The smartphones that no longer rely on the cloud or server systems can perform their own AI functions directly, and the future may no longer be a dream as an era of 'on-device machine learning' is opening up. Quoting industry sources, it is expected that operators such as Samsung Electronics, Huawei and MediaTek will be rolling out mobile application processors (machine tools) that support machine learning starting in 2018. For some time now, machine learning has passed High-end CPUs, graphics processors (GPUs), etc. Qualcomm, however, has been using a mobile system single-chip (SoC ) To efficiently and automatically perform machine learning with the goal of mounting machine learning chips on devices such as smartphones or robots, etc. In other words, this will allow devices equipped with Qualcomm's Snapdragon mobile chipset to capture images, objects and faces Classify and identify, if these functions can be implemented inside the device, future devices will no longer rely on AI services provided by the cloud or communication network. At present, more than 99% of global mobile chip architectures are in the hands of Arm (ARM). In 2017, Arm announced DynamIQ, a mobile chip architecture with machine learning function, adding an AI accelerator inside the chip. Compared with the latest mobile processor Theoretically, the functionality is expected to increase by more than 50 times. Arm Korea executives stated that smart devices such as smartphones and PCs are commonly used by consumers. If a viable AI ecosystem emerges, the device can process data faster, and Arm's goal , It is to enable devices equipped with Arm architecture chips, such as smart phones, to perform fast and efficient machine learning. Many mobile AP operators expected to design according to the standard architecture of Arm will launch new products in 2018. Samsung, MediaTek and other global leaders As for the mobile AP industry, the architecture of Arm is adopted at present, while for Qualcomm and Apple's AP, the architecture of Arm is partially changed and the architecture basically used is the same. If the mobile device officially entered the AI era, A great opportunity. Korean scholars pointed out that based on deep learning of devices, it is an inevitable trend of the times. Korean companies should Think of it as an opportunity to think hard about the various possibilities. For the mobile device industry, it can also be regarded as one of the antidote to the stagnant growth of the current market, and the dramatic leap in the function of mobile processors will drive software evolution. The demand for high-speed DRAM and high-capacity NAND Flash will also increase. On the other hand, most of the semiconductor architecture designs that support machine learning for mobile devices are still in the hands of Qualcomm, Arm, in the medium to long term, Samsung, LG Electronics. ) The license fees paid by South Korean smartphone makers are likely to increase. Related sources in the Korean semiconductor industry say that Samsung pays Arm for licensing fees each year. If Arm first breaks through the field of machine learning in mobile devices, it will be Samsung’s commitment to Arm’s. Dependence, I am afraid that will be deeper than now.
