From passive to active, every step is fraught with difficulties.

What are AIOS focusing on this year? The answer is smart assistants. With the support of large models, smarter smart assistants have stronger task execution capabilities, such as Honor's YOYO agent and vivo's Blue Heart Intelligence.

Overseas smartphone manufacturers have different ideas about smart assistants. Well-known blogger yawn revealed on social platform X that the latest beta version of Samsung OneUI 7 will include a feature called 'AI Brief', which can intelligently process some information required by users and report it at a certain time. For example, with this feature, Bixby may remind you of the weather, stock prices, and email summaries every morning.

(Image source: X)

In fact, Samsung is not the first smartphone manufacturer to have this idea. Before the exposure of AI Brief, similar features have been demonstrated and leaked by Apple's Apple Intelligence and Android 16, but none have been officially launched yet.

The AI priority notification feature, represented by AI Brief, may be a crucial step for system-level AI to evolve from passive assistance to active assistance.

What is the difference between AI priority notifications and traditional push messages?

Priority notifications are not exactly a new feature in mobile phone systems. Both the native versions of iOS and Android have similar capabilities: Apple added Siri 'Notifications Summary' in iOS 15, which organizes and displays important information by priority, allowing you to view notifications at specific times rather than being bombarded by fragmented information; Google introduced 'Priority Notifications' in Android 13, which allows users to separately extract messages from important contacts or applications to form their own notification content of interest.

Whether it's iOS or Android, without the support of large models, they can only classify notifications according to a fixed algorithm. Even if users' usage habits change over time, it is difficult for the system to automatically adapt to these changes.

(Image source: Apple)

The significance of AI priority notifications lies in adaptability, making users feel that the smart assistant is thinking like a human.

According to the leaked information, the AI Brief in the OneUI 7 beta actually uses device permissions to understand the user's location, app launch frequency, sensitivity to various notifications, and other information. It then combines large model algorithms to deeply learn your needs, ultimately providing the correct notification summary and reminders at the right time. For example, it knows what music you like to listen to in the morning and can remind you to bring an umbrella when it suddenly rains. Behind these functions, it has been learning your behavior and habits.

AI priority notifications require the use of the phone's local AI computing power to complete this series of understanding, learning, and analysis. Most notifications are time-sensitive, such as weather forecasts. Sometimes, sudden localized rainfall may arrive within half an hour after the notification is sent. Timely notifications can reduce the inconvenience caused by travel, and local analysis can improve processing efficiency and avoid delays in such emergency notifications. Similarly, since understanding and learning are private processes, not uploading data to the cloud becomes the best choice for most smartphone manufacturers to protect user privacy.

(Image source: Apple)

To reduce the pressure on AI computing power, Apple's AI priority notifications on iOS 18 actually identify each notification based on its text content. Compared to categorizing notifications based on user preferences, Apple prefers to simplify notification content so that users can immediately understand what a large number of notifications are about at a glance.

However, AI priority notifications do not provide much help to Samsung, Apple, and Google, as they are not as complete in AIOS as domestic manufacturers, especially customized systems like Honor, vivo, and OPPO that deeply integrate systems with large models. They inherently have strong task processing capabilities and can better help users handle various events when combined with priority notifications.

System-level AI becomes the new focus of the AI OS battle

Compared to the rawness when various AI systems were introduced last year, this year's AIOS have found the most suitable directions for themselves. However, regardless of the presentation method, the underlying logic remains the same: system-level AI.

System-level AI is actually global AI, using large model algorithms to deploy at all levels of the system, including the underlying system architecture, memory optimization, AI function invocation, and API interface openness. Currently, Honor's MagicOS 9.0 can be considered the most radical, not only obtaining the first L3 AI system certification but also cooperating with a large number of third-party AI service providers to create unique AIOS features.

(Image source: Samsung)

MagicOS has had functions similar to AI priority notifications for a long time. For example, during the MagicLive period, Honor could already provide users with useful help and suggestions by obtaining location information. On MagicOS 8.0, YOYO assistant can also actively suggest displaying the travel code when users take the subway and remind users to pick up packages when they arrive at the express station.

In addition to Honor, vivo's OriginOS, Xiaomi's MiUI OS 2, and OPPO's ColorOS 15 have made considerable efforts in system-level AI. Especially vivo, with its quick interaction combined with the Atom Island, has shortened users' operation processes and steps.

Although all manufacturers have implemented system-level AI this year, their directions are quite different. Xiaomi's MiUI OS 2 mainly uses large models for beautification and underlying system optimization. The most notable feature is the new AI dynamic wallpaper, which allows users to animate their chosen images. OPPO's ColorOS 15 upgrades Breeno Assistant to support continuous questioning and real-time image search, as well as features like AI reflection removal. These system-level AIs mainly allow users to autonomously select tasks to execute and have not yet derived scenarios where AI proactively provides services to users.

(Image source: Xiaomi)

Proactive intelligence is the ultimate form of AI OS evolution

In the future, the goal of smartphone manufacturers is to enable AI to provide proactive assistance at the system level, including allowing large models to proactively understand, learn, and assist users in executing tasks. As AIOS can reach more scenarios, users will naturally not only require the system to stay at the 'smart' stage but also hope it can increasingly 'understand me'.

This necessitates a transformation of system-level AI from a simple 'passive assistant' to an 'active partner', which can be simply understood as 'the phone being our life manager'. This transformation requires comprehensive upgrades, from anticipating needs to global collaboration, from immediate response to long-term learning, to better understand users' needs. The first step in this process is AI priority notifications, which Samsung, Apple, and Google are all working on.

However, the problem facing customized Android systems is also evident. Even though the current mobile-side + cloud computing power is sufficient for AI priority notifications to integrate different notifications for users based on different scenarios, third-party apps do not need to push notifications through the system, making it difficult for AIOS to integrate notifications from the bottom up and provide practical advice to users. Moreover, AI that is still in the learning stage may provide users with many wrong suggestions, a process that may be lengthy and may not retain users' patience to grow together with AI.

Currently, the best approach is to follow Apple Intelligence's 'notification simplification' solution, allowing the system to familiarize itself with the process of understanding notifications and then continuously improve hardware computing power so that the phone can purely rely on the mobile side to judge and integrate notification content. As large model algorithms gradually mature, AI can better evolve from passively executing tasks to proactively providing advice to users and even proactively executing tasks in specific scenarios, freeing up users' hands.

(Image source: PC-Tablet)

It is foreseeable that the evolution of system-level AI will be a protracted battle, evolving from current notification integration and semantic understanding to global task management, cross-device collaboration, and even emotional interaction. This evolution is not just a one-way upgrade of technology but is driven by user experience demands for the next interaction revolution. The manufacturer that achieves this progress first will ultimately win in this AIOS battle.

Source: Leitech