More than 90% of traffic deaths are caused by human error. The traditional theory of algorithm development, meanwhile, dictates that the larger and richer the data set the better the resulting algorithm. Tesla vehicles that use autopilot have begun to reach critical mass with an estimated total lifetime production of 200,000 electric vehicles. It is only with these higher autonomous miles that we begin to see shortcomings in the early algorithms for autonomous vehicles—mainly the challenge of identifying stationary objects that results in catastrophic crashes. The network effect—drawing big data from autonomous car use—will be important to designing the safest autonomous vehicles, which will, in turn, drive consumer preference. For manufacturers, autonomous algorithms mustn’t be a “winner take all market” because the developer of the best algorithms could resell its IP. As the data set underlying autonomous driving gets larger and richer, the quality of the autonomous driving should get better too.The data collected from miles driven with autonomous technology will begin to grow exponentially as the install base reaches critical mass—and that should rapidly accelerate the improvement of driving algorithms.