This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 3099067.
SANTA CLARA - Enterprises the world over are looking to AI for new answers to age-old problems. Beneath the results, though, there's a decent chance that those answers are being powered by an NVIDIA product.
NVIDIA has had a good couple of years. That could be putting it lightly. In fact, everyone from global ecommerce giants Baidu and Alibaba to Silicon Valley leaders such as Amazon, Facebook, Google, and Microsoft have adopted NVIDIA's GPU technology to power their AI solutions. This year, they even claimed to have built the fastest computer humanity has ever created.
To find out more about this central role that NVIDIA are playing in the AI revolution, we caught up with Paul Bommarito, VP of NVIDIA's Americas Enterprise Business, ahead of The AI Summit San Francisco.
Along with many new hardware announcements such as the DGX-2, AI was a big talking point at GTC 2018. How important is AI to the future of NVIDIA?
AI is very important to the future of NVIDIA, because as AI matures, there will be more demand for strong compute.
Another way to put this is that NVIDIA is essential to the future of AI. For 30 years, the dynamics of Moore’s law, which is the observation that the number of transistors in a dense integrated circuit doubles about every two years, held true. Microprocessor performance advanced at a rate of 50 percent per year as more and more transistors were fit onto a single chip. But that approach is reaching the limits of semiconductor physics, and today, CPU performance only grows by 10 percent per year.
The AI revolution has arrived despite the fact Moore’s law – the combined effect of Dennard scaling and CPU architecture advance – began slowing nearly a decade ago. Dennard scaling, whereby reducing transistor size and voltage allowed designers to increase transistor density and speed while maintaining power density, is now limited by device physics.
NVIDIA GPU computing has given the industry a path forward — and will provide a 1,000x speed-up by 2025. Most AI innovation is based on deep learning, which is a collection of statistical algorithms that leverage neural networks to gain insights fast. In order for deep learning-enabled solutions to thrive, they require massively parallel GPU power to crunch through iterations of calculations. That’s why we created DGX-2, the world’s largest GPUs, for our customers’ increasing parallelled AI workloads.
Why should large enterprises consider AI? How can they start thinking about their own everyday business problems in relation to AI solutions?
Large enterprises need to consider AI, because the computation power is now available for enterprises to build innovation, automate processes, improve the customer experience and develop intelligent agents to help with repetitive business processes. AI is the only path forward to measure return on investment and efficiency. Those who don’t start implementing AI will be left behind. AI is transforming every industry – healthcare, financial services, retail, telecommunications, oil and gas and automotive. The boom in AI development will lead to greater demand for hardware and services among growing AI developer communities.
NVIDIA enables companies to have the tools to retain AI talent. Instead of waiting for deep learning models to be trained and come back with results several months later, NVIDIA’s fastest GPUs allow data scientists and AI developers to see their model results within hours to days. Now the valued innovators can shorten deep learning training runs, iterate faster, get insights sooner, and improve productivity. These AI talents are rare and hard to come by. The last thing we want is to lose them over not having the infrastructure supporting their creativity.
Businesses need to assess three key areas before they can determine the right deep learning solution.
Understand your capabilities in terms of resources and staffing
Do you have an in-house AI team? To get started, businesses need to evaluate their capabilities. Leveraging initial deep learning work from others as a starting point is ideal. If your AI team is already experienced, you need to define which problems you want to solve first. You can focus on data preparation, neural network design to develop solutions that provide the highest accuracy and separate yourself from the competition.
Look to relevant use cases that match your situation
For example, are you in medical research and wondering where you should start with AI? We have found many use cases for Image Classification and Object Detection around X-rays and tissue samples. Also, there is work is being done to help identify patients with a potential for future medical issues. Or, do you work at a large financial services institution hoping to leverage time series predictions to identify future instrument values or anomaly detection to identify potential fraud? Every enterprise should start working with your deep learning talent to identify which use cases best match your organization and desired solutions, then recreate a method that was already proven to be successful.
Leverage available data and manage your data as well.
New deep learning networks depend on lots of data and complex models need more data than a simple model. Many customers are integrating external data sources to increase their capabilities. The faster you can process the data, the faster you can pull the insights, so investing in powerful hardware optimized with the right software only makes things easier for your data science team and expedites your company’s AI adoption. NVIDIA GPU solutions make it easy for enterprises to tackle these complex tasks in a timely manner.
What are the key obstacles to making AI work for global enterprises?
Vision -One of the first obstacles lies in building a vision for a focused AI strategy. Once there’s a vision, then it’s about setting the SMART goal. It’s important to come up with a few pilot projects, put together goals that are as specific, relevant, attainable, measurable, and timely as possible. Generating multiple ideas up front allows you to leverage the experience of the gathered team, as well as decide on a strategy to pivot quickly in case results are not forthcoming from the first pilot.
Data - Having the right data strategy has always been one of the most challenging steps. Data is the experience from which models learn. Without the right data, it’s hard to build models that can generate the right insights.
People -If you’re in a big organization, it’s important to have your executives and data scientists buy in on the goals. Together, you can look for areas or use cases that can benefit from deep learning, in support of solving a pressive challenge or tapping into a new opportunity.
Platform - Most companies’ data centers, where the algorithm training must take place, run on servers with traditional processors. This is hardly surprising, given that machine learning has only recently verged on mainstream business operations. An enterprise that intends to transform itself using machine learning will need to invest in the necessary combination of hardware and software to tap the vast promise of AI.
All told, the nascent business opportunities enabled by massive data collection and the implementation of algorithms will require rethinking the data center. Without investments in enterprise IT infrastructure, machine learning can’t deliver what it promises.
A critical step toward business transformation is to make sure an organization’s data center can support compute-intensive workloads. Those managing a company’s data center infrastructure need to ensure they have enough accelerated computing power and storage to handle all the data needed. This involves evaluating the whole picture to understand the incredible savings that can come from modernizing your architecture for the AI world.
Business leaders who perform due diligence to ensure their hardware is a match for their company’s machine learning ambitions will quickly understand the value of GPU computing.
How can enterprises meet the challenges of digital transformation in 2019?
Have a strategic AI vision -AI can’t solve all the problems. It takes a team of visionaries who understand the challenges in the organization and the capabilities of AI to further identify the use cases that will lead to successful digital transformation
Build a data culture -It’s important to understand how data is curated, collected, and stored. If there’s a strong data culture and strong data driven initiatives, lots will be change fundamentally.
Be mindful of regulatory - Now with GDPR establishment, there will be more and more regulations that will be in place to further protect customers’ data. That means enterprises need to understand what data can be collected and how long they can be stored for.
What does competitive advantage look like in the context of AI?
Domain knowledge in industries will help organizations better understand the data and context of problems/challenges. Outsourcing can be expensive and slow, so possessing in-house knowledge is a differentiation.
Fast computation power is going to make a tremendous difference in the AI space. Now with the GPU acceleration, some researchers said their work is doable within their lifetime. For such a transformational technology advancement, that means those who take advantage of that will be ahead of the game significantly.