DigiTimes had an interview with NVIDIA CEO Jen-Hsun Huang about deep learning and the company's solutions for autonomous cars. You can read it over here. Huang explained he's very optimistic about NVIDIA's autopilot solutions and mentioned that NVIDIA's self-driving solutions can receive verification by most car makers because the firm worked together with car makers to establish a telematics architecture as well as an application platform.
Q: What are the prospects of Nvidia autopilot solutions since a number of automotive companies have decided to deploy Nvidia's products to develop autonomous cars?
A: Since safety has always been the primary concern we have maintained engagement with the telematics market from the outset. Prior to the presentation of our self-driving solutions, we worked together with each car makers to establish a telematics architecture as well as an application platform pertaining to safe driving. This is why our self-driving solutions can receive verification by most car makers.
To be successful, an autonomous vehicle must be able to react decisively to the different situations, or road and weather conditions it may encounter. Passively handing over the decision making processes to cloud-based data centers is not a good practice.
Nvidia's in-house developed supercomputer solutions can collect data and sense outside conditions and then calculate the best solution for driving safely. Connectivity with the cloud is mainly used to study historic trends or to adapt to a driver's behavior patterns.