tech
April 16, 2026
Physical Intelligence, a hot robotics startup, says its new robot brain can figure out tasks it was never taught
The new model, called π0.7, represents what the company describes as an early but meaningful step toward the long-sought goal of a general-purpose robot brain.
TL;DR
- Physical Intelligence's new model, π0.7, can direct robots to perform tasks it was not explicitly trained on.
- The model exhibits compositional generalization, combining learned skills to solve new problems.
- This capability is a step towards a general-purpose robot brain that can learn new tasks through plain language coaching.
- The AI demonstrated an understanding of an unfamiliar air fryer, even cooking a sweet potato with verbal instructions.
- The research highlights the importance of prompt engineering in guiding the AI's performance.
- While not yet capable of complex autonomous tasks, the generalist model matches previous specialist models' performance on tasks like making coffee and folding laundry.
- Researchers were surprised by the model's emergent capabilities, noting it was difficult to trace the source of its knowledge.
- Physical Intelligence has raised significant funding and is valued at $5.6 billion, with discussions for a new funding round.
- The company is cautious about commercial timelines, with researchers emphasizing these are research results, not a deployed product.
Continue reading the original article