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

On September 7, 2023 Tencent Global Digital Ecology Conference was held in Shenzhen. At the meeting, Tencent announced for the first time the progress of the Tencent mixed meta model and announced the upgrade of MaaS capabilities from the three dimensions of technology base, platform capability and intelligent application. Based on the self-developed Tencent mixed meta model, it also supports more than 20 mainstream open source models in the industry, helps customers build exclusive large models and intelligent applications, accelerates the landing of the model industry, and promotes the "performance revolution".

After the upgrade, the Tencent Cloud industry model selection store will be further expanded to provide customers with model services of different sizes; large model training and fine-tuning tool chains can help customers build "more practical" models faster and at a lower cost; at the same time, relying on Tencent's multiple "trial farms" of PaaS and SaaS product matrices, users can also quickly equip themselves with plug-and-play model capabilities.

At the meeting, Tencent Cloud and CICT jointly released the Research report on the Standard system and capability Architecture of Industry large models, referring to Tencent Cloud's practice and experience in the industrial application of industry large models. It provides customers with standardized processes and comprehensive evaluation methods for building industry models, so as to better help industry partners to build exclusive models.

Tang Daosheng, senior executive vice president of Tencent Group and CEO of Cloud and Smart Industry Group, said that Tencent's idea of building a large model of the industry based on industrial practice has been widely recognized by the industry. Through the "industry large model + enterprise data fine tuning" program, Tencent helps industrial partners to create large model products that suit them, and solves the challenges faced by enterprises in the process of applying large models, such as professionalism, accuracy, data security, cost, and so on. Technological changes and the rapid development of customer requirements have promoted the rapid iterative upgrade of Tencent Cloud MaaS service capabilities.

Tang Daosheng, senior executive vice president of Tencent Group and CEO of Cloud and Smart Industry Group

Open multi-base model, Tencent Cloud MaaS continues to promote "performance revolution"

At the meeting, Wu Yunsheng, vice president of Tencent Cloud, head of Tencent Cloud Intelligence, and head of YouTu Lab, believed that the current enterprise exploration in the field of large models has passed its "taste period", and industrial practice will become an important standard for testing the value of large models. The "depth" of the application of large models in the industrial field will also continue to expand. Based on this, the large model of Tencent Cloud industry is upgraded from three levels: technology base, platform capability and intelligent application to speed up the model implementation process.

In terms of the technical base, the mixed-element general model developed by Tencent was publicly unveiled for the first time. Wu Yunsheng said that Tencent mixed Yuan Model is a solid base for Tencent Cloud industry model. It not only has excellent Chinese creative ability, but also has logical reasoning ability and reliable task execution ability in complex contexts. It has been successfully accessed and achieved remarkable results in a number of products and businesses within Tencent, and will be opened to industry customers one after another through Tencent Cloud.

Wu Yunsheng, vice president of Tencent Cloud, head of Tencent Cloud Intelligence, and head of YouTu Laboratory.

In order to meet the diverse model needs of enterprises and developers, Tencent Cloud also took the lead in announcing the launch and support of open source models, including Llama 2, Falcon, Bloom and other mainstream open source models in more than 20 industries, covering rich application scenarios such as intelligent dialogue, text generation, writing, and so on. Customers can use Tencent Cloud TI platform to train the above open source model, fine-tune the quick start model based on Jupyter Notebook, and deploy the model with low code operation.

At the same time, Tencent Cloud has also greatly improved the platform capability. The upgraded TI platform tool chain connects to Tencent Cloud HCC high-performance computing nodes and Xingmai high-performance computing network, providing network bandwidth of 3.2Tbps. The newly upgraded Taiji Angel large model acceleration framework improves the performance by more than 30% compared with the commonly used solutions in the industry. The brand-new tool chain, with high-performance large model fine tuning and deployment ability, covers the whole life cycle of large model development and application, so that customers can fine-tune their own model solution in a short time by adding their own unique scenario data.

In addition, enterprises can also access Tencent's upgraded SaaS and PaaS product matrices to quickly equip model capabilities through rich, plug-and-play smart applications. The upgraded intelligent customer service robot expands the knowledge boundary, greatly improves the reception dialogue experience, and reduces the operating cost; the enterprise point analysis AI assistant transforms the traditional analysis process and improves the output efficiency of enterprise data analysis insight through the core capabilities such as "conversational data analysis, auxiliary analysis model configuration update, intelligent conclusion extraction, one-click generation of analysis report" and so on. New release of AI big model blessed "universal mouth shape" version of the small number of Homo sapiens, do not need training, upload materials within 1 hour to obtain exclusive digital intelligence.

