On a certain e-commerce platform, the second-generation Doubao phone equipped with the Doubao mobile assistant—the Nubia NaviX Ultra—has currently reached 387,000 pre-orders. Considering that Nubia itself is not among the top 5 brands in smartphone industry shipments, this pre-order figure for the Nubia NaviX Ultra is indeed astonishing. With the first-generation Doubao phone paving the way, public attention to the second-generation Doubao phone is even higher.
As early as the second half of 2023, manufacturers such as Huawei, Xiaomi, vivo, and Honor announced the launch of on-device large models. In 2024, the concept of the "AI phone" began to gain traction. OPPO even stated in an internal letter, "2024 is the first year of the AI phone." Smartphone manufacturers have successively started to promote the "AI phone" banner. However, at the consumer level, "AI phones" have not sparked a frenzy, or rather, consumers are not particularly sensitive to the so-called "AI phones." At the smartphone level, Huawei, Honor, Xiaomi, vivo, OPPO, and Apple all have their own AI assistants, such as Apple's Siri. Yet, these AI assistants have never really caught on. The main reason is that their functions are relatively marginal and limited in what they can do. On the other hand, AI assistants like Doubao, Wenxin, Qianwen, DeepSeek, Kimi, and Yuanbao are much more user-friendly. They not only offer diversified functions but also continuously undergo technical upgrades. For users, third-party AI assistants are significantly more practical than the built-in AI assistants on smartphones.
The boundary between smartphone manufacturers and third-party AI assistants is very distinct. However, this strict segregation has also led to a more realistic problem: the so-called "AI phones" have failed to gain widespread popularity. The best solution might be for products like Doubao, Wenxin, Qianwen, DeepSeek, Kimi, and Yuanbao to engage in deep cooperation directly with smartphone manufacturers, similar to the customized phone collaboration between Doubao and Nubia. Yet, shortly after the first-generation Doubao phone was launched, it was dealt a blow regarding authorization issues with other apps. Current apps are "information islands," and each app is unwilling to grant permissions to other internet giants. After all, they learned their lessons during the PC era. Therefore, in the mobile internet era, every internet giant is building its own content library and operating in isolated silos.
For "AI phones" to become truly useful, two issues need to be resolved: first, a more convenient form of user authorization; second, integration among apps.
In their current form, authorization methods like fingerprint unlocking and facial recognition are not particularly fast, as they still require the user to tap the screen. Since the phone is already unlocked, manually opening AI assistants like Doubao, Wenxin, or Qianwen yields the exact same result. Unless there is a more convenient and faster authorization method, "AI phones" will remain in a lukewarm state.
As for app integration, it is more like a series of invisible door locks. Theoretically, through customization with Nubia, the Doubao mobile assistant can enable ByteDance's apps to function more effectively. ByteDance is jokingly referred to by the industry as an "app factory," and its apps have long occupied the top ten spots on the overall free chart of Apple's App Store, including Douyin Mall, Hongguo Short Drama, Dou Shengsheng, Hongguo Comic Drama, Doubao, Qishui Music, along with Toutiao, Douyin, and Jianying. Simply by integrating its own apps, the Doubao mobile assistant can help users accomplish a great deal, which is also one of the highlights of the Doubao AI phone.
The problem, however, is that ByteDance is not the only company in the entire mobile internet industry. Chatting involves WeChat, payments involve Alipay, and there are also shopping, browsers, banking, news, and more. Faced with so many apps, if AI can integrate them all and help users solve practical problems in an AI-driven manner, then the "AI phone" would truly live up to its name.
However, as of now, smartphone manufacturers, third-party AI assistants, and app developers are not yet prepared for mutual openness and integration. I think we might start from specific usage scenarios. When using smartphones, users mainly engage in the following scenarios: office scenarios, handling work matters; consumption scenarios, such as online shopping and ordering food delivery; anonymous lifestyle needs, such as setting alarms, memos, and voice recordings; entertainment needs, such as watching videos, watching live streams, and playing games; and communication, such as making phone calls and sending WeChat messages.
Among these main usage scenarios, "AI phones" can gradually solve problems starting from some specific ones. Currently, the most promising is the third scenario: utilizing AI to solve problems in relatively unimportant situations that can be handled completely anonymously without involving user privacy.
In terms of entertainment needs, if user authorization is obtained, AI can indeed help users take over, for example, "I'm busy right now, AI, help me play the game for a while," or "Help me switch to another live stream/short video." In recent years, AI assistants like Doubao and Qianwen have successively launched their own input method products, with voice input being the main focus. The purpose of launching voice input methods is obviously not to make money, but to better understand users through the voice input method. For the exact same pronunciation, each user might want a highly personalized result. The voice input method serves as the entry point for user interaction with the AI assistant. The longer users use the voice input method, the deeper the input method's vocabulary and its personal understanding of the user will become. Consequently, the corresponding output results will be more accurate, such as calling XXX or sending a message to XXX.
However, in consumption scenarios, current AI still has flaws. For instance, if I want to buy a piece of clothing and ask the AI to compare prices simultaneously across platforms like Taobao, Tmall, JD.com, and Pinduoduo, while ensuring the corresponding merchants are reliable, the AI would be overwhelmed by just this single requirement. Users want not only basic information but also price comparisons. Yet, product prices on various platforms are floating because the coupons received by each person are completely different, and there are also membership tiers. Consumption scenarios involve specific individuals and merchants, and not a single mistake can be made. Current AI is clearly incapable of achieving this. The concept of the "AI phone" has been touted for several years; I wonder when the true "AI phone" will actually arrive?