Edge AI, the rising trend of mobile intelligent computing

11/22 2024 543

Over the past year, the hottest keyword in the consumer electronics industry has been Edge AI.

Large models possess powerful comprehension and generation capabilities, while terminal devices are the preferred interface for human-AI interaction. The combination of these two factors makes the edge side the best path for large models to be implemented, perhaps even the only one. More and more consumer electronics terminal manufacturers, including mobile phones, PCs, and automobiles, are applying AI technology to their products and services. The prosperity of Edge AI cannot be separated from a core support point—the computing platform (chip).

Edge computing power comes from the chip. Therefore, high-performance chips are the source of continuous prosperity for Edge AI and the direction Qualcomm hopes to promote through technological innovation and ecological cooperation.

Recently, the Snapdragon Summit was held, where we saw Qualcomm continue to strengthen its layout in Edge AI products and support for domestic large models. Not only did it announce collaborations with large model vendors such as Zhipu and Tencent Mixin, jointly promoting the deployment and implementation of Edge AI models. Additionally, it launched the Snapdragon 8 Ultra mobile platform for AI phones and the Snapdragon Auto Platform Ultra for smart vehicles.

Cristiano Amon, President and CEO of Qualcomm, stated that AI, especially generative AI, is one of the biggest changes in mobile computing. Qualcomm focuses on generative AI, particularly how AI transforms computing experiences, the development of edge computing, and how it brings disruptive innovation cycles, striving to bring AI-first experiences to every computing space.

The rise of Edge AI stems from the deep integration of mobile computing, represented by Qualcomm, with AI. We might as well start from Qualcomm's direction of creating AI-first experiences to explore the source power driving Edge AI forward.

At the beginning of 2024, consumer electronics categories such as PCs, mobile phones, and automobiles began actively integrating with large models to create new AI XX hardware, ushering Edge AI into a phase of concentrated explosion.

It can be said that large models have opened up a new trend in the consumer electronics field. Delving into the heart of this trend, we will see a crucial force driving the gestation and maturation of Edge AI—mobile computing.

The irreplaceable role of Edge AI compared to Cloud AI lies in large models performing calculations locally, avoiding data being uploaded to the cloud, and addressing user concerns about data and privacy security. It offers stronger real-time performance, with applications such as gaming, photography, and AIGC experiencing lower latency, faster feedback, and better experiences when processed on the edge side.

Entering 2024, first-tier model vendors like OpenAI, Zhipu, and Tencent Mixin have recognized the value and constraints of deploying large models on the edge side, successively launching smaller, higher-performance model versions. Although large models are actively being miniaturized, due to limitations in edge-side computing power and memory space, model downsizing and compression can compromise the performance of standard, general-purpose large models.

The edge-side deployment of large models heavily relies on edge computing platforms providing robust performance, supporting ultra-large-scale data computations, and achieving a balance between performance and power consumption.

Making large models better suited to the edge side has become a top priority in the field of mobile computing, an issue Qualcomm is actively addressing.

At the recent Snapdragon Summit, Qualcomm announced collaborations with Zhipu and Tencent Mixin to jointly promote the deployment and implementation of Edge AI models. From Qualcomm's new initiatives, we can observe three changes in mobile computing + large models:

Change 1: Richer Edge AI capabilities. The NPU performance of the Snapdragon 8 Ultra has improved by 45% compared to the previous generation. Its powerful performance allows multiple AI workloads to run simultaneously, meaning more AI applications based on large models can be deployed on the edge side.

Change 2: Higher-quality edge intelligence experiences. Qualcomm has equipped the Snapdragon 8 Ultra with the fastest LPDDR5x memory to address memory concerns. Currently, Tencent Mixin is collaborating with Qualcomm to promote the deployment of the 7B and 3B Mixin large models on terminals.

Change 3: A more robust Edge AI ecosystem. AI developers are indispensable for the continuous prosperity of Edge AI. Qualcomm has prepared efficient development tools based on its own hardware for developers, including the Qualcomm AI Stack and AI Hub. Collaborating with Tencent Mixin and Zhipu AI, it helps model vendors better unleash platform computing power on Qualcomm hardware, supporting AI developers in easily integrating AI models into applications.

According to data released by Qualcomm at the Snapdragon Summit, fully leveraging the powerful edge-side AI performance of the Snapdragon 8 Ultra and the performance optimization brought by the Qualcomm AI Stack to models, the Zhipu GLM-4V edge-side vision large model can run at speeds exceeding 70 tokens/second on the edge side. Optimizing the edge-side performance of large models not only reduces development costs and enhances product effectiveness for developers but also facilitates the commercialization of basic, general-purpose large models.

Large models are the starting point for the trend of Edge AI in this era. Mobile computing provides momentum for large models on the edge side in terms of more applications, better experiences, and a stronger ecosystem.

Qualcomm's strong support for large models supports the prosperity of the Edge AI industry from the source, embodying its commitment to driving AI-first experiences.

For large models to be utilized by users on terminal devices, they must transform from a technological trend into the industrial prosperity and commercial value of Edge AI. For large AI models to enter terminals, hardware performance is fundamental.

The head of an AI mobile phone brand we interviewed stated: "AI is just beginning in the mobile phone industry. In the coming year, the AI experience will surpass your imagination. Edge-side models have high performance requirements, so performance will never be excessive."

Where does performance come from? The answer is the computing platform. Currently, Edge AI is concentrated in two major categories: one is personal mobile terminals, such as AI phones and AI PCs; the other is intelligent connected vehicles, with AI applications accelerating in smart cockpits, assisted driving, autonomous driving, etc. Qualcomm introduced the latest technologies and products for these two platforms at the recent Snapdragon Summit.