Work with customer partners to improve the quality and efficiency of the 20 + industry

At the meeting, Wu Yunsheng said, "Tencent Cloud will continue to improve MaaS capabilities and join hands with customers and partners to meet the 'performance revolution' in the AI model era." As a major model of the industry, Tencent Cloud has landed in more than 20 industries, including Culture and Travel, Pan-Mutual, Retail, Energy and so on, covering production, sales, customer service and other links, and has created a number of best practices among its head customers.

For example, when a technology company builds a large-scale modeling platform, it faces the problem of lack of manpower, material resources, resources and so on. Tencent Cloud MaaS improves the training performance of 2D and 3D modeling tasks in the field of self-driving, and realizes lightweight operation and access. Provide high-performance computing clustering and training acceleration, improve training performance by 40%; open high-performance distributed file storage and high-performance computing nodes, and greatly improve the utilization of computing resources.

The field of literature and travel is the centralized scene of customer service application, and there is a strong demand to improve the efficiency of model application. Tencent Cloud teamed up with an online travel OTA company to explore the application of the literature and travel industry model to improve the pain points of large amount of knowledge maintenance, long time-consuming and high labor cost in the traditional intelligent customer service dialogue configuration. The fine-tuned customer exclusive model can solve business problems end-to-end without configuring the dialogue process. Improve the task completion rate and reduce the cost of dialogue construction.

Content creation requires high speed of generation and understanding of large models, and is one of the areas that can best verify the breadth of large models. Tencent Cloud MaaS helps to deploy and accelerate the reasoning of China Literature Group's Wen Sheng Chu and Picture Sheng Tu. C-end user applications need to provide a solution with high QPS and low model time-consuming. Based on the Angel reasoning acceleration engine installed on Tencent Cloud TI platform, China Literature's reasoning ability has been improved by an average of 30%, reducing the cost of Wenshengwen to the original 1x4. In addition, the TI platform provides dynamic batch and automatic service scheduling mechanism, which greatly improves the service throughput, model optimization and deployment module one-click acceleration and release model, which greatly reduces the maintenance cost of service deployment.

In the field of government affairs, large models are needed to go deep into the application scene, which is more professional and reliable. Tencent Cloud helps a tax department to build an exclusive government model to solve the problem of limited knowledge boundary of traditional intelligent customer service, low problem interception and coverage, and high follow-up operation and maintenance costs caused by the continuous iteration of policies, regulations and guidelines. The upgraded customer service system has been intelligently upgraded from knowledge collection to knowledge operation, and has become a "knowledge expert in the tax field" and a "tax assistant following regulations and guidelines".

Actively participate in standard-setting to promote the high-quality development of large models in the industry

The landing and expansion of the large model of the industry needs a standardized process and evaluation system. At the meeting, Tencent Cloud and CICT jointly released the Research report on the Standard system and capability Architecture of Industry large models (hereinafter referred to as "the report") to help industry partners better build their own large models.

"Research report on Standard system and capability Architecture of Industry large Model" released

Wu Yunsheng said that the report includes key links such as industry demand analysis and resource assessment, industry data and model co-construction, industry fine tuning and optimal deployment, etc. it can help industry partners gain insight into the dimensions of industry application scenarios, model development, model capability, model application, model operation, security and credibility, so as to comprehensively evaluate the large model.

Wei Kai, deputy director of the Cloud Computing and big data Research Institute of the Chinese Academy of Information and Communications, said that as an important research institution in the national information and communication field, the Institute has been concerned about the development and application of AI technology, especially the innovation and promotion of large model technology. CICT will cooperate deeply with Tencent Cloud to promote the construction and improvement of the standard system of industry models, so as to provide technical support and services for the intelligent development of the industry.

"We hope that the research on the standard system and capability architecture of industry large models can help industry partners solve the problems of how to" apply "large model capabilities" to their own industries and scenarios, and can more effectively solve the cost, data, security, compliance and other large model landing problems. through the establishment of a standard system to provide guidance and support for the construction and application of large models in the industry. " Jiang Bo, vice president of legal affairs of Tencent, mentioned.

Tencent has been actively participating in the construction of large model standards while providing a solution for the large model of the transportation industry. In 2020, Tencent was elected as a member and deputy secretary-general of the artificial Intelligence Sub-Committee of the National Beacon Commission. As a core member, Tencent assumed more standard-setting work and technical leadership. Not long ago, Tencent Cloud, together with China Institute of Information and Communication, launched an industry large model ecological plan and took the lead in the formulation of China's first large model standard for the financial industry, providing important support for the intelligent and high-quality standardized development of the financial industry.

With stronger capabilities, higher efficiency, wider scenarios and deeper applications, Tencent Cloud industry models continue to iterate to improve the quality and efficiency of the industry through MaaS capabilities. In the future, Tencent Cloud will continue to upgrade its products and technical capabilities, help Qianhang Baiye create and deploy more practical AI applications with higher efficiency, high quality and low cost, and help more entities achieve high-quality digital intelligence transformation and upgrading.

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.