As the auto industry rushes headlong into autonomous vehicles and technologies there are some important learning curves ahead. Google admitted as much when one of its self-driving cars was involved in a mild fender-bender with a bus…not necessarily the fault of the self-driving Google car, but no doubt caused by reacting to an unfolding situation in ways different than a human driver would react. Humans understand that mass-heavy buses do not always yield right-of-way. On-board computers wouldn’t necessarily know this unless taught.

Teaching autonomous cars how to anticipate the actions of human drivers in varying real-world scenarios is critical, and this kind of deep learning is data-intensive. This is being addressed by many companies including video game-notable NVIDIA, which works with automakers on advanced electronics systems.

The company’s new DRIVE PX-2 graphics processing unit (GPU), the world’s first in-car artificial intelligence supercomputer, aims to provide 360-degree situational awareness and facilitate the deep learning required for cars to sense their surroundings and navigate autonomously, using processing power equivalent to that of 150 MacBook Pros. DRIVE PX-2 delivers up to 24 trillion deep learning operations per second, over 10 times more computational horsepower than the previous-generation product.

It’s a dangerous world out there with road debris, varying weather conditions, and unpredictable drivers. These are just some of the challenges as autonomous cars use artificial intelligence to drive better than humans. Unlike video games, there are real consequences on the road and supercomputer power like this will help keep autonomous drivers…um, passengers...safe.