- Baidu launched Apollo World, an AI simulation system it says can generate up to 100,000 km of virtual driving a day.
- The company is already using it to train its Apollo Go robotaxi fleet for global operations.
- The technology could help self-driving companies cut development costs, however, Baidu’s claim that one day of simulation equals three years of road testing has not been independently verified.
China’s Baidu has launched an AI system called “Apollo World” which generates virtual driving environments to train its Apollo Go robotaxis. Chinese and global self-driving firms continue to look for cheaper ways to develop autonomous technology, and the company announced the model at its AI Day in Shenzhen, Reuters reported.
A day of virtual driving, per Baidu
The headline claim is capacity. Baidu claims Apollo World can simulate up to 100,000 km (62,000 miles) of driving a day. The company added that this is equal to about three years of road testing by one self-driving car with a safety driver.
That comparison is Baidu’s own, however, and not any particular independent benchmark. The three-years-a-day figure refers only to simulated driving. It does not show how safe the cars are on real roads or whether they are ready for daily use.
The question is whether the simulations match the different roads and driving conditions in the cities where Apollo Go plans to operate. The company also needs to show how much more real-world testing the cars require.
Baidu added that it is already using Apollo World to train its Apollo Go robotaxis for use in other countries. Apollo Go is China’s largest robotaxi operator. Its rivals include WeRide, Pony.ai and Momenta Global.
Sensor streams, not just road scenes
Apollo World can simulate how roads and their surroundings change, rather than just create static images. It can also predict what might happen next. Live data from cameras and lidar sensors helps the self-driving software understand its surroundings and navigate them.
The model uses data from real road tests to train its system. It has reportedly recorded over 350 million kilometers of driving data. Apollo World also uses this data to create new road situations the cars have not faced before.
Baidu has also reportedly proposed a “2+3” standard for testing world models. If other developers adopt this standard, Baidu could help to set the rules for how these systems are judged in the long-term.
Baidu described Apollo World as an AI system that can understand and act on information from different sources. The company says it could also be used to develop robots, not just self-driving cars.
The cost problem driving the push
One of the biggest challenges in the robotaxi business is the cost of developing and running the technology. Many Chinese companies do not own all the cars they use. Instead, they provide the self-driving technology and work with carmakers, ride-hailing companies and other overseas operators. This keeps running costs down, but building and training the technology still costs a lot of money.
Frank McCleary, an expert in self-driving cars at consultancy Arthur D. Little, told Reuters that companies still spend heavily on training their AI systems and preparing vehicles. He said the speed of a launch depends on how well the system can learn to drive in a new place. He stressed that AI has helped companies to make even faster progress.
China’s robotaxi market is poised to grow quickly in a short period. Morgan Stanley expects the country to have 1 million robotaxis by 2035, Reuters reported. XPeng, Volkswagen’s partner in China, also plans to launch its robotaxi service in 2027. Tesla is taking a different approach with its Cybercab fleet, using AI models that are trained on videos of people driving.
The smartest crypto minds already read our newsletter. Want in? Join them.
Opeyemi Olanrewaju
Opeyemi specializes in creating and refining high-quality content focused on cryptocurrency, global financial markets and the economy. He graduated from the University of Ibadan with an MBBS degree. He has worked as Editor-in-Chief for his College’s editorial publication and previously at CFA. For over six years, he has helped safeguard uniqueness as news editor at Cryptopolitan.
