Song Qinghui, a well-known economist, said that many large model companies have entered the processing capacity of ultra-long texts in the hope of incubating some super applications based on the above functions on the basis of their basic general models in the ToC camp.
Network illustration
In the second half of the "100 model war"! Commercialization is accelerating, who can be the first to break through?
With the popularity of ChatGPT all over the world, China has opened the "Hurricane Mode" of AI large models, and related projects have begun to "blow out", and the industry has presented a competitive pattern of "100 model wars".
Since last year, traditional Internet giants such as Baidu, Alibaba, and Tencent have set foot in the field of large models, and launched advanced large model products such as "Wenxin Yiyan", "Tongyi Qianwen", and "Mixed Yuan". In addition, industry leaders such as Huawei, JD.com, and ByteDance have also followed suit and actively laid out the large-scale model market. It is worth mentioning that the three major operators of China Mobile, China Unicom and China Telecom, as well as scientific research institutes and universities such as Tsinghua University and Fudan University, have also released their own large-scale model products, injecting new vitality into the development of the industry. At the same time, "dark horse" companies continue to emerge in the general model market, such as Kimi, which brings new challenges and opportunities to market competition.
According to data from the Cyberspace Administration of China, as of March 2024, the number of large models that have completed the filing in mainland China has reached 117.
According to a report by the China Business Industry Research Institute, the market size of China's large model industry has increased from 1.5 billion yuan in 2020 to 7 billion yuan in 2022, with an average annual compound growth rate of 116.02%, and about 14.7 billion yuan in 2023. The agency predicts that the scale of China's large-scale model industry will reach 21.6 billion yuan in 2024.
However, behind the prosperity, the commercialization of domestic large-scale models is still in its infancy. At present, large-scale model training requires strong computing power support, and the cost is very high. Under the huge financial pressure, exploring the path of commercialization has become an urgent problem to be solved.
Referring to the layout of foreign top players, the C-end AI products ChatGPT, Gemini, Character AI, etc., which have a large overseas user base and a relatively mature business model, are all proving a point. Preempting the application entrance and cultivating the user's usage habits is the key to commercialization.
At present, many domestic companies are actively exploring the commercialization path of C-end large models. For example, Baidu launched the "Wenxin Yiyan" subscription model on the C-side. In April, "Wenxin Yiyan" will be upgraded to open up long text processing capabilities, with a text range of 2 million to 5 million words.
Alibaba's "Tongyi Qianwen" has also ushered in a major upgrade, opening up the long document processing function of 10 million words to everyone for free, becoming the world's largest AI application in document processing capacity. In addition, Kimi, the dark side of the recent Exploding Moon, has seen a surge in monthly visits after the launch of the 2 million word long text feature.
Famous economist Song Qinghui
Bedside economics ¥59.8 to buy
Song Qinghui, a well-known economist, said that many large model companies have entered the processing capacity of ultra-long texts in the hope of incubating some super applications based on the above functions on the basis of their basic general models in the ToC camp.
In addition, on the B side, with the evolution of large model technology, more and more industries have realized the value of large model products and have begun to actively advocate the use of large model products. Among them, finance, education, medical care, automobiles, and scientific research have become key areas of concern for all parties.
According to industry insiders, according to overseas successful cases, combining AI technology with traditional products such as software services and cloud services, and providing vertical customization capabilities based on enterprise data, is a major path for B-end AI commercialization.
In China, more and more large-scale model players have moved from ToC to ToB tracks. For example, 360 focuses on scenarios such as AI security and knowledge management, and has formed a profound demand analysis of the ToB market after serving more than 100 customers. For B-end customers, Baidu has also built a set of commercialization matrix, mainly for the customer's product grafting the ability of Wenxin large model.
In addition, a few days ago, Kingsoft Office officially released the WPS365 office new quality productivity platform, and carried out strategic interpretation and customer sharing around the one-stop AI office for B-end organizations and enterprises.
However, compared with the C-side, the large model not only puts forward more requirements in terms of cost control and data security in the commercialization process of the B-side, but also has its own specific needs and problems, which requires the large model not only to have strong general capabilities, but also to be able to customize and optimize for specific industries.
It can be said that among the large model players, whoever can meet the needs of intelligent upgrading in various industries will be able to take the first step towards commercialization, and may be able to occupy a place in the "100 model war" and have more right to speak.
In summary, the real value of large models lies in their ability to solve practical problems and create business value. At present, players in the large-scale model track mainly start from both ends of ToC and ToB to explore the commercialization of large-scale models. From general models to vertical models, it is an inevitable trend to accelerate application implementation and business model innovation. Original title: The second half of the "100-model war"! Commercialization is accelerating, who can be the first to break through?
This article is from: Finet