"Science and Technology Innovation Board Daily" reported on September 5 (Reporter Zhang Yangyang) Today, in the "Industrial Innovation Roundtable" session of the main forum of the 2024 Inclusion·The Bund Conference, we focused on the development process of large models, breakthro

"Science and Technology Innovation Board Daily" September 5th (Reporter Zhang Yangyang) Today, at the "Industrial Innovation Roundtable" session of the main forum of the 2024 Inclusion·The Bund Conference, we focused on the development history of large models, breakthroughs in the next cycle, and industry realization paths On issues of concern to the industry, Han Xinyi, President and Chief Financial Officer of Ant Group, Wang Xiaochuan, founder and CEO of Baichuan Intelligence, Yan Junjie, founder and CEO of Minimax, etc. had an in-depth discussion.

As in-depth participants in the AI ​​industry, the leaders of these leading domestic large model companies believe that application breakthroughs are the key to the development of China's general artificial intelligence technology (AGI).

In their view, the development of large models requires a long-term perspective. Large model technology has made breakthrough progress in the past two years. We are currently at a point in the development of the industry. Large models are moving from "parameter building" to "application building". Next, the industry faces two problems. One is how to reduce its reasoning cost, thereby promoting large-scale implementation. The second is how to use applications to drive the development of technology.

Han Xinyi believes that the development and application of AI technology is a continuous iterative process. It is necessary to discover technical difficulties in application and make the direction of technological development more focused.

Wang Xiaochuan denied the "AGI bubble" theory. He believed that the development of large models was in line with industry expectations in terms of results. He pointed out that in order to promote large models to the next stage of development, the key is to achieve two major integrations: one is the combination of knowledge search and large models, and the other is the combination of large model enhancements. The

large model is considered the only way to achieve true general artificial intelligence. Wang Xiaochuan believes that when talking about AGI, we can refer to the "human" standard, that is, the ability to think, know, and provide services like a human. If it falls into the scenario, it will be a human-like assistant.

is far from achieving such a standard, and Yan Junjie feels that it still needs patience. At present, he is more concerned about expanding the boundaries of AI through technological optimization. He said that if the model error rate can be reduced to single digits, then the model will be in a trustworthy state, and AI may become a form of productivity from an assistant. This change will also bring about truly popular products and bring benefits to the industry. come essential changes.

At the round table, the guests also shared their companies’ large model application practices.

Han Xinyi said that the goals of large model products can be condensed into: easy to use, useful and affordable. For users, this means a new, more intuitive interactive experience; for merchants, it means being able to understand user needs more deeply and provide more structured and personalized services. This kind of service is a delivery result, and its transformation is not only reflected in user interaction, it is also a deep supply-side change. Alipay is constantly exploring how to integrate artificial intelligence technology into people's daily lives and business operations more naturally to promote innovation and transformation of business models.

Han Xinyi said that at the Bund Conference, Ant released three products: the AI ​​life manager "Zhi Xiaobao", the AI ​​financial manager "Ma Xiaocai" and the AI ​​health manager, among which Zhi Xiaobao is an independent app. He expects these three AI stewards to provide users with good services in three areas, and at the same time connect the open ecosystem in an intelligent way to find business models in the new business environment.

Baichuan Intelligence focuses on investing in the field of large-scale medical AI, starting from the two entry points of pediatricians and general practitioners, hoping to achieve low-cost and accessibility of medical care next year, relieve users' anxiety, and reduce the phenomenon of hospital runs. .