In recent months, Baidu Intelligent Cloud has signed up several large central state-owned enterprises and industrial giants, from China Southern Power Grid to Sinopec, from BMW Brilliance to Kaiwu Tongxiang base. Baidu big model in the official announcement a few months later, into more and more industrial scenes.

In fact, the large model to enter the industry has a very high threshold, both to have the C-end of the common scene chat, writing, poetry and painting of a wide range of talents, but also need to B-end field know-how and industry know-how in-depth insight accumulation, only have one of them, are like a partial teenager, can not support its maximum potential.

Based on this, domestic technology companies are strengthening cooperation with the industry. After the head manufacturers such as Baidu public test Wenxin word, there is a new playing method and strategy: On the one hand, the ability to hone in C-end general scenarios such as office, writing, and marketing has been feeding back to the industry scene, on the other hand, product managers and solution engineers further go to the factory line, mine and field, and build industry capabilities with the assistance of the enterprise-class application platform Wenxin Qianfan large model platform.

These actions are yielding rapid results. Over the past few years, the AI capabilities of Baidu Intelligent Cloud have entered more and more core scenarios, and now with the support of the large model, the intelligent products and solutions precipitate on the Kaiwu industrial Internet platform have evolved from a single point of intelligence to the enterprise intelligent center. The generalization ability of the large model has further enhanced the scale replication ability of Baidu intelligent Cloud, and gathered a large number of manufacturing enterprises in Jiangsu, Chongqing, Guangzhou, etc., to obtain artificial intelligence technology dividends with a lower threshold, and achieve competitiveness reshaping.

01The big model knocks on the door of industry

Some trends show that large models are accelerating the exploration of applications in industrial manufacturing.

In the past few months, Baidu Intelligence Cloud has frequently made new moves in large industrial manufacturing sectors such as chemical, manufacturing, energy and automotive.

In early July, BMW Brilliance announced a strategic cooperation with Baidu, and the two sides said they would share superior resources to explore the integration and innovation of AI technology and the entire scene of the automobile manufacturing industry. In the field of petrochemical energy, large central state-owned enterprises have also expressed interest in large models. Last month, Sinopec and Baidu reached a strategic cooperation agreement to cooperate in six areas, including industrial artificial intelligence infrastructure construction, digital transformation and upgrading, and new technological innovation such as large models.

In the power industry, in May, Baidu and Southern Network signed a memorandum of understanding, the two sides to jointly promote AI and power dispatch integration innovation. Baidu and the State Grid jointly built, in the power industry in the application of large model of the case “equipment inspection knowledge assistant” at the end of June was also selected in Beijing’s first ten industry large model application cases.

In fact, compared with the Internet industry, the industrial manufacturing sector has been regarded as a more traditional field, and enterprises in the industry are generally more cautious about the application of new technologies. An industry source said that industrial customers really value is good to use, there is no real effect.

Hu Wei, general manager of Baidu intelligent cloud transportation logistics solutions, told Suzhi Front that after Wenxin’s public test, “a large number of central state-owned enterprises, including many industrial enterprises, took some scenes that were difficult to deal with themselves to discuss with us a lot, and wanted to use large models to solve problems in the industry”, Hu Wei said.

Industry observation, in the past few years, the domestic industrial manufacturing field has experienced the baptism of intelligent manufacturing and AI 1.0 stage, and many enterprises have a considerable degree of cognition of AI application. Especially in some large central state-owned enterprises, digital transformation has been promoted to a strategic height, and many conglomerates have set up specialized technology companies under the tier 1 company. This is also the reason why many head companies are moving quickly under the current boom of combining large models with the industry.

Since the large model is still in the early stage of application, how to land in the industry and what disruptive applications will be produced remains to be clarified. Manufacturers are exploring cooperation scenarios and entry points with industry customers. “This group of clients know that. We will put forward several intentionality directions to try and experiment, and finally find a meeting point. The application of different industries is also different.” Hu Wei introduced. At present, Baidu Intelligent Cloud in research and development, design, production, operation and marketing and other fields, with customers and partners to create landing solutions.

Mining is a relatively traditional industry in the popular impression. According to Wu Xueyi, general manager of Baidu Intelligent Cloud intelligent manufacturing solutions, at present, Baidu Intelligent Cloud has worked with enterprises in the industry to explore more intelligent upgrades through large models in mine safety production processes, safety and early warning scenarios. According to reports, in addition to the basic large model such as Wenxin Word, Baidu has 11 industry large models, covering transportation, energy and other key areas.

In this scenario, safety production knowledge and emergency response process are relatively standardized, and knowledge is solidified into the model through the learning ability of the large model. After adding multiple rounds of training, question-and-answer and precise guidance, the intelligent service supported by the large model reaches the production-grade standard with accuracy. In the mining or training links, intelligent service products can guide the operators to operate safely and properly.

“Partner companies have scenes and data, we docking training, multiple rounds of tests have had good results.” The partners now want to create a common platform to serve the entire mining industry.” Wu Xueyi said.

In the chemical industry, Wu Xueyi concluded that the current customer needs are divided into two categories. One is the industry research and consulting institutions, they hope to cooperate with Baidu, do large model research and development, the industry safety production, management and other standards, through the large model knowledge solidification, the formation of a common service platform.

Another type of demand is process upgrading. For example, a head chemical enterprise needs multiple rounds of quality inspection in the production process to ensure product quality, but because the stability of the product does not fluctuate much, some of the inspections are not necessary. They found Baidu, hoping to combine the previous data accumulation to create a large model of quality inspection and frequency reduction, reduce costs, and improve production efficiency.

In some areas where Baidu Intelligent Cloud has accumulated before, such as intelligent scheduling plate, previous AI solutions have been able to help enterprises coordinate transport capacity and freight demand, and greatly improve the efficiency of logistics scheduling in some scenarios. After the large model is supported, it has better human-computer interaction ability, based on weather and transportation link changes, more accurate sorting and scheduling information, and users can learn freight information by natural language interaction. In May, a Baidu partner released an intelligent logistics assistant based on the Wenxin model.

Landing in the industry, the technology maturity of applying large models in different scenarios is also different. The above-mentioned person revealed that in the safety production and other links, the application maturity is higher, and in the fields related to the process, the landing application of the large model still needs to be polished.

At present, Baidu Intelligent Cloud in the automotive manufacturing, construction, mining, logistics, textile and other industrial manufacturing fields to explore the use of large model scenarios and applications. These explorations have also achieved practical results. In the recent systematic evaluation of the knowledge question-and-answer ability of the large model of artificial intelligence in the Chinese industrial field conducted by the Institute of Trade Federations, Baidu Wenxin Word outperformed GPT3.5, ranking first in the comprehensive evaluation index in China.