Shulou(Shulou.com)11/24 Report--

Launch the test application of API invocation service for enterprise customers, serve more than 17000 customers, and explore the implementation of large model applications in nearly 500 scenarios in various industries. Since its launch in March this year, Baidu Intelligent Cloud Qianfan Model platform, as the world's first one-stop enterprise-level large model platform, has handed over the only "full score report card" for the industry.

On October 17, Baidu World 2023 was held in Beijing Shougang Park. Baidu Intelligent Cloud announced during the conference: a comprehensive upgrade of the "cloud and intelligence" strategy to provide full-stack service solutions for five types of customer landing model needs; for AI native application development, a major release of the "Qianfan AI native application development workbench" to accelerate the landing of enterprise AI native applications The release of China's first native AI application store, as well as China's first large model full-link ecological support system, enabling partners to grow, co-build and share a prosperous large model industrial ecology.

Shen Qiu, executive vice president of ▲ Baidu Group and president of Baidu Intelligent Cloud Business Group, meets the five major needs of enterprise large model landing: Baidu Intelligent Cloud upgrade "Cloud and Intelligence Integration" strategy

Since 2020, Baidu Intelligent Cloud has taken the lead in the industry to put forward the "integration of cloud and intelligence" strategy, and has carried out many iterative upgrades in technical architecture and services. At this Baidu World Congress, the strategic connotation of "integration of cloud and wisdom" has been comprehensively upgraded to "integration of cloud and wisdom, in-depth industry, ecological prosperity and AI benefits".

Shen Xi, executive vice president of Baidu Group and president of Baidu Intelligent Cloud Business Group, said that the deep combination of artificial intelligence and cloud computing is the key to the rapid landing of AI native applications, which is also the "cloud and intelligence as one" concept that Baidu Smart Cloud has always advocated and implemented. At present, all applications and services of Baidu Group run on Baidu Intelligent Cloud based on the "cloud and intelligence in one" technical architecture. In addition, for the five types of customer landing model needs, Baidu Intelligent Cloud based on the Qianfan large model platform to build the "large model super factory" respectively gives the best service solutions.

First of all, Qianfan platform can provide extremely efficient and cost-effective heterogeneous computing services for customers who only need computing power. In the part of large model training that customers are most concerned about, through distributed parallel training strategy and microsecond interconnection ability, Qianfan platform can achieve a speedup of 95% for cluster training on a million-card scale. Through prevention in advance, timely detection, location and solution, the ineffective operation of the cluster due to failures and other reasons can be avoided to the maximum extent, and the proportion of effective training time will be increased. The proportion of effective training time of million-card cluster will reach 96%. Fully release the effective computing power of the cluster, and greatly reduce the customer's computing power and time cost.

In addition, Qianfan platform is also compatible with domestic and foreign mainstream AI chips such as Kunlun Core, Teng Teng, Haiguang DCU, Nvidia, Intel and so on, supporting customers to complete computing power adaptation with minimum switching cost.

Face-to-face intelligence works with Zhihu. Based on the AI computing power cluster provided by Baidu Intelligent Cloud, we train the "knowledge chart AI" large model and multimodal large model Luca. The proportion of effective training time on the thousand card cluster is as high as 99%. While ensuring the continuity of the model training, it can achieve efficient convergence of the model training effect. In addition, Zhihu, good Future, Horizon and other enterprises are also training and managing large-scale clusters in a more stable, efficient and economical way based on the AI computing services provided by Baidu Smart Cloud.

Secondly, at the model level, for customers who want to directly call existing large models, Qianfan platform manages 42 mainstream large models at home and abroad, and enterprise customers can quickly call the API of various large models, including Wen Xin big model, to obtain large model capabilities. For the third-party large model, Qianfan platform also targeted Chinese enhancement, performance enhancement, context enhancement and other optimization. At present, Qianfan platform has served more than 17000 customers, and the usage of large model API continues to rise at a high speed.

Third, for customers who want to carry out secondary development based on existing large models, Qianfan platform provides a full lifecycle tool chain for the retraining, fine-tuning, evaluation and deployment of large models, with the largest number of 41 high-quality industry data sets in the industry. to help customers quickly optimize the effect of the model according to their own business scenarios. At present, many industry head customers, including China Postal savings Bank, du Lesser Fullness of Grain, Jinshan Office and Hebei Expressway Group, are developing exclusive models that meet business needs through the tool chain services provided by Qianfan platform.

Fourth, at the application level, some enterprises need to develop AI native applications based on large model services. A series of capability components and frameworks provided on Qianfan platform can help enterprises complete application development quickly and respond flexibly to the needs of users and the market.

Finally, another part of customers want to directly and easily purchase mature AI native application products to enable business development.

Empower enterprises to efficiently develop AI native applications: release "Qianfan AI native application development workbench"

In order to meet the requirements of agile and efficient AI native application development and reduce the threshold of AI native application development, Baidu Intelligent Cloud has released the "Qianfan AI Native Application Development Workbench", which precipitates the common patterns, tools and processes for developing large model applications into a workbench to help developers focus on their own business without having to involve superfluous energy in the development process. Specifically, Qianfan AI native application development workbench is mainly composed of application components and application framework services.

