Author: Huang Yu
Editor: Liu Baodan
Seventeen years ago, Apple released the iPhone, bringing smartphones into the era of intelligence. At that time, Nokia executives were still mocking Apple, thinking that this phone was a toy for children. Later, outdated companies like Nokia were eliminated by the times.
Now, the wave of AI large models is coming, bringing new opportunities to redefine human-computer interaction. For major smartphone manufacturers, this is another new starting line, which will definitely bring about a new round of reshuffling in the mobile phone industry.
This time, no one dares to take it lightly. Since the second half of last year, the AI large models of major domestic smartphone manufacturers are no longer just confined to &#34;PPT&#34; presentations, but have begun to accelerate their implementation, especially by integrating large models into new systems.
Following Huawei, Xiaomi, Vivo, and OPPO, Honor also officially released its self-developed operating system, MagicOS 8.0, on January 10, 2024. This system will incorporate a self-developed 7 billion parameter AI large model called &#34;Magic Large Model&#34; into the new Magic6 series phones. This is also Honor's first platform-level AI large model for end devices.
The development of AI large models in smartphones is still in the stage of cultivating user awareness. In order to seize the initiative, major smartphone manufacturers are also eager to demonstrate their uniqueness.
After being beaten by OPPO in launching the &#34;first end-side application with a 7 billion parameter AI large model&#34; gimmick, Honor emphasizes that MagicOS 8.0 features the industry's first intent-based intelligent terminal interaction paradigm called Intent-based UI (IUI).
Honor CEO Zhao Ming said that rewriting the underlying operating system is not what Honor pursues; it is the reconstruction of the operating system with AI.
The traditional human-computer interaction mode is based on the logic of iOS, with the machine as the center for resource allocation and interaction between human and machine interfaces. This was established by Steve Jobs in 2007. In Zhao Ming's view, 17 years have passed, and no one has imagined the next generation of operating systems. Honor hopes to provide a different answer.
According to the introduction, IUI supports multimodal interaction methods such as natural language, voice, images, gestures, and eye movements. It can intelligently recognize user intentions, make quick inference decisions, and proactively provide personalized services to users.
In simple terms, compared to traditional human-computer interaction, IUI aims to lower the threshold for using the operating system and improve interaction efficiency.
IUI benefits from the &#34;new kernel&#34; of MagicOS 8.0, which is the platform-level AI. Unlike application-level AI, platform-level AI can comprehensively enable the operating system (OS) as a technical foundation.
The traditional OS kernel is mainly responsible for managing hardware resources, while the &#34;new kernel&#34; of platform-level AI is responsible for &#34;managing&#34; factors related to humans, such as personal knowledge bases, location and status, habits and profiles, etc., helping the OS identify user intentions and schedule system services.
Zhao Ming pointed out that AI-enabled operating systems and AI-reconstructed operating systems have always been the ideal, or rather the strategy and goal, of Honor. In 2022, Honor launched the platform-level AI in MagicOS 7.0, which is the direction and concept. With MagicOS 8.0, this solution has been systematically implemented.
MagicOS 8.0 has built a large-scale model ecosystem for edge-cloud collaboration. The edge-side platform-level AI large-scale model serves as the central control, understanding user intent, decomposing and orchestrating tasks, while connecting to cloud-side large-scale models, distributing, integrating, and scheduling atomic services to complete complex task cycles.
In this ecosystem, the edge-side AI large-scale model is self-developed by Honor, which filters out personal privacy through the edge-side protection network and ensures that personal information does not go to the cloud, safeguarding privacy and security. For the cloud-side large-scale models, Honor chooses to cooperate with other companies, such as Baidu's Wenxin large-scale model.
It is understood that Honor has launched the &#34;Hundred Models Ecosystem Plan&#34; to provide an entry point for large-scale models, allowing cloud-side general large-scale models and various industry-specific large-scale models to be integrated into MagicOS through a shelf-like approach.
Compared with self-developed large-scale models on both the cloud and edge sides, Honor has clearly chosen a path with smaller investment.
Nowadays, there are mainly two paths for smartphone manufacturers to deploy large-scale models. One is to develop lightweight and locally deployed large-scale models on the smartphone side, mainly represented by Honor and Xiaomi. The other is to adopt a cloud-side collaborative architecture, deploying large-scale models at the level of tens of billions or hundreds of billions in the cloud, and deploying large-scale models at the level of billions on the smartphone side, launching a large-scale model matrix, such as Huawei, Vivo, and so on.
AI large-scale models have undoubtedly become the focus of competition for major smartphone manufacturers.
Chen Haijin, an analyst at Debon Securities, said that hardware represented by smartphones will not miss the era dividend of AI wave. The combination of hardware and AI is a revolutionary promotion of work and lifestyle, and is expected to give birth to a new &#34;Apple moment&#34;, with terminal hardware expected to become the entry point for AI.
Ivan Lam, a senior analyst at Counterpoint, told Wall Street News that the application of AI large-scale models is a topic that smartphone manufacturers must take seriously.
Major smartphone manufacturers also understand the importance of AI large-scale models for the future development of the smartphone industry.
Zhao Ming pointed out that the ability of AI, especially platform-level AI, is definitely a very important point for future product differentiation from competitors. OPPO's management even stated directly that companies that do not layout large-scale model smartphones this year will lose competitiveness in the market.
Wang Peng, deputy researcher at the Beijing Academy of Social Sciences, also told Wall Street News that in the fierce competition in the smartphone market, smartphone manufacturers can seek differentiation and open up new areas of innovation by self-developing AI large-scale models, thereby enhancing their brand image and market competitiveness.
Therefore, against the backdrop of the current bottleneck in software and hardware innovation in smartphones, the application of AI large-scale models is an important trend for major smartphone brands' future innovation, and it is also an inevitable strategy to continue seeking breakthroughs in the high-end road.
Exploring the application of AI large models is undoubtedly a correct but challenging path.
The first and foremost challenge is the cost of training large models. Zhao Ming revealed that Honor has accumulated RMB 10 billion in AI research and development expenses over the past few years. After entering the training of AI large models, the investment in computing resources has increased significantly, possibly doubling every year. Future investments in AI will also increase.
Currently, major smartphone manufacturers are still in the early stages of exploring AI large models, and the actual use cases have not been clearly defined. Voice assistants are still the main landing method.
Regardless, AI large models have pressed the accelerator for a new round of elimination in the smartphone industry.
At this important turning point in the industry, all major smartphone brands need to seize the opportunity of AI large models, break through the innovation bottleneck of smartphones' software and hardware, and create their own differentiated advantages to achieve overtaking or consolidate their industry position.
In the era of AI, standing still will eventually be eliminated by the market. Whether it is Honor or other smartphone brands, they cannot afford to be complacent.

AAPL

Reconstructing operating systems with AI.

Honour is going to challenge Apple's dominance.