Pony AI founder and Chief Technology Officer Dr. Tiancheng Lou argued simulations and static datasets alone cannot deliver commercially scalable Level 4 robotaxis.
Lou made the case in a Financial Times op-ed, saying companies must collect massive amounts of real-world traffic data and build advanced "world models" because autonomous vehicles change how nearby drivers, cyclists and pedestrians behave.
“It is on the road, not in the lab, that the autonomous driving race will be won,” Lou states.
He singled out everyday ambiguity-hesitation, informal signals and partial commitments-as the hardest problems, giving examples such as a scooter cutting across a pickup point or a driver edging into a gap.
Related reading
- Anthropic opens Bluetooth API to let makers connect hardware devices to Claude
- Pony AI targets robotaxi cheaper than a Tesla Model 3 as Chinese operators race to cut costs
- China freezes all new robotaxi licences nationwide after Baidu fleet meltdown stranded 100 vehicles in Wuhan…
Lou said world models can help systems infer cause and effect from real-world interactions, for example how other road users react when a robotaxi slows, but those models cannot replace exposure to live traffic and the fleet operations that generate the necessary data.
The piece framed Pony AI, an artificial intelligence technology company focused on autonomous vehicles that says its PonyPilot service operates robotaxis in major Chinese cities, against a crowded field that includes Waymo-which said last year Waymo One provided more than 250,000 paid trips weekly across Phoenix, San Francisco, Los Angeles and Austin with expansion plans-and Tesla, whose Full Self-Driving push has drawn investor criticism.
