On February 18thMeizu Technology announced that the company has decided to go all in AI and will stop the new traditional "smartphone" project。On the same day, Chen Mingyong, founder and CEO of OPPO, also issued an internal letter saying that 2024 is the first year of AI mobile phones, and the OPPO AI Center will be established, and resources will be tilted towards the center.
The reason why many mobile phone manufacturers have invariably deployed AI mobile phones is of course related to the huge business imagination of the rapid development of AI technology. Since smartphones belong to the mass computing platform, the introduction of AI technology may trigger a new wave of replacement.
However, it is worth noting that the portable characteristics of smartphones determine their limited computing power and battery life. In this context,How to give smartphones a revolutionary AI experience is also a challenge that manufacturers have to face
The attractiveness of traditional mobile phones is declining, and new technologies are urgently needed to break the situation
Once upon a time, smartphones were still a vibrant emerging industry, with shipments continuing to rise and new forces emerging one after another. YetIn the past two years, the market vitality of the smartphone industry has long been incomparable
Source: IDC
According to data disclosed by IDC, in 2023, China's smartphone market will ship 27.1 billion units, a year-on-year increase of **50%, the lowest shipment in 10 years. Since 2017, the year-on-year shipment volume has become the "main theme" of China's smartphone market.
The decline in the vitality of the smartphone market is certainly related to the completion of the sinking reach of related devices and the disappearance of incremental dividends, but on the other hand, it may also be due to the excess performance of mobile phone products and the lack of innovation, which is difficult to persuade consumers to change their phones.
According to the data disclosed by TechInsight,In 2023, the replacement rate of smartphones worldwide will drop to 23About 5%, the replacement cycle is as high as 51 months
It is undeniable that most smartphone manufacturers have not relaxed their pursuit of innovative technology. In the past few years, the smartphone industry has set off a lot of involution competition frenzy in imaging and charging.
Source: oppo
Taking charging as an example, as early as the 2022 MWC, OPPO demonstrated the 240W super flash charging technology, which can charge the equivalent 4500mAh battery to 100% in 9 minutes. Subsequently, many smartphones began mass production of 240W fast charging.
However, in 2023, the upper limit of charging power in the smartphone industry has not only not been broken, but most flagship mobile phones even generally use 100-watt fast charging.
The reason for the above-mentioned "technology ebb tide" phenomenon in the smartphone industry is largely due to the fact that the competition of related technologies is dominated by smartphone manufacturers and does not necessarily meet the needs of consumers. For example, due to the limited battery capacity of smartphones, 240W fast charging is not a few minutes faster than 100W fast charging, but it has to pay a higher cost and larger volume, which is more than worth the loss for consumers.
Therefore, the biggest challenge facing the smartphone industry at present is:The existing technology can already meet the needs of most consumers, and it is urgent for emerging technologies to give birth to new selling points to attract consumers to change their phones
ChatGPT detonated the technology industry, and mobile phone manufacturers saw hope
Just when the smartphone industry is feeling anxiousThe explosion of ChatGPT has made many smartphone manufacturers see the hope of breaking the game
It is understood that in November 2022, OpenAI launched ChatGPT, an AI-driven natural language processing tool, which can chat and communicate like a human, and can even complete tasks such as composing emails, scripts, and copywriting.
Although ChatGPT currently mainly exists on the web side, due to its need to reach users through different terminals, and smartphones are mass computing platforms, with the gradual maturity of technology, smartphones may become the ideal terminal for AI technology to land.
Source: humane
In fact, SAM Altman, CEO of OpenAI, has a deep understanding of the significance of portable terminals for AI technology. In March 2023, Humane, a former Apple executive startup, raised $100 million in funding from investors including Microsoft, Sam Altman, and others. In November, Humane released its first AI hardware, the AI PIN, equipped with a GPT large model. This product has been praised by many ** as the iPhone of the AI era.
From this point of viewDue to the completely different selling points and functions, mobile phones with the blessing of AI technology should be called AI agents rather than mobile phones。This also explains why Meizu will stop the new project of traditional "smartphones" while all in AI.
Because revolutionary AI technology and experience cannot be delegated to traditional mobile phones through simple system updates, the mobile phone industry may usher in a "third spring" as AI mobile phones gradually mature.
According to a research report disclosed by AVIC**, it is expected that AI mobile phone shipments will reach 100 million units in 2024 and 5 million units in 20272.2 billion units, with a CAGR of 83% from 2023 to 2027, and the penetration rate of AI mobile phones will reach 40% by then.
The computing power on the device side is limited, and the implementation of AI mobile phones faces many challenges
There's no denying thatPortable smartphones are ideal terminals for the implementation of AI technology, but it should be noted that AI technology requires abundant computing power and power, and it is difficult for small smartphones to bring a first-class experience in the industry at this stage
At present, there are two major technical schools of AI mobile phones, represented by AI PIN and Rabbit R1, hand over the computing to the cloud; One faction is represented by Honor, OPPO, and Xiaomi, which pursue local computing.
Source: Sina Technology.
AI mobile phones that focus on cloud computing power, such as AI PIN and Rabbit R1, have solved the problem of insufficient local computing power, but the response speed is not ideal due to the need to connect to the Internet at all times. There is a ** report, ask rabbit r1 a question, it takes about 20 seconds to give feedback. And in a noisy environment, the recognition accuracy of the Rabbit R1 is also very low.
Due to the instability of cloud computing and considering that mobile phone users have a large amount of private data, smartphone manufacturers such as Honor, OPPO, and Xiaomi focus on device-side AI.
For example, in January 2024, the Honor Magic6 series of mobile phones released by Honor will support a large device AI model with 7 billion parameters.
In an interview, Zhao Ming, CEO of Honor, said that the end-side large model "has more and more processing power, and 1 billion parameters can't be used, it's a toy, 5 billion, 6 billion, 7 billion, it's quite good to use, and it may be 10 billion or 20 billion in the future."
Although smartphones can achieve functions such as natural dialogue, AIGC elimination, and AI article summarization based on the 7 billion parameter AI device-side large model, the device-side computing power and battery life of smartphones are still difficult to support cutting-edge AI technologies.
Source: MediaTek.
Taking the flagship phone in 2023 as an example, the memory capacity is about 16GB, the bandwidth of the memory is about 50GBS, and the inference computing power is only 50TOPS. On the other hand, a large model with 13 billion parameters requires at least 13GB of memory and 130GB of bandwidth, and it is estimated that the computing power needs to reach 70 tops.
Dr. Zhongli Lu, Vice President of MediaTek's Computing and Artificial Intelligence Business Group, said, "If you run a 13 billion model on the device side, it will require more computing power, memory capacity, and memory bandwidth than the current flagship can provide. ”
In view of the application prospect of AI technology in the smartphone industry at this stage, Li Nan, former president of Meizu Blue and founder of Angry MIAO, posted on social **, "When there is no popularization of end-side LLM on **insensitive and power-insensitive trams, end-side LLMs on mobile phones are all gimmicks." ”
All in allFrom the perspective of ease of use, smartphones are undoubtedly the best terminal for AI technology to land, which is the primary reason why smartphone manufacturers are all in AI
However, it should be noted that smartphones also need to be portable, with limited performance and battery capacity, making it difficult to provide computing power and power resources for AI technology without control.
From this point of view, AI phones are still a long way from the real "AIlass" moment.