GTC 2017: Nvidia heralds the age of AI

• AI has become irreplaceable in today’s world and has great potential for growth
• Announces initiatives for new Volta GPU platform, robotics and AI for enterprise

NVIDIA’S GPU Technology Conference (GTC) 2017 reached its climax on day three when its chief executive officer Jensen Huang (pic) took to the stage to deliver his keynote speech to a packed auditorium.

Huang made a bold statement that the era of artificial intelligence (AI) has arrived and Nvidia’s efforts will shift towards the creation of new supercomputers

He sets the scene where hundreds of millions of people rely on AI-powered search, language translation and speech recognition services. This has all fuelled demand for more AI computing power.

Adding to that, in the last year alone the investments in AI startups have risen to US$5 billion, a clear indication that there is a vast amount of interest in the space.

“Artificial intelligence is driving the greatest technology advances in human history,” said Huang during his GTC keynote. “It will automate intelligence and spur a wave of social progress unmatched since the industrial revolution.”

The biggest announcement coming out of the keynote was definitely Nvidia’s new Volta GPU platform, which it says is the world’s most powerful GPU computing architecture that has been specifically made for handling artificial intelligence and high-performance computing (HPC).

Volta is Nvidia’s seventh-generation GPU architecture. It was built with 21 billion transistors and Tensor Cores that are designed to speed up AI workloads. That translates to being able to deliver the equivalent of 100 CPUs for deep learning, so Nvidia said.

Demand for accelerating AI has never been greater as developers, data scientists and researchers are increasingly relying on neural networks to power advancements in fighting cancer, making transportation safer with self-driving vehicles and providing intelligent customer experiences.

To put Volta’s performance gains in perspective, the company says that the new architecture provides a 5x improvement over its previous GPU architecture, Pascal and by 15x against its two-year-old Maxwell architecture.

Nvidia followed up by announcing its first Volta-based processor, the Tesla V100 data centre GPU, that brings speed and scalability for AI computing and training, as well as for accelerating HPC and graphics workloads.

This huge leap in computing power came at an investment of more than US$3 billion on the part of Nvidia, a massive bet that it hopes will pay off.

With data centres demanding for greater processing power, they need to effectively scale to support the rapid adoption of AI-based services. This includes natural language virtual assistants, such as Google Assistant or Amazon’s Alexa, and personalised search and recommendation systems.

“Deep learning, a groundbreaking AI approach that creates computer software that learns, all demand for processing power. Nvidia engineers have spent three years creating Volta to help meet this trend, enabling the industry to realise AI’s life-changing potential,” he said.

Bringing AI to enterprise

Businesses have collected mountains of data daily and many struggles to cope with this large volume of data. Often times, it is stored on a disk and soon forgotten.

But Nvidia intends to make the case that it can, in fact, move those mountains of data. The company announced during GTC that it is partnering with enterprise software company SAP to create a new wave of AI business applications.

SAP is said to leverage on the advancements made by Nvidia in its GPUs, systems and software, particularly its new SAP Brand Impact solution that measures brand attributes such as logos in near real time with incredible accuracy.

Powered by Nvidia deep learning, SAP Brand Impact is made up of deep neural networks trained on Nvidia’s DGX-1 deep learning system to provide video analysis. Equipped with eight Volta GPUs, the DGX-1 sells for US$149,000.

Huang provided an example during the keynote showing how efficient the AI was at instantly tracking, identifying and capturing the brand logo placement on fast moving vehicles during a motorsport event.

The results are practically immediate, accurate and auditable, all delivered within a day.

Do the Robot

Nvidia also announced that it wants to usher a new era in the robotics industry by introducing its Isaac robot simulator, the name an obvious an homage to science fiction writer Isaac Asimov who penned the popular ‘Law of Robotics’.

The concept behind Isaac, which is built on an enhanced version of Epic Games’ Unreal Engine 4, is that utilises simulated real-world conditions rendered by video-game graphics technology to train intelligent machines before they are deployed.

“Robots based on artificial intelligence hold enormous potential, but building and training them has posed challenges due to time and budget constraints, Huang said.

Prior to deployment, robots need to extensively train and tested with physical prototypes, which can be costly and time-consuming.

By simulating these tasks and interactions in a highly realistic virtual environment that have the same laws of physics so that it knows how to behave in the real world.

Not constrained by time and space, Isaac can be used to jump start a robot’s brain by running multiple instances of the training process and drastically cut down training time.

Nvidia also introduced a set of robot reference design platforms, which are basically a generic set of robots that either traverse land, air or water, so that software developers need only focus on coding their robot rather than worrying how to build one.

The VR connection

It is clear that Huang is a big fan of collaboration so it comes to no surprise that Nvidia’s announced its new Project Holodeck, a VR collaborative space that is an obvious nod to the simulated reality facility featured in the science fiction series Star Trek.

Now, Nvidia’s Holodeck isn’t exactly like its science fiction counterpart that transports users into a fully immersive simulated environment.

Instead, it is more of a collaborative tool within a virtual reality environment that creators can import high-resolution models into VR to collaborate and share to make design decisions easier and faster.

Holodeck is built on an enhanced version of Epic Games’ Unreal Engine 4 graphics engine. Together with Nvidia GameWorks, VRWorks and DesignWorks that delivers photo-realistic graphics and physics simulation, designers can easily import high quality render models without simplifying models for VR.

The demonstration showed how a group of engineers could explore a Koenigsegg Regera supercar at full scale and able to consult with each other on the design changes in real time.

Nvidia expects Holodeck to be used as a collaborative tool for product designers and architects. It also believes that it would be used by enthusiasts wanting to enjoy 3D and 360-degree content within a VR space.

Chong Jinn Xiung reports from GTC 2017 in San Jose, at the invitation of Nvidia. All editorials are independent.

Related Stories:

GTC 2017: Nvidia aims to spread knowledge of Deep Learning

AMD: It’s all about Games and VR

Culture shift imperative in cloud adoption: HSBC CIO

For more technology news and the latest updates, follow us on Facebook,Twitter or LinkedIn.