Cloud vendors explore new ways to play in the industry

After the large model began to enter the industry, the industry observed that the landing path and method of Baidu intelligent cloud in the industrial field are also evolving.

The first is a greater focus on AI-native thinking. Hu Wei told Digital Intelligence Front, “This round, customers have high expectations for large models.” They not only hope that the large model can realize some point-like applications, but also expect to build the central capability of enterprise intelligence on the basis of the large model, and thus evolve the intelligent application of all aspects of the enterprise.

Taking Baidu intelligent customer service as an example, after the launch of the large model, when communicating with customers, we found that customers do not want to disperse investment, and hope that through the large model, it can not only solve the problem of customer service, but also solve the problem of employee office assistant, automatic induction of corporate rules and regulations, and enterprise search. “They want the big model to cover all areas of enterprise management, business applications and customer service.”

Large models will bring about the refactoring of many industry applications, although at present, many applications still invoke the power of large models in a superimposed way of thinking. But in the long run, the familiar enterprise applications of the industry will gradually evolve into a new model in the coming period of time. For example, whether writing documents or making PPT, in the near future, users only need to put forward ideas and frameworks, and office software will be intelligently produced. The user then suggests changes, and it further completes the changes. “That’s how AI is native to thinking, and the whole way of thinking has been reconfigured.” Mr. Hu believes that the big model will do everything before.

The second is to help customers build their own intelligence capabilities. After the launch of the big model, many customers’ goal is not only to buy a big model, but also to gain the ability to eventually have a “big model only for themselves.” This big model knows its past and present life best, has all the data of the customer, understands its operation mode, and can incubate various enterprise applications.

Hu Wei introduced that the demands of some large central state-owned enterprise customers are very clear, they put forward to Baidu, they do not need Baidu to help write code, “Baidu is more of a coach role, give me good tools, teach me how to use, in the long run my ability can continue to improve.”

To this end, Baidu launched the Wenxin Qianfan large model platform, which is a tool chain for the whole process of large models, including data management, model training, evaluation and optimization, predictive service deployment, and plug-in services. Previously, Baidu Wenxin word was a To C product to provide consumers with large model capabilities. For B-end enterprises, the core service is the Wenxin Qianfan large model platform.

It is reported that Wenxin Qianfan is now exploring with some central state-owned enterprises and has begun to land. In this process, Baidu must not only find a balance between technology and customer business, but also find a balance in the quality of delivery and the cultivation of AI capabilities for customers.

The third is to take root in the industry. Because the research and development of large models requires larger data and deeper industry knowledge, and the current needs of industrial customers are also focused on the core business level, such as technology, research and development. The universal model integrates business scenarios, Wu Xueyi told the digital intelligence front, which requires them to go deeper into industry. The architects in the team have worked on the front line of the industry for many years, and the product managers have also gone to the factory line and the mine, “to dig down deep enough, there is mud at the foot.”

Ecology also plays a key role. Especially in the early stage of industry development, the first batch of partners are often enterprises with large amounts of data, industry knowledge and resources, and fully cooperating with them can also achieve twice the effect with half the effort.

Now, Baidu Intelligent Cloud is working with all parties to explore innovative models for serving small and medium-sized industrial enterprises. For example, the Wenxin Qianfan large model platform and the corresponding computing power platform will be sunk to the park. On the one hand, enterprises in the park can call the services of the platform for model training, which can realize innovation incubation at a low cost. On the other hand, the needs of a characteristic industrial park have commonality, and the large model may promote the common service platform. Together with partners, Baidu trains a set of large models for the common needs of enterprises in the park, such as equipment operation and maintenance diagnosis, safety production services, and joint operations with the park and local operating partners, so that small and medium-sized enterprises in the park can call services quickly and at low cost, and enjoy the technical dividends brought by the large model.

There are no small challenges in the large model landing industry, such as accuracy and cost performance should be improved. However, as a new thing, whether Baidu’s big model or the big model of other manufacturers in the industry, the development iteration is very fast. “If it was at a high school level when it came out in March, today I think it’s at least at a college graduate level.” There are Baidu people evaluation.