Mobile Platform:

The Snapdragon 8 Ultra chip, with performance breakthroughs and innovative capabilities in technology, brings a significant leap in AI phone intelligence.

The Snapdragon 8 Ultra's breakthroughs in edge-side generative AI, multimodal processing, and image processing can support a plethora of AI use cases for AI phones. For example, the revolutionary breakthrough in AI-ISP technology, closely integrated with the NPU, makes real-time processing of high-resolution image data possible. Mobile phone manufacturers can introduce more AI applications in photography functions, such as automatic scene recognition, real-time beautification, and auto-focus, which is one of the product strengths currently valued by young consumers.

In addition, the Snapdragon platform also actively supports AI applications in the PC field.

At the recently held 2024 World Internet Conference, Qualcomm won the "WIC Leading Science and Technology Award" for its Snapdragon X Elite platform, specifically designed for Windows 11 AI+ PCs. One of the core advantages of the Snapdragon X Elite is its customized Oryon CPU, offering extremely fast response speeds and providing robust performance for AI PCs. For instance, Honor's newly launched AI PC, supported by the computing power of the Snapdragon X Elite, realizes AI functions such as AI meeting summaries and bidirectional AI noise reduction, significantly enhancing users' meeting and communication efficiency. Currently, the Snapdragon X series supports 58 PC products that have been released or are under development, effectively supporting the growth of this new AI PC category.

Automotive Platform:

The trend towards automotive intelligence is becoming increasingly prevalent, with smart cockpits and intelligent driving becoming core competitive points for automakers. To provide the underlying driving force for smart cars, Qualcomm has launched the new Snapdragon Ride Vision Platform Ultra and Snapdragon Ride Platform Ultra.

Among them, the Snapdragon Ride Vision Platform Ultra also adopts Qualcomm's self-developed Oryon CPU specifically designed for automobiles, significantly enhancing performance. It can help automakers integrate capabilities such as large models and AIGC with cockpits, creating new experiences like AI voice assistants, AIGC applications, complex situation awareness, and natural language interaction.

The Snapdragon Ride Platform Ultra, based on Qualcomm's most advanced NPU and Oryon CPU, brings better intelligent driving to automobiles, realizing a series of intelligent driving functions such as high-precision positioning, environmental perception, decision-making planning, and vehicle control.

It is worth mentioning that to enhance the product capabilities of AI phones, AI PCs, and smart cars, more powerful mobile computing hardware is only a necessary condition. In addition, manufacturers and developers also need a series of development efforts to unleash hardware performance and optimize Edge AI performance.

Currently, Snapdragon hardware platforms used in mobile phones, PCs, and automobiles have integrated Qualcomm's self-developed Oryon CPU. Coupled with the empowerment of the Qualcomm AI Stack, it is expected to achieve efficient development of edge-side AI applications on various platforms.

With the coordination of hardware (mobile platforms, PC platforms, and automotive platforms) and software (AI Stack, AI Hub), Qualcomm is removing the performance constraints of Edge AI, driving the accelerated evolution of various consumer electronics products such as mobile phones and automobiles, and seizing the technological dividends of large models.

In summary, large models are the starting point for this wave of Edge AI, and computing hardware is the source of performance that enables the new concept of AI terminals to take root. To support the sustainable development of large models and AI terminals, Qualcomm is also contributing to the prosperity of Edge AI from the source.

The reason for seeking the source power of Edge AI development today is consistent with the needs of domestic consumer electronics manufacturers to seize the opportunities and technological trends of the times. Close cooperation and coordinated progress with the source are conducive to gaining momentum for future growth.

Currently, Qualcomm is one of the important AI forces closely cooperating with domestic manufacturers.

On the one hand, Qualcomm and the Snapdragon platform are driving the globalization and high-end development of domestic Android mobile phone manufacturers. Leading domestic brands are directly competing with Apple in the high-end market. The close cooperation between Qualcomm, Snapdragon, and domestic manufacturers serves as a guarantee for the global competitiveness of domestic smartphone brands, enabling leading AI applications, AI phones, and AI PC hardware to set benchmarks globally. Currently, flagship phones equipped with the first batch of Snapdragon 8 Ultra, such as Xiaomi, Honor, and iQOO, have established distinct market recognition in terms of cost-effectiveness, imaging capabilities, and gaming performance.

Furthermore, the growth of new energy vehicles is expected to advance further.

Currently, domestic new energy vehicles have established strong competitiveness in the global market. Seizing the opportunity of intelligence and achieving greater development can further consolidate their advantageous position. The Snapdragon automotive platform serves as the cornerstone, supporting the integration of large model-based intelligent driving and smart cockpits into vehicles, enabling new energy vehicles to maintain their competitive advantage and tread more steadily and further in the global market.

For Edge AI, large models are the new-generation operating system. To effectively utilize this new operating system, full support from the ecosystem, tools, and resources is necessary. Through the Snapdragon software and hardware ecosystem, Qualcomm can provide ecological support for the development of Edge AI, helping more developers make good use of large model technology and promoting the overall prosperity of AI products and applications.

In summary, the market for consumer electronics is vast, and the intelligent upgrade has only just begun. AI will sweep through various familiar terminal categories. Standing at the forefront of the times, the role of Qualcomm and Snapdragon as source drivers is becoming increasingly clear.

The close integration of Qualcomm, Snapdragon, and AI brings about a lasting bond between mobile computing and Edge AI. The story of Edge AI, which has been on a rapid trajectory for a year, will continue to unfold.

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.