On-Device AI: The next generation in deep learning technologies

As we enter the age of the Quaternary Industrial Revolution, interest in artificial intelligence (AI) technologies, which learn and deduce the way we humans do, is rising. We’re expecting AI technology to find its way into an increasingly broader number of applications, from our smartphones and cleaning robots to washing machines. Today we’ll be looking into “on-device AI”, a new generation of deep learning technology set to take the industry by storm. AI in the palm of your hand

In conventional AI technology, data collected from a smart or mobile device is transmitted to a central Cloud server, where it is analyzed and sent back to the original device. However, as the name suggests, with on-device AI, the smart device both gathers and analyzes data without the need for a remote Cloud server. Because data is processed within the user device, on-device AI provides improved latency. And by doing away with a centralized server, on-device AI solves the security issues that have been pointed out for Cloud-based AI. The ability to provide AI services without a network connection means real-time translation and other tasks can be performed even without an Internet connection. Edge devices meet AI

In the classic AI arrangement, the Cloud server, which processes data, serves as the brain. The smartphone or tablet PC serves as the ears and eyes, interfacing with the user to gather data. But under this scheme, the actual AI is far away from the site where the data is being generated, limiting the ability of the AI to understand what is actually going on. But with on-device AI, edge devices, which are the hardware devices that come into direct user contact, incorporate artificial intelligence. The AI processes data gathered from environments it experiences firsthand, and can make more suitable decisions for the user. How on-device AI will change our lives Samsung Electronics is deeply involved in the research of innovative technologies such as deep learning algorithms and on-device AI. At the “Samsung AI Forum 2019” held in November, the company unveiled “on-device AI interpretation” technology, providing on-device interpretation without the need for a central server. Samsung’s Exynos 9 (9820) premium mobile application processor (AP) is a prime example of on-device AI. The processor incorporates a neural processing unit (NPU), which is a system semiconductor able to handle a variety of computations in real-time without delay. This makes AI processing up to seven times faster possible, all self-contained in the mobile device. Samsung Electronics aims to further boost NPU performance to support performance-hungry on-device AI applications. Progress in on-device AI technology will allow devices to develop a better idea of user diets and lifestyle patterns such as exercise, leading to further improvements in individually tailored user services. We look forward to the many positive changes that user-centered on-device AI technology will bring to our daily lives.