Chip designer ARM has announced two new designs aimed at delivering AI capabilities to a wide array of devices. The first is a machine learning (ML) processor that can be included in devices such as smartphones or tablets and optimized to carry out machine learning tasks that would otherwise bog down the CPU or GPU. The second is an object detection (OD) chip design that is suitable for continuous monitoring of an area on devices with lower-power budgets, such as security cameras, as well as on other high-power devices.
ARM designs will bring AI to all smartphones and into the smart home
Ovum and the wider electronics industry have been waiting for ARM to unveil its AI solution for some time now. Dedicated AI chipsets already exist in the Apple iPhone X and the Huawei Mate 10, leaving companies relying on ARM-designed chips from the likes of Qualcomm feeling left behind. Apple and Huawei’s ability to design their own chips gave them a head start, but the focus so far has been on first-party implementations by the manufacturers themselves rather than on providing an open platform for developers.
ARM’s designs will change this as these are the designs that the AI industry centers itself around and that developers will learn to expect. This is simply due to ARM’s incredible reach in the mobile market. ARM-designed chipsets power every smartphone including Apple’s. As a result, ARM designs help provide a common approach to hardware that should mean that programs wanting to take advantage of AI features on a device are able to, regardless of the device or chipset manufacturer. Ovum expects that ARM’s designs will gradually spread across the industry and lead to the wide adoption of AI capabilities in smartphones over the next five years. Ovum expects the first chipsets based on these designs to come to market in late 2018 with devices following shortly after.
The OD chipset is very exciting from a smart home perspective. The key challenge in delivering AI performance to most smart home devices is that the devices in question have much less processing power to work with. The OD chip meets this challenge for a very specific niche use case. Object detection and recognition is one of the most valuable tasks for a smart home AI device because of its impact on security systems, so this makes it an obvious first choice for ARM to develop. Ovum expects that ARM will follow up the OD design over the next few years with similar single-use designs in areas such as natural language recognition.
On-Device AI: Demand Analysis, Chipset Vendor Strategies, and Future Outlook, CES004-000007 (January 2018)
Daniel Gleeson, Senior Analyst, Consumer Technology