At The Edge: Thinking Big & Small With AIAt The Edge: Thinking Big & Small With AI
At The Edge: Thinking Big & Small With AI
September 13, 2018
SAN JOSE - Making AI work for practical, everyday applications is a challenge. Part of the issue stems from the need for real time decisionmaking, which the latency of working with data in the cloud - where many AI platforms are based - necessarily inhibits. Between smartphones and the data-generating sensors of the IoT, there's a clear need for AI processing 'at the edge' - which will enable faster, real-time AI decisionmaking in everyday life.
To find out more about what AI at the edge can mean for businesses today, as well as the wider strategic questions and implications of artificial intelligence for businesses, we caught up with Jem Davies, Fellow, Vice President, and General Manager of Arm Technology, a chipmaker that provides processing architecture for 90% of all AI-enabled devices in the world.
What does AI at the edge mean in practice for enterprises?
The edge is where we all reside – not in the data center or in the cloud. Our devices are right here in our hands, in our homes and in our offices. Each of those devices has the potential to capture and process data in real time. If that data takes too long to process, it could be too old; it may prevent our devices from making immediate decisions and acting decisively to avert problems, maintain operational efficiency and provide instantaneous situational analysis.
AI at the edge means increased responsiveness, with improved security. Arm’s power-efficient processors support AI algorithms even for highly constrained, battery-powered edge devices such as wearables and sensors, providing the ‘intelligence’ needed to keep the processing on-device. Once you start doing that, rather than sending everything back and forth to the cloud, you quite naturally reduce latency – not to mention costs. And the less you shuffle your data back and forth, the less you’re exposing it to risk. User expectations on privacy and security mean that most people – and companies – quite rightly prefer to keep as much of their data as possible on-device. Edge compute restricts access and guarantees control.
How can large enterprises start thinking about their own everyday business problems in relation to AI solutions?
As we see AI introduced into more devices, we expect to see a world in which most ‘things’ are equipped with a new level of smartness; AI will have a massive effect on just about every segment I can think of, so the first step is simply to gain an understanding of what AI could do for you – and the answer is that AI applied to most situations will improve efficiency and user experience.
That said, enterprises need to think about the problem intelligently – pardon the pun – considering where their investment in AI can bring them tangible benefits, and a sound return on their investment. Whenever anything becomes a hot topic, there’s almost a pressure to become part of the buzz, to leap straight in and be seen to be active. But AI’s here for the long haul; there’s no rush. Don’t invest in AI per se, invest in a solution that tackles a business problem.
It’s also important to consider that, to be effective, AI needs data. Does your company have a sufficiently robust information infrastructure to allow you fully capitalize on the benefits AI can bring? Do you track the data that you will need?
Related: AI Is Nothing Without The IIoT
What are the key obstacles to making edge AI work at scale?
The whole thing about edge AI is that it’s naturally scalable: the available hardware multiplies with the number of people that have devices, and the data generated by those devices aren’t bothering the cloud at all.
There’s a famous quote from Google saying that if every Android device in the world performed three minutes of voice recognition each day, they’d need twice as much computing power to cope. In other words, one of the world’s largest computing infrastructures would have to double in size.
Ultimately, the world doesn't have the bandwidth to cope with transmitting all the data being produced – and the power and cost of transmitting that data to be processed in the cloud is simply prohibitive – so, wherever we are today, I think that, from necessity, there will be a natural shift to edge AI, and that will place an increased requirement in computing capability in all those edge devices.
What does competitive advantage look like in the context of AI today?
According to a recent MIT Sloan Management study, the apparent appetite for AI is not backed up by actual deployment. Curiosity and ambition significantly outweigh levels of execution. Almost 85 per cent of executives believe AI will allow their companies to obtain or sustain a competitive advantage, but only about one in five companies has incorporated AI in some offerings or processes, and only one in 20 companies has extensively incorporated AI in offerings or processes. Less than 39 per cent of all companies have an AI strategy in place. So, currently, simply having an AI strategy will put you ahead of 60 per cent of companies out there.
That said, AI can be used to help provide better experiences for customers, including digital experiences. In today’s world, customers often have options to go elsewhere with their dollars, so providing a more satisfying experience of interacting with your organization can be a powerful competitive edge.
How can enterprises meet the challenges of digital transformation for 2019?
Determine which parts of your organization can be improved by using AI. Part of the challenge is identifying the technology you need to meet your requirements; AI isn’t one-size-fits-all. Arm’s Project Trillium architecture allows you to select a comprehensive solution that’s tailored to your use case – from Arm Cortex-M processors for smart, connected embedded applications to purpose-built, ultra-efficient Machine Learning and Object Detection processors for on-device inference and neural network capabilities.
Once you’ve chosen which flavor of AI works for you, identify the data that will be needed to fuel that AI and ensure that it’s captured in a way that can be used by the AI later.
One of the key value propositions of the Internet of Things (IoT), for example, is the data that lies untapped in the billions of devices deployed around the world. But the complexity and diversity of so many different technologies, vendors, device types and connectivity protocols has meant that making effective use of it that data has, until now, been more of a vision than a reality. We’ve addressed that with the Arm Pelion IoT Platform– the industry’s first connectivity, device and data management platform for hybrid environments. It delivers the flexibility, security and efficiency needed to enable organizations to unlock real business value from IoT devices and data.
Once your data is under control, you also need to ensure that your computing infrastructure – including edge devices – is up to the task. Don’t be afraid to ask for guidance. No one expects you to have all the answers! Talk to people who are experts in the field. Even we may not have all the answers, but we’re here to help.
Jem Davies is Fellow, Vice President, and General Manager of Arm Technology
About the Author(s)
You May Also Like