Zhongguancun Science and Technology's large model capabilities have passed the China Academy of Information and Communications Technology standard verification, reaching the highest level 4+!

From September 17th to 18th, the "2023 Trustworthy AI Conference and Nanjing Artificial Intelligence Industry Development Conference" hosted by the Nanjing Municipal People's Government and China Academy of Information and Communications Technology was successfully held. At the meeting, the Academy of Information and Communications Technology released a series of "Trusted AI" evaluation results. Zhongguancun Science and Technology's large model has successfully completed the standard compliance verification of China Academy of Information and Communications Technology's trustworthy AI model capabilities due to its richness of functions and superior performance, and received the current highest rating of level 4. It is among the first batch of dialogues to receive this level of rating. AI company.

In addition, Zhongguancun Science and Technology, as an important unit that promotes the development of large models in the field, was invited to participate in the "Industry-oriented Large-scale Pre-training Model Technology and Application Evaluation Methods" of the Academy of Information and Communications Technology. Part 1: The joint preparation of the "Financial Large Model" standard will empower the development of large models in the financial industry and help the digital transformation and upgrading of the industry to intelligence.

Obtained the current highest rating standard compliance verification, one of the first batch in the industry

Since China Academy of Information and Communications Technology began to build the "Trusted AI" evaluation system in 2018, the system has It has gradually become the most authoritative artificial intelligence evaluation system in China. The "Trusted AI" evaluation system starts from four aspects: AI basic software and hardware, typical technologies, products and services, and AI trusted governance, and forms multiple evaluation sub-systems to provide guidance for the research and development, product selection, capability improvement and development of artificial intelligence technology. Security guarantee provides a reference

Zhongguancun Science and Technology's large model has comprehensive basic capabilities such as semantics, speech, and vision. It successfully completed the model capability module evaluation of the trusted AI large model standard of China Academy of Information and Communications Technology. The compliance verification of 5 competency domains and 46 competency items finally received the highest rating of Level 4, and the functional richness and performance superiority were recognized.

Zhongguancun’s large-scale model in the field of science and technology and finance supports three task areas: intelligent semantics, intelligent speech and intelligent vision. It can perform semantic analysis, dialogue system, sentiment analysis, text classification, text generation, knowledge graph, text content recommendation, OCR, face recognition, image classification, semantic segmentation, action recognition, speech recognition, speech synthesis, picture generation, etc. tasks, and have high accuracy and acceptability in these tasks

Artificial intelligence technology is an inevitable choice to release the multiplication effect of digital superposition and build comprehensive competitive advantages. Based on years of accumulated enterprise service experience and industry knowledge, Zhongguancun Science and Technology has launched a large-scale model factory in the field, covering the entire process of data collection, annotation, training, pre-training basic models, and tuning, forming an original atomic task collaborative gain chart, which is highly efficient. Fine-tuning to avoid catastrophic forgetting

In order to ensure the application and actual effect of large-scale models and meet the customized needs of enterprises, Zhongguancun Science and Technology has created a prompt engineering platform. Through its highly encapsulated artificial intelligence capabilities and flexible configuration of orchestration tools, the platform can help enterprises quickly launch large-scale model applications and significantly reduce costs.

In general enterprise application scenarios, Zhongguancun Science and Technology has launched a series of "super employee" AIGC applications, including ChatPilot knowledge assistant, investment advisory assistant, service assistant, compliance assistant, etc., to open up the "last mile" of the digital transformation of enterprise dialogue scenarios; in segmented business scenarios, Zhongguancun Science and Technology provides enterprise knowledge intelligent training, enterprise knowledge Q&A training, knowledge summary and abstract extraction, investment research analysis, customer service assistant, quality inspection assistant and other large model applications to help enterprises reduce costs, increase efficiency and improve quality.

Currently, Zhongguancun’s large-scale model applications in the field of science and technology finance have been applied in multiple industries and scenarios. For example, the marketing assistant provided support to a financial planner of a wealth company, shortening the time for writing marketing copy from 10 minutes to 10 seconds; in the field of government affairs, the application of the medical insurance omniscient model helped the masses to answer medical insurance questions instantly. All information on medical insurance can be obtained in one stop; in the retail industry, the application of technician assistants allows only 2 people to complete the work that originally required 30 technicians, and the accuracy of semantic understanding has also increased from 85% to 85%. 94%

The first financial industry large model standard, jointly prepared

As it faces vertical fields and specific scenarios Large industry models are constantly emerging and are expected to become the technical base for front-end application scenarios in various industries.

In April 2023, China Academy of Information and Communications Technology began to formulate a series of standards for the "Industry-oriented Large-scale Pre-training Model Technology and Application Evaluation Methods", and took the lead in formulating large model standards in the financial industry. By building a trinity evaluation system of technical capabilities, application capabilities, and service capabilities, we help all parties comprehensively evaluate the technology and application capabilities of large industry models, further promote the industrialization process of large models, and accelerate the integration of large models and industry applications

In order to promote the intelligent upgrading of the financial industry, China Academy of Information and Communications Technology officially released the "Industry-oriented large-scale pre-training model technology and application evaluation method Part 1: Financial Large Model" standard at this conference. As a leading provider of conversational artificial intelligence technology solutions, Zhongguancun Science and Technology has become the first unit in China to compile large model standards for the financial industry with its independently developed domain large models and application innovations in new scenarios. Cooperate with Huawei, Baidu, Tencent Cloud, Ant Group, 360, Industrial and Commercial Bank of China, China Construction Bank, Agricultural Bank of China, Postal Savings Bank, Hang Seng Electronics and other enterprises to provide important support for the high-quality and standardized development of intelligence in the financial industry

According to reports, the industry standards released this time cover the key capability requirements of large-scale financial models, including three major aspects: scenario adaptability, capability support, and application maturity. In addition, the standard also starts from the characteristics of the financial industry, covering multiple financial application scenarios such as investment research, investment consulting, risk management, marketing, customer service, etc., and stipulates in detail the data compliance and traceability of large financial models. , privatization deployment, risk control and other aspects of requirements.

By training and tuning the basic large model with specific domain data, the domain large model has the professionalism and scene adaptability to meet the specific needs of the industry. This enables enterprises to more conveniently form a capability center that better matches their own industry, and has become an important measure to promote the penetration of large models into the industry. The three core technologies of domain large models, big data analysis, and multi-modal interaction provide enterprises with out-of-the-box, system-seamless, and affordable domain-specific large models, helping enterprises to lay out the entire chain of marketing, customer service, and operations. Business ecology realizes the value release of new human-machine collaboration.

On the other hand, we will speed up the application of large-scale industry models in various fields and create model demonstration cases in key fields. In the financial field, Zhongguancun Science and Technology has developed an enterprise intelligent knowledge base for Noah Wealth. It provides support for intelligent customer service products in the form of ChatPilot knowledge assistant, which significantly improves the customer service system's accuracy in identifying the intention of the question and provides more accurate reply. This technology can reduce the workload of text customer service by more than 70%, and improve the answer effect by more than 50%

In the future, Zhongguancun Science and Technology will continue to invest in the research and development of AI core technology, deepen the development of business scenarios, and combine its own expertise in technology and scenarios , services and other advantages to provide enterprises with richer, smarter, more convenient and lower-cost intelligent services, and accelerate the digital transformation and development of enterprises.

The above is the detailed content of Zhongguancun Science and Technology's large model capabilities have passed the China Academy of Information and Communications Technology standard verification, reaching the highest level 4+!. For more information, please follow other related articles on the PHP Chinese website!