Overview
- Pony.ai, which announced the upgrade Friday, says PonyWorld 2.0 marks a new phase in how its AI improves itself.
- The system adds a structured intention layer that records why the driverless stack chose an action and checks the result to find weak scenarios.
- It then issues specific data-collection tasks to human teams, who gather those real-world cases and feed them into a cloud training loop to refine the model.
- Pony.ai says the upgrade is live across its Level 4 driverless fleet and R&D operations to improve safety, ride comfort, and traffic flow.
- After reporting breakeven unit economics in two major Chinese cities, the company targets more than 3,000 vehicles across about 20 cities by the end of 2026, with nearly half overseas, and these claims come from company materials without independent validation in the coverage.