Physical AI is becoming a major next step for the field. Instead of only generating text, images, or code, these systems need to understand movement, space, timing, sensors, and real-world consequences.
General Intuition’s work with video game action data, research into physical training signals, and Nvidia-linked AI research all point toward the same goal: foundation models that can transfer more easily into robots and real-world machines.
Why This Is Hard
- Robots face environments that change constantly.
- Real-world training data is expensive and slow to collect.
- Small errors can create safety risks.
- Models need to connect perception, planning, and movement.
If physical AI works, robotics could move faster. If it fails, companies will rediscover why the real world is harder than language.
Will robotics get its own foundation-model breakthrough?