This year’s Chinese mobile phone market is undergoing major changes.
The two major variables have converged and begun to show their influence, making the market hot again:
Advertisement
One isHuawei's full comeback is unprecedented.
two isAI has stirred up the situation and become a battleground for military strategists.
Under this blend, the throne of the first quarter has now been assigned to—
glory.
Advertisement
AI becomes a key growth engine for mobile phone manufacturers
Huawei's comprehensive return adopts the most low-key posture and has gained the highest level of attention in history.
There was a heated discussion in the market that after Huawei's return, “Honor will be the most affected.”
But as the latest data comes out, this is not the case.
The latest quarterly mobile phone tracking report released by IDC (International Data Corporation) shows that in the first quarter of this year, China’s overall smartphone shipments were approximately 69.26 million units. Specifically, in terms of manufacturers,Honor takes first place with 17.1% market share“has become an important force in driving the market to continue to improve.”
Wang Jiping, Vice President of IDC China, believes that,With this glory reaching the top, AI has become a key growth engine:
At the beginning of this year, Honor released the AI-enabled full-scenario operating system MagicOS 8.0. It is blessed with a large magic model and uses platform-level AI as the core and base. Smart service. The first Magic6 series equipped with MagicOS 8.0 has been well received by the market.
According to Wang Jiping, thanks to the increase in AI functions and all-round upgrades in imaging, screens, etc., the shipments of the new flagship Magic6 series in the first quarter exceeded the combined shipments in the first two quarters of the previous generation;
On the other hand, driven by the Honor Magic V2 equipped with platform-level AI and other folding screen family products, the share of Honor folding screen mobile phones has increased by as much as 675.4% year-on-year since last year.
Relying on the performance of the Honor Magic6 series and the folding screen family, Honor's share of the high-end market above US$600 increased significantly in the first quarter of this year, with shipments increasing by as much as 123.3% year-on-year. Its high-end market share is second only to Apple and Huawei.
Mobile AI≠AI mobile phone
Then the question comes——
AI mobile phones are a topic that every terminal manufacturer must cue after the big model application storm begins. In the first quarter, major domestic and foreign Android manufacturers that made aggressive bets launched their first products one after another.
Everyone is ambitious, and the bottom line is of course that we cannot lag behind in this competition.
Why is Glory the one that stands out?
Honor CEO Zhao Ming believes that the key is“Mobile phone AI is not equal to AI phone”:
For the technical capabilities and experience of using AI on mobile phones, mobile phones provide platforms and algorithms.
But only when AI becomes the foundation and is ubiquitous on mobile phones can it be called an AI mobile phone.
Looking at the AI mobile phones that appeared on the market in the first quarter, it is not difficult to see that although the promotional words are tightly bound to “big models” and “AIGC”, there are actually differences in genres in the specific implementation routes.
The route chosen by most manufacturers is to take the lead in focusing on applications. Based on AIGC technologies such as text generation and image generation that have been refined in the cloud, AIGC applications for device-side and device-cloud collaboration are launched on mobile phones. For example, more intelligent portrait cutout, translation, meeting summary, etc.
Manufacturers represented by Honor are doing the opposite.
Zhao Ming has publicly stated on many occasions that mobile phone manufacturers should avoid falling into the trap of the explosion of applications. In the field of AI,The core value that terminal manufacturers can bring to consumers lies at the operating system level.
Specifically, the latter route actually starts from first principles, considering that the core issues that consumers really care about are:
Can the vigorous large-scale model and AIGC really bring a better and more difficult-to-replace practical experience to terminal devices?
So, what is the experience that big models can actually enhance when they go into mobile phones?
The most intuitive thing is to move from “people looking for services” to “services looking for people”.
At the level of technological exploration, Qualcomm has demonstrated the ability of smart assistants to proactively make decisions based on user habits: when planning a travel route, it will proactively ask if you want to buy a cup of coffee at a passing coffee shop.
On a practical level, you can take Honor as an example:
First of all, MagicOS 8.0, with platform-level AI as its new core, brings the industry’s first human-computer interaction based on intent recognition.
It can identify human-computer interaction intentions and is a system-level active service.
It can be when traveling, when you arrive at the subway station, the bus code will automatically pop up on your phone; when you arrive at the airport, your boarding pass will automatically pop up; when you arrive at the cinema, the ticket collection code will pop up.
You can also notify the leader of a meeting in the group. Just take a screenshot and click “Create Schedule”, and the system will be able to identify the time, location, and meeting topic, and create the corresponding schedule in the calendar.
One of the core functions, the “any door” based on multi-modal capabilities, can complete “one-step direct” transfer between apps.
Similar to this, there are many more experiences brought by any door. To sum up, it simplifies the traditional multi-step operation into a one-step operation and achieves the goal in one step.
Behind the user experience, Honor also more clearly proposed the AI four-layer architecture theory of terminal devices:
The first layer is at the system level, using AI to enable cross-system and cross-device integration. That is to say, different operating systems can be seamlessly connected through AI decision-making to realize data sharing between mobile phones, tablets, PCs and other devices.