Application component services are composed of AI and basic cloud components, which encapsulate the underlying service capabilities, allowing each component to perform a specific function. "AI components" include not only large language model components such as question and answer, CoT,Chain of Thought, but also multi-modal components such as text diagrams and voice recognition, while "basic cloud components" include traditional cloud service capabilities such as vector database and object storage.

The application framework is to selectively connect and combine these components so that they can complete the task of a specific scene more completely. At present, retrieval enhanced generation (RAG) and agent (Agent) provided on Qianfan platform are commonly used AI native application frameworks. Under each framework, Baidu Smart Cloud also provides a rich template room to support developers to develop AI native applications quickly and efficiently.

▲ Baidu Intelligent Cloud has released the AI native application development workbench search enhanced generation (RAG,Retrieval Augmented Generation) framework, which can make more efficient use of enterprise proprietary knowledge and make accurate answers to relevant questions with the help of large models. It is a necessary core competence for professional knowledge question and answer AI native applications.

The conference also demonstrated how to quickly develop knowledge question and answer applications for Sany based on RAG framework: as long as you select the preset RAG framework in the Qianfan AI native application workbench and configure the corresponding parameters, you can quickly realize the development and launch of Sany's official website intelligent customer service application. Shen says that to build such a "little assistant", even if you need to process thousands of ten-thousand-word documents, it costs only a few hundred yuan, and then users only cost a few cents for each consultation.

▲ quickly completes Sany knowledge question answering Application Development Agent (Agent) based on RAG application framework, as a popular application framework in the industry, it can automatically disassemble the tasks given by human beings, automatically plan and call various components to complete the tasks together, and get feedback according to the task completion effect to improve their own ability. At present, Agent framework has been widely used in industry, transportation and other fields.

Zhongtian Iron and Steel has built an intelligent "enterprise scheduling center" based on the Agent framework provided by Qianfan AI native application development platform, which realizes the automatic perception, decomposition and execution of task instructions. For example, when the steel output is found to be substandard, the user only needs to ask the question once, and the big model can automatically call various resources and API of the platform to complete the BI fetch and third-party root cause analysis, find out the reasons for the substandard, adjust the production schedule in time and send e-mail to notify the dispatcher.

▲ Zhongtian Steel builds enterprise dispatching center based on Agent application framework to prosper industrial ecology: the first domestic AI native application store and full-link large model ecosystem come out collectively.

In order to meet the needs of enterprise customers to easily choose and purchase high-quality AI native applications and empower partners for commercial success, Baidu Intelligent Cloud launched a major launch: Baidu Intelligent Cloud Qianfan AI Native Application Store, the first domestic AI native application store for enterprise customers to conduct one-stop transactions, aims to establish a convenient and reliable connection for both supply and demand of AI native applications in the large model era. Greatly improve customer application selection and procurement efficiency, and create more business opportunities for partner merchants. At present, enterprise users can enter the Qianfan AI native application store through Baidu Intelligent Cloud official website and buy the AI native applications they like.

▲ Baidu Intelligent Cloud Qianfan AI Native App Store at this World Congress, Baidu launched China's first generative business intelligence product-Baidu GBI (Generative Business Intelligence). BI (Business Intelligence) is one of the typical application scenarios that show the "ceiling" of large model capabilities. Baidu GBI, as the native application engine of AI, has realized the comprehensive reconstruction of the ability of "intelligent question number" based on the Wen Xin Da model: facing any database, it can complete query and analysis through dialogue in one step. Compared with traditional BI, it greatly improves the efficiency of data analysis and reduces the professional ability requirements of data analysis.

Taking the analysis of business data as an example, Baidu GBI can shorten the working hours of data analysis and report writing that can be completed by business data analysts from days to minutes. As a kind of AI native application framework products, Baidu GBI can also achieve minimal cross-industry and cross-domain capacity transfer, and achieve efficient development and construction of government and enterprise BI-like AI native applications in scenarios such as urban governance analysis, energy consumption insight, industrial equipment operation management, financial investment decision-making, and traffic congestion management.

▲ Baidu GBI function display Kingdee, as one of the first partners to enter the Qianfan AI native application store, jointly launched Kingdee Cloud Sky GPT on August 8, 2023, and took the lead in launching the industry's first large financial model, helping customers to carry out financial management and analysis more efficiently, improve the accuracy and standardization of financial processing, and provide a strong guarantee for the long-term and stable development of the enterprise.

Large model technology is not only reshaping the service model and product form of cloud computing, but also reshaping the ecological pattern of the cloud computing industry, which puts forward new requirements for the capacity and system construction of cloud computing manufacturers and their partners covering thousands of industries.

In order to help partners achieve commercial success in the era of large models, Baidu Intelligent Cloud has built and launched China's first large model full-link ecological support system, and provides 10 integrated partners, 100 application partners, more than 10,000 start-ups and agent partners, including Qianfan Community, AI native application incubation, sales opportunities, marketing, enabling training. And Qianfan AI native app store and other full support.

Yuan Foyu, vice president of Baidu Group, said, "in this rapidly changing era, Baidu should not only enter no man's land courageously on the road of large model technology and commercial exploration, but also work with many partners to create a prosperous industrial ecology." jointly explore new business models and business opportunities to achieve real win-win cooperation. "

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.