Ecological Dialogue | Aixin Yuanzhi: Large model + AI new paradigm creates more value for the development of end-side and edge-side intelligence
Click the Arm Community above to follow us
With the vigorous development of artificial intelligence (AI) in various fields, the demand for low latency, high bandwidth, privacy, etc. on the edge side is also growing day by day, prompting more computing power to gradually migrate to the edge side and realizing many applications. applications in the field. Accelerating the release of the unlimited potential of AI on the terminal and edge sides means that smart chips need to carry algorithms and adapt to various application scenarios. As an AI visual perception chip R&D and basic computing power platform company, Aixin Yuanzhi has made useful attempts for the industry to explore the implementation of AI and large models on the terminal and edge sides. In this issue, Arm will cooperate with its partner Aixin Yuanzhi Semiconductor. dialogue to discuss the practice of deploying large visual models on the end side and edge side.
interlocutor
Ma Jian, Vice President of Business Development of Arm IoT Division ( hereinafter referred to as: Ma Jian)
Liu Jianwei, co-founder and vice president of Aixin Yuanzhi Semiconductor ( hereinafter referred to as: Liu Jianwei)
Q1
Ma Jian: As more and more AI applications expand to the edge, the process of intelligent development in various fields is accelerated. Can Mr. Liu share with us his insights or market development trends in the deployment of AI on the terminal and edge sides?
Liu Jianwei: Indeed, the development of semiconductors has helped artificial intelligence gradually move to the edge and terminal sides, transforming it from fixed terminal scenarios represented by smart cities and smart homes to mobile terminal scenarios such as smart driving. The technical needs of these three major fields are interlinked, and they all also require visual perception and intelligent computing. As a platform chip company that supports the development of edge-side and end-side intelligence, Aixin Yuanzhi has always focused on this.
In recent years, with the development of AI, the level of intelligence and demand for products in cities, homes, such as sweeping robots, home cameras, smart speakers, etc., as well as driving products have continued to increase. In addition, the computing power of chips has continued to improve, making edge intelligence gained popularity. According to our observations, smart cities are currently undergoing continuous upgrades, L2/L2+ smart driving applications are being implemented on a large scale, and the demand for intelligent terminal equipment is also expanding. We believe that the smart chip market with sensing and computing capabilities has the opportunity to usher in larger-scale development and market opportunities.
Q2
Ma Jian: Aixin Yuanzhi is one of the pioneers in combining large models with AI vision chips, and has rapid practice and deep insights in the field of vision. Can you share your company's latest progress in this regard, and how cooperation with Arm can help develop this process?
Liu Jianwei: We are the first to demonstrate the integration with large models on our flagship AI vision chip AX650N. This chip deploys Swin Transformer at the edge. It has the characteristics of high performance, high precision, easy deployment and low power consumption. It is the industry's leading Transformer landing platform. It also implements the Arm Cortex-A55 cluster for pre-processing and post-processing, which is optimally combined with our Aixin Tongyuan mixed-precision NPU and Aixin Zhimo AI-ISP.
In the field of smart vision, we are very pleased to see that Arm has launched a new smart vision reference design for local customers this year, which further meets the differentiated processing needs in multiple scenarios and improves the load of machine learning and device processing work. performance, helping to accelerate the innovation and implementation of intelligent vision products. Our Aixin Zhimo AI-ISP combines deep learning algorithms with traditional ISP processing units to achieve better image quality. As a general-purpose processor for AI computing, Aixin Tongyuan NPU is designed with operators as atomic instruction sets, which can provide surging computing power for AI applications at a limited cost. Therefore, we also look forward to strong cooperation with Arm in the integration of IP in the future to inspire more innovative sparks for intelligent vision application solutions.
In addition to product technology, the Arm ecosystem with rich resources also allows us to have no worries during the product development process. We can focus on Aixin's own core technology, achieve continuous innovation, and complete the goal of "inclusive AI creates a better life" Company mission!
Q3
Ma Jian: I am glad that more local companies like Aixin Yuanzhi have joined the Arm ecosystem, bringing more vitality and innovation power to the entire ecosystem. Finally, tell us about your company’s development plan for next year?
Liu Jianwei: Aixin Yuanzhi’s vision is “AI universalization” to provide convenience to the public and create a better life. Next year, we will focus on two directions: First, from the overall product layout, we will continue to consolidate the development path of one body and two wings, continue to expand in the fields of smart cities and smart driving, and continue to meet customer needs in different scenarios; second, Increase efforts to promote ecological construction and establish a "Love Core Circle of Friends". We hope to create an ecological environment of co-inclusion and co-creation, and work with more developers to optimize product experience. This is why we chose to join the "Arm Ecosystem".
* This article is an original article by Arm. For reprinting, please leave a message to get authorization and indicate the source.
京公网安备 11010802033920号