The second layer is at the stand-alone level, using AI to reconstruct the operating system. Based on artificial intelligence technology, the terminal will understand you better the more you use it, and the more you use it, the better it will be.
The third layer is at the application level, that is, the application of AI on the terminal side.
The fourth layer is AI device-cloud collaboration. That is, on the premise of ensuring user privacy and security, network-side AI such as AIGC and cloud large models can be presented on mobile phones.
In fact, this four-layer architecture also outlines the current competitive status of the AI mobile phone market:
The third and fourth floors are easier to penetrate and are the first to usher in the outbreak.
However, the first and second layers are more critical to terminal manufacturers and are the core values that differentiate terminal companies from Internet companies.
The next phase of mobile phones: smart agents
To make a simple summary using the currently hotly debated technology words, these two different application routes can be divided into “tool school” and “intelligent school”.
The so-called tool school pays more attention to the tooling of single-point AIGC technical capabilities, emphasizing the direct efficiency improvement brought by tool changes in specific task scenarios.
The “intelligent school” pays more attention to the in-depth integration of large models, AIGC capabilities and terminal operating systems, which means to give full play to the capabilities of large models in perception, memory, understanding, reasoning, and decision-making, and to continuously learn user habits and understand user intentions. Various capabilities of the terminal are scheduled around user needs.
Compared with launching AIGC applications to solve a certain type of problem, the layout of “Intelligence School” appears to be more long-term. The core is people-centered and meets consumers' essential needs for mobile phone intelligence:
On the one hand, simplifying the human-computer interaction model allows terminals such as mobile phones to become more and more like human assistants, proactively adapting and matching people's needs.
On the other hand, it can make full use of the large amount of user personalized data retained in the terminal, allowing the intelligent assistant to evolve and continue to grow according to user characteristics.
In fact, this is also the key to whether AI mobile phones can become the new form of mobile phones in the third stage just like smart phones replacing functional phones.
Through terminal devices such as mobile phones, can we really have a “Jarvis” that can learn by itself and continuously adapt to user needs?
However, it is worth noting that for this route of advanced mobile phone intelligence, even with the rapid progress of large-scale model technology, the technical challenges still cannot be underestimated.
The key difficulty is that the agent route involves using AI to reconstruct the operating system.
Zhao Ming also further explained this. Compared with application, the construction of this core capability is more threshold:
Honor’s route is to first build platform-level AI and the core capabilities of AI phones. To a certain extent, we have slightly controlled and inhibited the development of mobile AI applications because we were worried that we would place too much emphasis on single-point applications and neglect the construction of core capabilities.
On this basis, at this node of 2024, we will work hard on AI end-side applications. There is no threshold in itself because it does not involve the reconstruction of the operating system.
This also explains why Honor took the lead in the competition of AI mobile phones in the first quarter.
First of all, it is deployed early, so it can take the lead when cutting-edge technologies bring innovation opportunities.
On the first generation of Magic in 2016, Honor was the first to bring the function of automatically turning on the screen when looking at it, automatically pushing boarding passes at the airport, etc., allowing the phone to actively identify the user's intention, and has since established the Honor IUI (Intention Identification Human-Computer Interaction). )’s direction of evolution.
In 2022, Honor began to establish platform-level AI capabilities on MagicOS 7.0, pushing a variety of card reminders in the three major scenarios of smart travel, smart life, and smart entertainment, including flight reminders, express delivery reminders, repayment reminders, etc., and gradually established The prototype of IUI.
With the migration of large model technology to the terminal side, this year, Honor not only released the 7 billion parameter “Magic Large Model”, but also comprehensively upgraded the platform-level AI capabilities based on the terminal side large model, and revolutionized it again on MagicOS 8.0 Human-computer interaction mode realizes intention recognition human-computer interaction, making it possible to personalize the operating system.
On the one hand, Honor platform-level AI allows the phone to pay more attention to the individual user, such as location, status, personal habits, knowledge base and other human elements.
On the other hand, the blessing of the device-side large model (Honor's self-developed 7 billion parameter magic large model) can further increase the platform-level AI's learning and understanding of human factors elements.
Secondly, it is a big investment.
Up to now, Honor’s AI R&D expenses have accumulated to 10 billion, and it has achieved more than 2,100 AI patents.
One More Thing
Coincidentally, Apple, which was revealed to be announcing a new AI layout at this year's WWDC, has recently revealed its “intelligent style” tendency to integrate large models with iOS in a deeper way.
In March, a new paper from Apple, “ReALM: Reference Resolution As Language Modeling,” also revealed Apple's technological progress in improving the voice assistant's ability to understand and respond to commands.
To put it simply, it is to let the large model analyze what operations the user has done on the mobile phone screen, so as to learn the user's usage habits and understand the user's intention.
Combining multiple sources of information, Apple now has great momentum to put theoretical research into practice.
It seems that whether it is Huawei's comprehensive return, Honor's strong rise to the top, or Apple's pursuit of new changes, the terminal situation in 2024 is destined to be more closely intertwined with AI and large models. AI has become the new long-term focus of competition in the terminal market.
Having Jarvis in everyone's hands is indeed more and more worth looking forward to.