NVIDIA kicked off the press conferences at CES 2016 and did not disappoint. The team wowed audiences last night when they unveiled the Drive PX2 – an in-vehicle A.I. supercomputer, designed to make it possible to realize the self-driving car. The compact PX2 has more processing power than 150 MacBook Pros.
Through the realization of the self-driving car, NVIDIA plans to change the landscape with regard to safer driving, new mobility services, and urban redesign.
NVIDIA’s vision is to provide a complete computing platform that the entire auto industry can use to implement self-driving. This vision involves two aspects:
1) Mobility Teammate Concept (virtual copilot)
2) No driver at all
NVIDIA uses deep learning/deep neuro-net to train networks faster than has ever been possible. Networks are trained to recognize a multitude of shapes, to the point of super-human image recognition.
Their strategy for successfully realizing the afforemention vision is threefold:
1) Incorporate GPUs to accelerate the process of deep learning by 20-40 times. Training can be done in days instead of months.
2) Create computing platforms so anyone who wants to do so can.
3) Create end to end platform for deep learning; from the moment network training is started to deploying a network in a vehicle.
Volvo is the first company to partner with NVIDIA to utilize the deep learning vehicle computer.