Nvidia Corp. and Baidu Inc. today announced they’re teaming up to bring the processing power of Nvidia’s latest graphics chips to artificial-intelligence services in the Chinese Internet giant’s services, in particular self-driving cars.

The two companies said at Baidu’s annual Create conference in Beijing that the two companies are looking to AI to a number of products and services ranging from cloud computing to self-driving cars and home digital assistants.

“We see AI transforming every industry,” Ian Buck (pictured), Nvidia’s vice president and general manager of accelerated computing, said on a press call. “We aim to democratize AI everywhere.”

The companies announced four areas where they’ll work together:

* The next-generation Nvidia Volta graphics processing units will be available on the Baidu Cloud.

* Baidu will use Nvidia’s Drive PX platform exclusively for its self-driving car initiative and develop those cars working with Chinese automakers. The two companies have already worked together on Baidu’s Apollo self-driving car project.

Among the 50 partners on the Apollo project are vehicle manufacturers that include Chery, FAW Group, Changan Automobile, Daimler, Ford Motor Co. and Great Wall Motors. Parts and service suppliers include Bosch, Microsoft Azure cloud, ZTE, Velodyne and TomTom, as well as ride-hailing companies UCAR and Grab Taxi. Lu Qi, Baidu’s chief operating officer, also said the company will will set up subsidiaries in the United States and Singapore to advance Apollo.

Baidu is aiming to get fully autonomous cars on highways and major roads by year-end 2020.

* Baidu’s PaddlePaddle, its platform for deep learning that it open-sourced last August, will be tuned for the Volta GPUs. Deep learning neural networks try to roughly emulate part of the way the brain works to allow computers to learn on their own rather than being programmed specifically. The branch of machine learning has been responsible for recent breakthroughs in speech and image recognition and robotics.

Buck said the current Tesla P100 offers 10 teraflops of mathematical horsepower per GPU. With Volta, a new core called TensorCore ups that to 120 teraflops per GPU. “That’s going to give a dramatic boost to AI performance for PaddlePaddle and all AI frameworks,” such as the Google Inc.-developed TensorFlow, Buck said.

* Baidu’s new Duer AI assistant, similar to Google and Amazon.com Inc.’s digital assistants, will be used in Nvidia’s Shield streaming video device in the China market. Nvidia said a new version of the streamer that includes custom software for China, will be available later this year.

Photo: Nvidia