OpenAI disbands its standalone robotics team, moving its staff to research and Applied units, saying models like GPT-6 Astra can be directly applied to robotics
OpenAI has disbanded its standalone robotics team, reassigning its researchers to general AI and applied product units. The company states that its frontier models, such as GPT-6 Astra, can now be directly applied to robotics, eliminating the need for a separate specialized robotics team. While industry observers note that general-purpose models show strong performance on spatial and reasoning tasks, some experts argue that specialized models and deployment data remain necessary for real-world robotics. Open article →
Notable comments
Commenters largely agree embodied robotics can use frontier LLMs for high-level goals, but emphasize hard real-time control, local safety/latency constraints, and multi-layer architectures as reasons standalone “direct control by LLM” is unrealistic today.
HN · dwattttt At ~30ms per "choice"/output, you can choose what lower level motors/servos/systems drive towards, but you can't generate the signals that drive them.
HN · gopalv The idea is probably not have a frontier model doing this, but to make a frontier model build these out of its own brain (well, directly edit the weights instead of RL tuning).
HN · jacquesm More like high level motion controller, so 'servo 23 1700 us' or for better grade servos 'servo 23 pos 145400', possibly concurrent with motion commands to other actuators and possibly synchronized, possibly using other kinds of protocols (G-codes).