Intel AI Summit 2020: The Future of AI for Enterprises

BrandPost By Intel
Oct 05, 2020
AnalyticsArtificial Intelligence

ai vendor relationship management artificial intelligence hand on virtual screen
Credit: ipopba / Getty Images

“Nineteen per cent of organisations are not using AI at all, and 54 per cent are still limited to trials, evaluations, and proof of concepts,” Nash Palaniswamy, General Manager, AI and HPC Solutions, Intel, said at the opening keynote of Intel AI Summit 2020, highlighting just how much of an opportunity and value there still is for AI to deliver value to a wide range of enterprises in every sector.

Intel is helping organisations realise the value of AI across ecosystem, software and hardware solutions, Palaniswamy added. Intel’s AI solutions encompass five key areas:

  • Intel® AI Builders – in which Intel works with ISVs and systems integrators to develop solutions that accelerate the adoption of AI across verticals and workloads.
  • OEM systems – in which Intel provides read-to-deploy, end-to-end solutions that have been jointly designed and delivered by Intel & partners.
  • Public clouds – where Intel works with cloud service providers such as Amazon, Baidu, Google and Microsoft to deliver AI-optimised solutions.
  • Intel® Select Solutions – a series of AI-optimised solutions designed to meet real-world demands, and are pre-configured and rigorously benchmark-tested to accelerate infrastructure deployment.
  • Intel® oneAPI Toolkits– this is a unified, standards-based programming model designed to help developers improve efficiency and achieve innovation by providing tools needed to deploy applications across any architecture – XPUs (including CPUs), GPUs, FGPAs, and other accelerators.

Intel has developed this approach to AI based on its vision of AI that will become ubiquitous soon. Intel’s Exascale for Everyone vision, introduced by Raja Koduri, Chief Architect, Senior Vice President of Architecture, Graphis & Software, Intel, in The Path To AI Everywhere keynote, is built on the premise that there will soon be 100 billion intelligent connected devices, and the drive towards that milestone is doubling the compute demands worldwide every 3.4 months.

In the same presentation, Wei Li, Chief Machine Learning Software Architect, VP of Machine Learning & Performance, Intel, overviewed three AI tends that Intel has observed in the drive towards this hyper-connected, AI-driven future:

  • Models – AI is getting smarter. This is being achieved by making the models bigger (ELM0 just a few years ago featured 94 million parameters, now GPT-3 boasts 175,000 million). In addition, there is the effort to run AI work efficiently so AI can proliferate in low-powered devices such as a watches or cameras.
  • Use Cases – The field of AI is evolving rapidly. The breakthrough first-iteration AI focused on solving computer vision use cases. Now AI covers many more use cases such as natural language processing and recommendation systems.
  • End-to-End – For enterprises to deliver AI solutions they need to collect data, conduct feature engineering (sometimes via machine learning), then train the AI via deep learning before finally deploying the model. An AI solution requires machine learning, deep learning and analytics all working together.

Addressing the challenges of AI

Enterprises understand the drive to AI, and it sits at the core objective of most transformation exercises. The availability of data is there, and compute and storage is affordable in a way that is unprecedented. There are also plenty of use cases, frameworks and models to base AI deployments on. Most organisations, however, are still in the proof-of-concept phase, or are using AI in a limited number of business processes. Progressing AI beyond the hype is still a key conversation piece in the development of the technology.

That was a key theme that was discussed in a fireside chat featuring Sumner Lemon (Director of Digital Transformation, Intel), Foo Wui Ngiap (Head of Technology, Grab), Glenn Gore (Global Head, Solutions Architecture, AWS), Simon Thomas (Global Head of Data & AI, Avande) and Parviz Peiravi (Global CTO/Principal Engineer, Financial Services Industry Solutions, Intel), there remains some hesitancy across industry when it comes to adopting AI, and an uncertainty about the expected results of AI projects.

“People think that AI is a miracle pill,” Ngiap said. “They expect radical results in a short time frame. One of the biggest lessons we’ve learned is that you’ve got to start small, and you’ve got to grow.

“There’s a myth around machine learning and AI,” Gore added. “It can be very confronting, and the first inclination can be to push it away – I would encourage enterprises to instead get hands-on with it. It’s not as complex as many think it is.”

As Thomas noted, AI is so common now many people are using it without even realising it. “My fridge and microwave have AI implemented into them,” he joked. That ubiquity means that the barriers to AI adoption that many might assume are commonplace are simply no longer there. “The hype around AI started six or seven decades ago, but there hasn’t been any time in history where all the requirements for AI are in place like they are now. The most powerful capabilities are available to everyone thanks to the cloud. It’s democratised at this point.”

Intel can take an industry thought leadership position on AI, having undergone the transformation and implementation process itself, Archie Deskus, Senior Vice President & Chief Information Officer, Intel said in a later interview session.

“It’s been more than ten years. It takes time,” Deskus said. “There’s cultural change to manage, and you’ve got to give the process the time to drive the success. This is not something that happened overnight. It evolved organically, and for us the process started with business intelligence.

“The first challenge was around collecting reliable connected data. We were challenged by having unreliable or siloed, unconnected data. Moving to data lakes required IT to go back to invest in a process to ensure we had the proper collection of data.

“We encountered some resistance in fear and uncertainty in some parts of the business, and our approach there needed to be to go slowly. Creating the pull to deliver AI successful across the organisation required education both from below and on top.”

As a result of these AI deployments, Intel has been able to achieve everything from improvements to the internal work and processes, through to leveraging AI to scour the Internet to find new customers in need of its solutions. Intel has truly been able to use AI to transform its entire business.

For more information on how AI can empower your enterprise and deliver meaningful benefits across the organisation, watch the Power Your Enterprise With AI webinar, in collaboration with Intel and CIO. Click here to watch on-demand.

The entire Intel AI Summit 2020 is available for playback OnDemand. To catch up on any sessions you may have missed or wish to watch again, click here.