits 80.8% on RoboTwin tasks, beating strongest baseline by 16 points — trevordarrell · 2026-10-27
- AW-0's code stream fixes world model prediction: picks mango where baseline grabs apple — trevordarrell · 2026-10-27
- AW-0 runs at 230ms on a single RTX 5090, 2.4x faster than autoregressive code planning — trevordarrell · 2026-10-27
- AW-0 debuts: one generative world action model plans in code, predicts, and controls robots — trevordarrell · 2026-10-27
- Skyfall AI's SKY-001 world model turns sketches into CAD builds in 3.9s, vs 199s for GPT-6 Astra — rohanpaul_ai · 2026-10-27