Recently, Xiaomi's series of heavyweight moves in the field of AI big models have stirred up a thousand waves in the technology circle.
It is reported that Xiaomi is building its own GPU 10,000-card cluster and will invest heavily in the field of AI big models. It is reported that Xiaomi established the AI laboratory big model team in April last year, and the team already had 6,500 GPU resources when it was established.
In addition, it is reported that Lei Jun recruited Luo Fuli, one of the key developers of DeepSeek-V2, with a high salary, and she will lead the Xiaomi AI big model team.
What kind of big game is Xiaomi playing with such a series of moves that it has invested so heavily in AI big models?
The arrow is on the string and must be shot
The hottest topic in the past two years must be AI large models.
As we all know, with the popularity of ChatGPT, a new round of artificial intelligence craze has set off around the world. All walks of life are actively embracing AI large models, and various AI large model products are constantly coming out, performing a " The grand occasion of "Battle of 100 Models".
According to data from the "2024 Artificial Intelligence Large Model Industry Development and Application Research White Paper", China's large model market size has reached 21.6 billion yuan in 2023, and is expected to exceed 117.9 billion yuan by 2028.
In this context, regardless of whether the above-mentioned rumors about Xiaomi are true or not, for Xiaomi, it is a must-choose question to rush into large-scale AI models. Not to mention other things, the development of Xiaomi's mobile phone business, automobile business, smart home business, etc. will inevitably require the strong support and in-depth empowerment of AI large models.
In the mobile phone business, the iterative innovation of AI large model technology has brought huge development space to the AI mobile phone market and significantly improved the intelligence level of mobile phones. Through technologies such as deep learning and natural language processing, AI large models can understand users' voice instructions, achieve more accurate speech recognition and semantic understanding, greatly improve the interactive experience of mobile phones, and provide users with more convenient, safe and personalized services. .
According to Counterpoint’s forecast, the penetration rate of AI mobile phones will increase from 11% in 2024 to 43% in 2027. At the same time, according to IDC estimates, global shipments of new-generation AI mobile phones will reach 170 million units in 2024, accounting for approximately 15% of overall smartphone shipments. Judging from this growth trend, AI mobile phones are gradually becoming the mainstream of the market and inspiring the recovery of the smartphone industry.
Based on this, major mobile phone manufacturers will naturally increase their research and development and investment in AI technology, including Samsung, OPPO, vivo, Honor, etc. As a leader in the domestic smartphone market, Xiaomi has to increase its investment in large AI models to maintain and enhance its competitiveness.
In the automotive business, autonomous driving technology is the future trend of the automotive industry, and its core lies in AI algorithms and big data processing. By training and optimizing large AI models, accurate perception and intelligent decision-making of the vehicle driving environment can be achieved to improve the safety and reliability of autonomous driving. At the same time, the AI large model can also provide rich functions in in-car entertainment, intelligent navigation, voice interaction, etc., to enhance the user's driving experience. At present, AI large models are increasingly used in the automotive industry, including intelligent customer service, in-vehicle voice systems, autonomous driving perception algorithms, etc.
Not long ago, Xiaomi SU7 was officially connected to the VLM visual language large model. Through this visual language large model, the system can identify complex road environments such as construction areas, dim roads, large areas of waterlogged roads, and special traffic rule areas such as bus lanes. And through text and voice prompts. It is foreseeable that the application of AI large models in smart car driving will continue to strengthen in the future, and Xiaomi Motors will also need to continuously iterate AI large model technology to improve users' driving experience.
In the smart home business, the emergence of large AI models has brought technological innovation and development opportunities to smart homes. Through the powerful learning and interaction capabilities of AI large models, home devices can be made intelligent and interconnected, and smart devices can understand and identify complex and ambiguous instructions issued by users, thereby improving user experience and achieving more intelligent interactions. experience. In recent years, as consumers' living standards have improved, their demand for these smart home products has also continued to increase. Xiaomi also urgently needs to improve its AI capabilities to meet the more diversified smart needs of users.
Xiaomi's AI history
In fact, as early as 2016, Xiaomi had already begun to lay out the AI field and established the first visual AI team, thus embarking on the journey of AI exploration.
In 2017, the Xiaomi AI Laboratory was established, further strengthening Xiaomi's R&D strength in the field of AI. The establishment of the laboratory provides an important platform for Xiaomi's exploration and innovation in AI technology.
In 2019, Xiaomi opened its artificial intelligence platform to attract developers to develop applications using its AI technology.
In 2020, Xiaomi upgraded its core strategy to "mobile phone × AIoT", launched new smart devices and applications such as smart speakers and smart cameras, and strengthened cooperation with other companies and research institutions.
In March 2021, Xiaomi launched its first self-developed AI chip Pengpai C1, established an autonomous driving team, and established the Xiaomi Robotics Laboratory in August to further deepen its layout in the field of AI.
In 2023, Xiaomi established a large model team and set "lightweight and local deployment" as the main breakthrough direction of Xiaomi's large model technology. At the same time, Xiaomi upgraded its technology strategy, put AI at the core, and put forward the principle of "deepening the underlying technology, long-term continuous investment, deep integration of software and hardware, and comprehensive empowerment of AI". It plans to invest 100 billion in technology research and development in 12 technical fields and 99 subdivided tracks in the next five years.
So far, Xiaomi's AI team has more than 3,000 people, covering multiple cutting-edge directions such as vision, acoustics, voice, NLP, knowledge graph, machine learning, large models, and multimodality, promoting AI technology to fully empower mobile phones, AIoT, automobiles, robots and other business sectors.
Indeed, Xiaomi has its own unique advantages in the field of AI. Its huge "people, cars, and homes" ecosystem, which has been built for many years, is like a tightly interwoven intelligent network, giving AI large models an extremely broad soil for taking root.
This time, whether it is the GPU Wanka cluster construction plan or Luo Fuli's heavy investment, it is one of Xiaomi's actions to fully develop AI large models, so as to further help Xiaomi gallop on the AI track. In the future, we are also looking forward to Xiaomi's in-depth cultivation in the field of AI large models.
This paper is from Ulink Media, Shenzhen, China, the organizer of IOTE EXPO (IoT Expo in China)
