With all the big AI models competing, how can we better empower various industries?

Domestic large models, let a hundred flowers bloom. General large models, industry large models, vertical large models... major technology and Internet companies have chosen the model types they are best at to start, and competition has become extremely fierce.

In this context, large AI models are gradually moving towards the in-depth application stage. Among them, there are many giants such as Alibaba, Huawei, and Baidu "flexing their muscles", as well as players in vertical fields such as iFlytek, SenseTime, and NetEase. Industry insiders said that the application of large models in thousands of industries is accelerating, and the market is rapidly expanding.

As of the end of July this year, a total of 130 large artificial intelligence models have been launched in our country. The first batch of 11 large model products registered through the "Interim Measures for the Management of Generative Artificial Intelligence Services" have also been approved to open services to the whole society.

Some experts said that in the general field, large models will show a "winner takes all" pattern, dominated by a few companies with data, computing power and technical advantages. In the vertical field, leading companies will rely on their professional knowledge and business model advantages to make better use of large models to improve efficiency and make profits.

The wind has risen and the wave has arrived. In this industry change full of opportunities and challenges, how will AI large models empower various industries?

Application acceleration

The AI ​​large model itself is a tool to realize industrial value, and the real realization of its own value must be in industrial applications. According to iResearch's forecast, the scale of the industry driven by artificial intelligence is expected to reach 1.67 trillion yuan in 2025.

The just-concluded Hangzhou Asian Games is known as the first cloud-based Asian Games in history. Based on Alibaba Cloud, 100% of the core system is moved to the cloud. At the Hangzhou Asian Games Digital Technology Experience Center, visitors can use the art installation "Imagine the Future Asian Games" to have Alibaba Cloud's large painting creation model "Tongyi Wanxiang" paint on-site to generate Asian Games theme images.

Alibaba Cloud is one of the first technology companies to invest in large model research and development. It began to lay out large models in 2019. In April 2021, it released the large language model "PLUG"; in October 2021, it trained a multi-modal M6 large model. The parameter scale is 10 trillion; Alibaba Cloud released the "Tongyi" large model in September 2022; Tongyi Qianwen was officially released in April this year.

At that time, Alibaba Cloud Tongyi Qianwen opened the invitation test, and in just one month, more than 200,000 enterprise and institutional users applied to access Tongyi Qianwen test. Among them, enterprises such as OPPO, Fudan University, and China Southern Power Grid have used Alibaba Cloud's computing power and AI capabilities to conduct new AI explorations.

At the same time, Geely Automobile, Zhiji Automobile, Chery New Energy, Hao Mo Zhixing, Swire Coca-Cola, Bosideng, Zhangyue Technology and many other companies have stated that they will explore and co-create technology cooperation with Alibaba Cloud in large-scale model-related scenarios.

Picture source: Tu Chong Creative

Huawei has also proven through practice that large models are important, but what is more important is to use large models to solve the pain points of industries and products, to create products and services that can make enterprises and users pay for them, and to truly create value for thousands of industries. .

In July this year, Pangu Model 3.0 was officially unveiled at the Huawei Developer Conference 2023 (Cloud). At that time, Huawei Managing Director and Huawei Cloud CEO Zhang Pingan announced that Pangu Model was positioned as "born for the industry", saying that it "does not compose poetry, only does things."

Looking at the entire industry, compared with general large models, the research and development progress of large models for vertical industries such as biomedicine, finance, and industry is slightly slower because the latter are more difficult in terms of professional data collection, scene adaptation, and customization. high.

Different from large models in other industries, Pangu 3.0 is a series of large models that adopts a hierarchical decoupling design. Huawei Cloud has prepared five basic large-scale models at the bottom layer, including natural language, vision, multi-modality, prediction, and scientific computing, to meet a variety of skill needs in industry scenarios.

On top of the five basic large models, Huawei Cloud has also developed N large industry models, which can provide industry-wide large models trained using industry public data, including large models for government affairs, finance, manufacturing, mining, meteorology, etc. ; It can also be used to train customers’ own proprietary large models based on industry customers’ own data and on the basic layer of Pangu’s large model.

As a result, Huawei Cloud provides more detailed scenario models, focusing more on specific industry applications or specific business scenarios such as government hotlines, branch assistants, lead drug screening, conveyor belt foreign body detection, and typhoon path prediction.

AI Empowering Industry

What is the core competitiveness of domestic large models? As the development and application of domestic large-scale models enters the deep water area, the answer to this question is increasingly available: it starts with technology, is stronger than application, and benefits the ecology.

"The value of large models lies in their applications. Only by empowering the development of smart economy and smart society in diverse practical application scenarios can we find the industrial value and achieve the big models themselves." Tsinghua University Artificial Intelligence Research Institute Vision said Deng Zhidong, director and professor of the Intelligence Research Center.

From technology to application is not an easy task. The "National Artificial Intelligence Innovation and Application Pilot Zone "100 Scenes of Intelligence"" released by the Ministry of Industry and Information Technology in 2022 shows that among the nine major AI implementation fields such as public security and transportation, manufacturing, agriculture, energy, finance, and medical care are important economic pillars in our country. , is still an undeveloped area for AI implementation, and the scene penetration rate is low.

Since the second half of this year, Beijing, Shenzhen, and Shanghai, as the “first tier” cities in the development of the domestic artificial intelligence industry, have intensively issued policies to support the acceleration of the development and application of large models. On August 31, the first batch of large models registered under the "Interim Measures for Generative Artificial Intelligence Service Management", including Baidu Wenxinyiyan, Baichuan Intelligence, and SenseTime, were announced to be open to the public. Subsequently, large models of more enterprises were laid out and launched one after another.

Picture source: Tu Chong Creative

On August 31, the first day Wen Xin Yi Yan was opened, Baidu official platform data showed that in just 24 hours, Wen Xin Yi Yan responded to more than 33.42 million questions from netizens, with an average of more than 23,000 questions answered in one minute. . On that day, Wenxinyiyan APP appeared on the hot lists of many app stores.

On September 5, the iFlytek Spark cognitive large model was opened to the public, and the number of users exceeded 1 million within 14 hours of being online.

On September 7, Tencent’s Hunyuan model was officially unveiled. Jiang Jie, vice president of Tencent Group, said that Tencent will fully embrace large models. At the same time, it was also announced that it will open up the Hunyuan large model capabilities to the outside world through Tencent Cloud to serve various industries with self-developed technology.

The "China Artificial Intelligence Large Model Map Research Report" points out that domestic general-purpose large models are continuing to expand their application fields and create a cross-industry universal artificial intelligence capability platform. The applications are moving from office, life, and entertainment to medical, industrial, Education and other industries are accelerating penetration, and large-scale model applications in vertical fields such as biopharmaceuticals, remote sensing, and meteorology are also being implemented in depth.

With the increasing number of large model products, the competition in the large model industry has transitioned from the technology competition stage to the implementation application stage, and the era of AI empowering thousands of industries is accelerating.

The above is the detailed content of With all the big AI models competing, how can we better empower various industries?. For more information, please follow other related articles on the PHP Chinese website!