Ian Ferguson, vice president of ARM Global Marketing and Strategic Alliance, revealed that the strategy of next year is blunt: ARM is still too early to talk about AI, and will not launch a dedicated AI chip design next year, which will specialize in the IoT market. Artificial intelligence AI has become one of the most popular technologies this year, and has become a major focus of many large-scale technology and Internet companies, such as Google, Microsoft, Amazon, Facebook, etc., and even the processor leader Intel Painting will launch AI dedicated server chips next year to seize the AI ​​market opportunities. However, Ian Ferguson, vice president of ARM Global Marketing and Strategic Alliance, recently bluntly revealed that next year's strategy is blunt: "It is still too early to talk about AI, and will not launch a dedicated AI chip design next year, but will specialize in the IoT market." Development of ARM architecture AI chips will have to wait Although the AI ​​trend is sweeping the world, many large technology and Internet companies are beginning to combine AI with their own services for service purposes such as computer vision, speech recognition and robotics, and even more and more small-scale innovations. Or Internet companies, such as Api.ai, Drive. ai, etc., have begun to apply AI in a variety of industries. However, Ian Ferguson bluntly said, “It’s too early for ARM to talk about entering the AI ​​market.†Ian Ferguson explained that the AI ​​market is still quite fragmented, and the AI ​​technologies lack integration and have not yet developed into mainstream architecture. There are different ways. For example, some companies such as Nvidia use their own GPU card plus ARM processor to build an AI application platform. Other companies use a custom computing engine and a control processor. To develop AI applications, such as Google's development of custom computing hardware TPU for AI, other companies also use ARM architecture with Programmable Logic Devices to develop different AI applications. Ian Ferguson said that many hardware platforms that many companies use to develop AI applications will use the ARM solution to handle the control part. Take Nvidia's approach as an example. On their AI vision application computing platform, they usually The GPU is used to collect external information for arithmetic processing, and at the same time, it is also equipped with an ARM processor as a coordination function for the back-end control application. In the development of AI strategy, Ian Ferguson admits that next year will not introduce a special ARM core design for AI optimization or AI, but will focus on the establishment of basic control components, which can be put into the chip by hardware manufacturers. In order to assist customers to build an AI platform. He also expects that more and more companies will use ARM architecture processors to perform control processing functions when they develop AI applications next year. As for the part that requires a lot of parallel operations, the future will be handed over to other hard Completion, such as GPU or custom computing engine. In the long run, Ian Ferguson believes that more AI application examples are needed to decide whether to launch an AI-specific chip design. For example, many AI applications are currently being processed in the cloud, and the application will move closer to the future. On the terminal device or gateway device generated by the data, special ARM chip processing may be required. Unlike the entry into the AI ​​market, ARM is still very active in the IoT layout next year. Ian Ferguson said that next year's IoT industry development focus will focus on industrial and smart city applications. In order to expand the IoT industry layout, next year will have the opportunity to acquire some new IoT-related companies, and will also seek more business partners or The cooperation opportunities of the community, such as Arduino and Raspberry Pi, are all objects of future cooperation. It is hoped that through close cooperation with each other, more software developers can join the development ranks. Banknote Value Counting Machine Two CIS Suzhou Ribao Technology Co. Ltd. , https://www.ribaoeurope.com