NVIDIA is a major innovator in the AI space, building computing platforms focused on redefining product development and business intelligence.
As Senior Director of Business Development for Deep Learning and AI at NVIDIA, Kimberly Powell focuses on the AI ecosystem and has seen the number of organizations NVIDIA is enabling grow from a couple of dozen three years ago to more than 4000 today.
“Deep learning”, Kimberly says, “has led to the latest AI revolution – the algorithms are data driven and accessible with GPU computing. NVIDIA’s GPU is available on all our computing platforms and we are accelerating every deep learning framework”.
With deep learning at the heart of NVIDIA’s vision, we spoke to Kimberly to find out why deep learning is the hottest topic in AI right now, as well as hear about their latest products and the industries they are working in – all ahead of her keynote at The AI Summit in San Francisco on 28-29 September.
“Deep learning can do a lot of things that machine learning and computer vision cannot do”, Kimberly says.
“In 2012, deep learning algorithms beat human-engineering computer vision algorithms in the image recognition challenge. In 2013, deep learning beat humans in detecting breast cancer. In 2015, deep learning beat human at image recognition challenge. In 2016, AlphaGo beat the best human Go player, Lee Sedol, by applying deep learning.
“The complexity of such problems are near impossible for a human-engineered algorithm; instead, deep learning can learn from data and result in an algorithm with far more depth and accuracy. Industries are able to put their big data to use with deep neural networks, and changing the way businesses run and products are developed.”
Like AI Business, Kimberly believes that AI is triggering the on-set of the fourth industrial revolution, which will disrupt almost every business, across every industry. She explains why GPU computing is “a natural fit” with these developments:
“The onslaught of big data, and the ability to harness it, requires tremendous computing power. Deep learning algorithms are billions of software neurons and trillions of connections trained in parallel – a natural fit for GPU computing. Our portfolio of deep learning solutions is positioned to address all needs – whether you want it accessible on a PC, in the data center, or in the cloud”.
So what is the most recent deep learning innovation NVIDIA has built?
“Most recently, we created an AI supercomputer-in-a-box called the DGX-1 and it’s built with eight of our latest Pascal GPUs. Pascal introduced five miracle technologies that boosts deep learning by 65x from just 4 years ago. The DGX-1 is a plug-and-play server with integrated hardware and software that has the computing power of a 250-nodes making deep learning accessible to every academic lab, AI startup, and data scientists in F1000 companies. We just delivered our first DGX-1 to Elon Musk’s OpenAI so they can build autonomous agents like chat bots, autonomous cars and robots that are accessible to everyone”.
Rather than focus NVIDIA’s solution towards any particular industry verticals, Kimberly remarks that “forward-looking companies in almost every industry are adopting deep learning”:
“This is to address exponentially increasing amounts of data by exploiting improvements in machine learning algorithms and advances in computing hardware. In turn this is helping them find new ways to tap into the wealth of data and to develop new products, services, and processes – and create a disruptive competitive advantage”.
But in terms of applications of AI software, she says that NVIDIA see the healthcare and transportation industries as “leaders of the pack”:
“Both industries have incredible human life at stake, every nation is seeking state-of-the-art technology to improve patient outcomes and make driving safer. We’re tackling self-driving cars head-on – it requires learning, reasoning, and planning – all in real time. We just announced a self-driving AI solution in partnership with Baidu for all carmakers”.
“More global industries, include security and defense organizations, are using deep learning for face recognition, video surveillance and cybersecurity. Media and entertainment is using it for video captioning, content-based search and real-time translation. Internet services companies are using it for image and video classification, speech recognition, recommendation engines and natural language processing”.
Within industries themselves, Kimberly points out that there are countless examples of how F1000 companies, researchers, and startups are using deep learning right now to accelerate innovation and transform their bottom lines. She cites a few key ones, including those that implement NVIDIA’s technology:
“In the medical industry, understanding the human genome holds enormous promise for personalized medicine, but genomic biology generates a torrent of data that humans cannot grasp. Deep Genomics deep learning-based model can accurately predict disease-causing variants in DNA. Their work has led to insight into the genetic basis of autism and cancer.
“In finance, banks process millions of transactions per day, but many can only use a small sample to model fraud. Skymind is using distributed deep learning to create more accurate models for fraud detection that can learn from massive amounts of data and improve as they go.
“In robotics, Sergey Levine and Pieter Abbeel at the Berkeley AI lab are using reinforcement deep learning to enable a robot to learn fine motor skills — like hanging cloth, opening bottle caps, putting a square peg in a square hole. BRETT is basically developing hand-eye coordination through trial and error. Such robots will revolutionize manufacturing. They also recently received a DGX-1 to accelerate their research.
In cybersecurity and graphistry, too much data slows down the process of identifying issues with cybersecurities for big companies. Using NVIDIA technology, the graph analysis cloud platform is able to help the company’s response and hunting team sift through 100M+ alerts a day. Customers can visually correlate what’s going on in their systems and beef up their security in seconds rather than a few hours. For the first time you can see all your data, from your users to software services.
“Online fraud is another example, with PayPal the beneficiaries of deep learning. Deep learning algorithms are able to analyze potentially tens of thousands of latent features (time signals, actors and geographic location are some easy examples) that might make up a particular type of fraud, and are even able to detect “sub modus operandi,” or different variants of the same scheme. Deep learning will give the PayPal team the ability to adapt to these new patterns faster than before. It’s possible, for example, that PayPal might someday be able to deploy models that take live data from its system and become smarter, by retraining themselves, in real time.
NVIDIA certainly appear to be leading the charge with deep learning in AI computing.
Kimberly Powell will deliver her keynote on ‘The Deep Learning Revolution’ at The AI Summit in San Francisco on 28-29 September.
She will be joined at the event by fellow CxOs from the world’s leading enterprises and the most exciting AI software developers, gathering to explore the huge opportunity that AI presents all industry verticals.
To find out more, and to register to attend the event, visit: theaisummit.com