There's no reason Tesla/MobilEye couldn't be using an off-policy reinforcement learning algorithm, no.