TL;DR
The hardest part of teaching a car to drive itself is the rare, dangerous moment, and real cars can drive a long time before they meet one. Baidu says its new AI system, Apollo World, builds a living copy of real roads and can churn out up to 100,000 km (about 62,000 miles) of simulated robotaxi driving a day, which it equates to roughly three years of testing by one real car with a safety driver aboard. It even feeds the driving software fake camera and lidar data in real time, so the car "sees" a world that does not exist. Baidu says the simulator is already training its Apollo Go robotaxis, which it plans to run around the world. If practice gets that cheap, the robotaxi rollout already spreading from Austin to London could move a lot faster. For the full odds and the timeline of breakthroughs that gets driverless rides into more cities, subscribe.
Below the line, for paid subscribers: The full story: where it stands, how we got here and what's dated next; our labeled speculation on what it adds up to; and every source.