APUS Li Tao: Customize large models for China to let AI create more value

On October 24, the 3rd China (Ningbo) Software Summit and Programmers Festival was held. Ni Guangnan, academician of the Chinese Academy of Engineering, Shen Changxiang, academician of the Chinese Academy of Engineering, and many corporate celebrities gathered on site to discuss the development of the software industry. . Li Tao, Chairman and CEO of APUS, a domestic AI large model company, focused on the theme of "AI leads the future with new intelligence" at this summit, analyzing how large models can lead industrial change, create value and promote social progress, and creatively proposed the AI ​​Big Model The “six elements” view of model value creation.

(APUS Li Tao speaks on the spot)

AI large model brings comprehensive changes

Artificial intelligence large model technology has developed rapidly and is gradually being implemented in various scenarios and creating rich application value. It has become an important driving force in promoting industrial upgrading and an important force in changing society.

Li Tao proposed that under the trend of AI reconstructing the industry, large models play the role of the underlying "operating system" to integrate, schedule, and allocate resources such as computing power, data, and applications, and comprehensively promote the construction of industrial networks. For example:

In the field of industrial Internet, multi-modal large model capabilities can monitor and analyze production line data in real time, provide support such as predictive maintenance, quality control and production optimization, and improve the efficiency and quality of industrial manufacturing;

In the field of agricultural Internet, the multi-modal large model restarts the construction of agricultural digital networks, uses AI to enhance the intelligence level of the entire chain of "production, supply, marketing, and service management", and promotes innovative applications such as digital machinery and digital live broadcast to improve production and sales efficiency;

In the field of the Internet of Things, the multi-modal capabilities of intelligent hardware and large models are connected, enabling them to independently possess perception and cognition, providing comprehensive, efficient and reliable guidance for innovative research and development;

On the journey of exploring the metaverse, the powerful reasoning capabilities of large models help the metaverse build cognition, supporting the rapid generation of exquisite design, complex modeling, human-computer interaction, mechanism establishment, and space operation and maintenance in space, allowing the metaverse to be quickly generated. The concept of the universe becomes possible.

(AI changes industrial factors)

Using large models as the driving force not only brings about changes in the Internet industry, but also brings about iterative upgrades of IT elements. Li Tao believes that with the popularization of AI, natural language will be widely used in programming in the future, and people can talk directly to machines; at the same time, functional positions such as programmers, product managers, and designers will also transform into "designers" of needs. By handing over tedious, mechanical, and repetitive work to large models and handing over value-creating work to people, social resource allocation will be more reasonable.

Building China’s large model competitiveness

Whether it is oriented to industry or IT elements, AI large models demonstrate powerful reconstruction capabilities. "Although large models seem to be 'omnipotent', not all models can be verified in practice, and not all models can create real value." Li Tao said. He believes that the goal of the development of large AI models is value creation, and to achieve value creation, "six elements" must be present: strong computing power, global knowledge base, high-quality data, continuously evolving algorithms, value alignment, and scenario integration value creation .

(Six elements of AI large model value creation)

Compared with leading foreign large models, China's large model industry started late, and technical bottlenecks such as algorithm computing power are difficult to break through in a short time; and due to limitations of Chinese corpus data, there is little high-quality data for training large models, and China's Big models aren’t “smart enough” yet. "Chinese corpus accounts for only 3% of the global knowledge base. Breaking through the data bottleneck is also a difficult problem that China's large model manufacturers must unite to tackle." Li Tao said.

Although some models seem to be "smart enough", there are problems such as misaligned values. In this regard, APUS believes that China's AI application research and development must still be based on China's large models. Therefore, China's large model manufacturers must have a sense of independence and shoulder the responsibility of independent innovation; build differentiated barriers based on the needs of the Chinese market and create accurate products that meet China's needs and values. AI large model to create new AI productivity.

Connect applications with value creation

The current large model market is still in the exploratory period of commercial implementation. It has long been an industry consensus that large models can create good AI applications and serve market needs well. Li Tao pointed out: "China's digital construction is a continuous iterative process, which requires a large model 'base' to support the rapid transformation of intelligent scenarios. Facing complex needs, APUS will continue to customize large AI models for the Chinese market to integrate AI applications with value creation. ”

(APUS’s AI strategy)

In terms of data, APUS has formed unique global user data and high-quality corpus based on the past 9 years of global business and 2.4 billion global users; at the same time, APUS cooperates with domestic corpus manufacturers to jointly build a database for the Chinese market. The “red corpus” of positive values. In terms of computing power, APUS has invested in the construction of two major intelligent computing centers around the world, and has linked up with cloud computing manufacturers such as Alibaba, Baidu, and Tencent to provide demanders with a variety of computing power combinations and elastic expansion solutions, helping enterprises and developers to achieve rapid implementation of research and development. , Costs are significantly reduced.

(AI large model built for China)

At present, APUS has cooperated with multiple industry partners such as medical, network information, manufacturing, and e-commerce to jointly create large-scale industry solutions, using AI to help partners upgrade their hardware infrastructure, build basic model platforms, and implement specific Business application innovation.

At the same time, APUS also continues to open up its ecosystem and cooperates deeply with manufacturers such as Baidu, Tencent Cloud, and Alibaba Cloud to build flexible industrial collaboration. For example, the APUS large model has opened model plug-ins, and developers can independently access external models and solve complex scene requirements through the capabilities of "model groups".

(APUS builds flexible industrial collaboration)

"Building China's big model" is a long-term correct thing. APUS is also adhering to this original intention to continue to consolidate platform capabilities and train large models to reason and evolve towards higher levels. Li Tao said: "We expect that APUS large models can play a fundamental role in thousands of industries; we also expect more developers to join the APUS ecosystem and use large models to create more viable AI applications."

With the theme of "Digital Empowerment, Intelligent Future Creation", this conference focuses on hot topics such as domestic open source ecology, AI large models, and industrial software empowerment. The activities include academician reports, release ceremonies, special reports, signing ceremonies, celebrity reports, press conferences, final reports, etc. In addition to Ni Guangnan and Shen Changxiang, representatives from companies such as China Unicom Cloud, Kingdee Cloud, and Tiger Industrial Cloud Group were invited to deliver the "voice of technology."

The above is the detailed content of APUS Li Tao: Customize large models for China to let AI create more value. For more information, please follow other related articles on the PHP Chinese website!