What could possibly go wrong, one may ask, when development of real autonomous cars turn to a computer game such as Grand Theft Auto for help. The idea isn’t as ludicrous as it may seem, because there aren’t enough hours in a day to rack up mileage in the real world in order to teach cars to drive on their own, according to a Bloomberg report.

“Just relying on data from the roads is not practical. With simulation, you can run the same scenario over and over again infinitely, then test it again,” said Davide Bacchet of Nio, the startup company of EP9 fame, which aims to bring an autonomous car to the US market in 2020.

It’s said that the benefits of computer game simulation are two-fold: the software in games are able to generate data that is very close to what will be encountered in real life, and the virtual environment allows simulations to be conducted much more quickly than in real-world exercises.

The game “is the richest virtual environment that we could extract data from,” said Alain Kornhauser, a Princeton University professor of operations research and financial engineering, who advises the Princeton Autonomous Vehicle Engineering team.

Meanwhile at Toyota’s research institute in California, its engineers try to simulate the toughest conditions possible by running what’s known as the Quick Brown Fox test: extensive tests conducted in the most challenging weather and traffic conditions.

Despite obvious limitations, the human brain is still superior to a computer when it comes to assessing real-world risk, such as anticipating a child running in chase of a ball that has bounced across a street. There lies the challenge for all parties aiming for a market lead in autonomous mobility – to make computer systems better and safer drivers than humans.

Simulation should be “an acceptable equivalent to real-word testing,” albeit followed up with validation, Gill Pratt, chief executive officer of the Toyota institute told a House Energy and Commerce subcommittee. It would appear that this is a path increasingly adopted by developers of autonomous systems, with the apparent benefits of time-saving progress.