NVIDIA’s Future Lies In AI, Cars And Deep Learning

NVIDIA’s Future Lies In AI, Cars And Deep Learning

You can’t have Computex without NVIDIA, and the annual tech conference wouldn’t be the same without a conference or two from the GPU giant. But while Pascal and the GeForce GTX 1070 and 1080 are drawing a lot of headlines right now, what’s really intriguing is the world with which NVIDIA has inextricably tied its future to.

That future, according to the GPU giant’s CEO and co-founder Jen-Hsun Huang, is in three parts: artificial intelligence, deep learning and driverless cars.

“100 of the world’s top supercomputers are accelerated by NVIDIA GPU. The NVIDIA GPU called Tesla is the most successful supercomputing, high performing computing accelerator in the world today.”

The reminder from Jen-Hsun Huang, NVIDIA chief executive and co-founder, is a handy reminder that NVIDIA doesn’t see themselves as just a GPU manufacturer. But while PC gaming remains a valuable target market for the company, one rapidly growing sector is deep learning.

“Software can now write software,” the NVIDIA CEO explained. “A computer can learn to write software. It can learn how to predict, it can learn how to recommend, it can recognise information, it can learn how to recognise voice and language, pictures, videos, the computer has figured out how to learn.”

Deep learning appeals to NVIDIA because it’s a high intensity, high performance application. That’s precisely the kind of market the company wants to target.

NVIDIA also showed off their DGX1 supercomputer, the fastest supercomputer per node in the world. It runs off eight Tesla P100 processors, and according to Huang it would take $USD1 million worth of servers to match the performance of the DGX1. NVIDIA isn’t looking to be a server vendor in the long term, however, and from 2017 we’ll see OEM offerings equipped with the Tesla P100.

The other big ticket appeal for NVIDIA is the burgeoning field of driverless cars. Huang pitched NVIDIA’s Drive PX2 chips for AI-powered, driverless cars as not just being an advancement for auto-pilot technology, but also for society.

“With this little processor, we can achieve amazing levels of auto-pilot capability. You’re still the pilot, but this is like cruise control like you’ve never imagined before. From the moment you enter a freeway, it will drive completely by yourself. Because it’s paying attention to all the cameras around the car, it pays more attention than you do. And hopefully we can help save lives.”

He went on to say that he wasn’t aware of a single medicine being invented today that has the potential to save over a million people per year. Given that NVIDIA spent several billion dollars on the Pascal architecture, it makes sense that they want to target future markets.

$USD2 billion alone went towards the development for AI, but with the increased adoption of deep learning across various industries — particularly hyperscale data centres — it’s a bet that seems likely to pay off for the company.

It’s already starting to become the battleground between the tech giants. Amazon has already begun testing the offering of AI services to companies through rentable computers using NVIDIA’s new Tesla chips. Google is already targeting clients with their machine-learning services in a bid to gain market share off Amazon and Microsoft.

So the question remains: if this is going to be the era of the AI, how long will it take before an AI can mimic the full capacity of the human brain?

According to the NVIDIA CEO: decades.

The author travelled to Computex 2016 as a guest of Intel.