PowerAI democratizes deep learning for customers

By Sejuti Das Jul 5th 2017
PowerAI democratizes deep learning for customers

With PowerAI combined with Power Systems will help businesses around the world to deploy a fully optimized platform for the deep learning with enhanced performance, says Sumit Gupta, VP-Cognitive Systems at IBM.

Businesses around the world are looking to transform ways to deliver value from information technology, which is creating new demands on enterprise IT infrastructure. Organizations need to align their technologies according to business demands, drive innovation by discovering value from data and reduce cost by securely deliver business services.

And, therefore organizations are aggressively focusing on the pervasive use of machine learning, deep learning, and natural language-based sensory data, which is driven by human to machine, and machine to machine interaction at a massive scale. To maximize IT flexibility and drive efficiency, integration of cognitive and deep learning with systems infrastructure have become absolute imperatives for both large and small businesses.

On this note, Sumit Gupta, VP-Cognitive Systems at IBM, talks about how IBM is focusing on delivering technology innovation by infusing the prowess of cognitive in its PowerSystems servers.

How do AI innovations transform the way companies design, build and deliver services? How is IBM focusing on delivering technology innovation by using cognitive in PowerSystems servers?

With artificial intelligence reaching a tipping point where it's being integrated into our daily lives, there are still major challenges for the data scientists and developers dealing with the technology. And therefore, IBM has created a multi-faceted strategy towards AI.

With the new PowerAI model, deep learning understands the historical data and recognizes patterns to detect frauds instantly. PowerAI further democratizes deep learning for users to improve upon tasks like fraud detection, and crime analysis.

Sumit Gupta, VP-Cognitive Systems at IBM

Watson and PowerAI are complementary technologies, however, the later one includes machine learning and deep learning. Watson is a platform which is focused on APIs (application program interface), and probably has the best voice API in the world. It is used to fundamentally change application interface or framework. The second trend--open source AI, that is gaining traction in the community is focused more on the machine and deep learning and is being generated by open source AI. These are the open source packages that act as building blocks to build new cognitive applications.

Currently, several organizations, especially banks are using open source AI in order to detect frauds and threats. The organization gathers all possible information of the customers and trains their system to understand the fraud and its tricks.

Therefore, IBM has built a software offering, using open source frameworks, like TensorFlow, CAFFE, Torch, etc and built the platform called Power AI. It is a software product that is optimized for running on a cognitive system from IBM. PowerAI gives higher level tools that make the system easier and automated. With this new AI model, deep learning understands the historical data and recognizes patterns to detect frauds instantly. PowerAI further democratizes deep learning for users to improve upon tasks like fraud detection, and crime analysis. It also helps in reducing the complexity and risk of deploying these open source frameworks for enterprises on the power architecture.

Advancements in artificial intelligence are leading to significant innovations in hardware technology, making robust hardware the next great frontier. What are your thoughts?

IBM believes that the path forward for cognitive infrastructure is an accelerated datacenter with enhanced servers and GPU accelerators. We, at IBM, are working towards accelerating the network and storage as well. Considering organizations are currently working on compute intensive tasks, every piece of datacenter system has to run much faster to support these AI technologies. Therefore datacenter and hardware are getting revolutionized, which is making them the next great frontier.

We, at IBM, working in close proximity with NVIDIA, in order to embed high-speed connection--NVLink, between the IBM POWER8 CPU and the NVIDIA Tesla P100 GPU accelerator, and it is made available only on PowerAI which gives us a unique advantage from a server perspective. Another new interface that we are working on next generation chip is called OpenCAPI, which is a new connection that gives organizations a high bandwidth, low latency, connection for memory, accelerators, network, storage, and other devices like ASICs.

Along with some major investment and innovation towards such technologies, IBM is also partnering with the ecosystem, supporting the new OpenPOWER software environment. Power8 is one such RISC microprocessor that supports OpenPOWER. It was designed to deliver unprecedented performance for emerging workloads, such as business analytics, big data applications, cloud computing, and scale out data center workloads. We also provide open interfaces in our CPUs to help other technologies vendors or partners to innovate and add value to the servers and the data center.

Understanding data has always been a huge challenge for most companies, be it in India or global. Your thoughts?

Currently, organizations have figured out ways to bring their data into a data lake, and have created proper data lake strategies, which was the initial challenge for the majority of organizations. Now, the next piece of this puzzle is to use that data to get useful insight, and that is what we are focusing right now. Watson is one such platform, that enables developers to draw insight from their data lake, using APIs. But, PowerAI makes it even easier for customers to gain insights from their data, because of the layman approach it creates, which is also known as an easy button for machine learning.

“In banking and financial industry, organizations are using deep learning and PowerAI for fraud detection, credit risk analysis, anti-money laundering pattern detection, chat bots for customer service and algorithmic trading.”

Our objective with PowerAI is to make the journey to AI as easy, intuitive and productive as possible. It not only reduces the frustration of waiting but also increase the productivity of the system. Power Systems are designed for next era of computing, in great contrast to x86 servers which were designed for the client/server. PowerAI now includes CAFFE, Chainer, TensorFlow, Theano, Torch, NVIDIA DIGITS, and several other machines and deep learning frameworks and libraries.

How is the industry innovation, with respect to deep and machine learning, capturing growth against different vertical segments?

Undoubtedly the decline in business profitability across multiple industries is starting to threaten future investments, innovations, and stakeholders' value. But, the new factor of production--advanced artificial intelligence is going to help organizations kick-start their profitability. The rush of AI advancements is working on the confluence of three factors, which are combined together to create the balance for AI growth--big data, powerful graphics processing units, and decades-old AI computation model along with deep learning.

“In the retail industry, organizations are using advanced AI for gaining insights on point of sale analysis, environment and weather forecasting, situation analysis, personalized shopping.

Each of the vertical segments, whether be it banking, financial institutions, telecom, retail, and government, are trying to do different things with AI to enhance business profitability. In banking and financial industry, organizations are using deep learning and PowerAI for fraud detection, credit risk analysis, anti-money laundering pattern detection, chat bots for customer service and algorithmic trading.

On the other hand, in the retail industry, organizations are using advanced AI for gaining insights on point of sale analysis, environment and weather forecasting, situation analysis, personalized shopping, staffing, optimizing supply chain, omnichannel shopping, managing product inventory, and chatbots for delivering better customer service.

IBM also sees a lot of activity happening in the manufacturing industrial space. Organizations are using drones for industrial inspection and automating the backend by using deep-learning-based computer vision. They also use AI for crunching production time, managing demand-side constraints, improving repair and overhaul, gaining relationship intelligence, and revolutionizing product and service quality. IBM is also working with clients who wish to build robots for either managing manufacturing or delivering customer service.

LATEST interviews