The pandemic has highlighted the benefits of, and need for, autonomous vehicles (AVs). The path to satisfying regulators and the American public that these vehicles are as safe as, or safer than, human drivers, remains unclear. It’s estimated that AVs will need to drive 11 billion miles in order to reach the same error rate as human drivers. Since this is impractical, simulations are being used to test and teach AVs in highly realistic virtual environments. These simulations have the advantage of enabling automakers to train AV systems on edge cases generated by artificial intelligence - situations that rarely, if ever, happen, and that could be prohibitively dangerous to create and test on roads with human drivers.