miniarcadepacman| Yuan Xiaohui, a senior expert at Tencent Research Institute: Various industries have three common needs for large models: "content generation and creative design"

Every reporter Chen Hao every editor Wei Guanhong

May 10th is the eighth "China Brand Day". With the arrival of "China Brand Day" in 2024, the "2024 eighth Brand value list of Chinese listed companies" jointly sponsored by Chengdu Media Group and Shenzhen newspaper Group, jointly sponsored by the Daily Economic News and Shenzhen Business Daily, and the academic support provided by the Chinese Enterprise Research Center of the School of Economics and Management of Tsinghua University will be held in Shenzhen on May 9th.

In the past year, the craze of artificial intelligence has swept the world, and how to enable large models to empower various industries has become a hot topic. What is the stage of domestic large models at the application level?Miniarcadepacman? At present, which industries have begun to use large models to improve productivity? With these questions, a reporter from the Daily Economic News interviewed Yuan Xiaohui, a senior expert in Tencent Research Institute.

Yuan Xiaohui said that from the perspective of the ability of large models, various industries have three common needs for large models, including content generation and creative design, information extraction and professional assistance, task scheduling and intelligent interaction. "under the guidance of important policies such as' artificial intelligence +', the large model is expected to accelerate the landing application in traditional industries and develop in the direction of multi-mode, artificial agent, end-to-side and miniaturization with the support of cloud and intelligence infrastructure. More deeply embedded in the workflow of various industries, so as to promote productivity."

On the ability of large models: there are three common needs in all industries

Yuan Xiaohui believes that from the perspective of the ability of large models, various industries have three common needs for large models, including content generation and creative design, information extraction and professional assistance, task scheduling and intelligent interaction. The promotion of brand value by the big model also includes the above three aspects:

One is to assist content generation and creative design to enhance brand awareness and influence. For example, the use of large models for content generation, such as social media posts, blog posts, press releases, etc., can help brands maintain high-frequency and high-quality content output and increase brand visibility in the public and industry. At the same time, high-quality AI technology display can also be used as a symbol of brand innovation, enhance brand image and attract more partners and investors.

Second, information extraction and professional assistance to help brands optimize marketing strategies. Staff mainly use the ability of large model summary, planning and data analysis to identify market trends and consumer behavior patterns to help brands develop more effective market positioning and marketing strategies.

Third, through task scheduling and intelligent interaction, the large model, as a new quality productivity, can enhance customer experience. Large models can optimize the user experience by providing fast and accurate customer service. For example, the online customer service chat robot can quickly respond to customer inquiries and solve customer problems, thus improving customer satisfaction and loyalty. Personalized recommendation system can recommend related products or services according to users' historical behavior and preferences, and enhance users' purchasing experience.

"the application of the big model in various industries in China is in the early stage of development." Yuan Xiaohui said that among all industries, digital native industries (such as software, games and other industries) are the forerunners, while producer services (advertising, finance, etc.) in traditional industries are making relatively rapid progress. progress in asset-heavy industries (construction, manufacturing, energy, etc.) is relatively slow. She said that there are two core influencing factors behind this, one is the adaptability of the large model to the industry demand, and the other is the availability of industry data. Often, the higher the model capability matches the industry demand, the higher the availability of industry data. The faster the progress of large model application.

Talking about the development trend of large models: it is expected to accelerate the landing of traditional industries.

Yuan Xiaohui stressed that at present, there are two issues that need to be paid close attention to in the application of large models in various industries:

First, there is a lack of criteria for judging the success of the application of large models, and enterprises are easy to pursue some technical indicators, such as paying attention to the technical performance of large models, including accuracy, recall rate and promotion rate, etc., but these indicators are far from enough. the level of technical indicators can not directly reflect the value of the large model. More importantly, pay attention to business indicators, such as the number of users, usage, revenue, etc., these indicators can directly reflect the value, as a more convincing criterion for judging success.

The second is to pay too much attention to the short-term output of the investment income while neglecting the long-term input. Traditional industries tend to treat large models equally with mature business, requiring clear input and output, and even hope that the input can achieve positive profits in the short term, which is not conducive to the development of large model applications. The large model is still iterating rapidly, and there are a lot of uncertainties, such as technological mutation, bubble, etc., it is difficult to have a good financial performance in the short term, and it is difficult to evaluate accurately in the long term. A more reasonable way is to take the large model as a research and development or incubation project, not to force the absolute value of short-term financial indicators, but to focus on the relative improvement of business, technology and other indicators.

According to Yuan Xiaohui, according to the "Industry Big Model Research report" to be released by Tencent Research Institute on May 10, various industries have begun to use large models to improve productivity, but to some extent. Specific to the scene, the speed of the application scene presents the characteristic of "smile curve", that is, the two ends of the industrial chain with high added value, and the application of the large model falls to the ground quickly.MiniarcadepacmanIn the middle of the low added value, the application process of the large model is slow.

"the reason is that the big model brings about a paradigm shift in intelligence as a service, which is particularly suited to the knowledge-intensive and service-intensive areas at both ends of the smile curve." Yuan Xiaohui said.

Yuan Xiaohui believes that under the guidance of important policies such as "artificial intelligence +", the large model is expected to accelerate the landing application in traditional industries, and with the support of the infrastructure of cloud intelligence, it will develop in the direction of multi-mode, artificial agent, end-to-side and miniaturization, and will be more deeply embedded in the workflow of various industries, so as to improve productivity.