Xinhua News Agency and Peking University released a large AI model evaluation: safety and reliability are the focus, and 360 Intelligent Brain performs well

Recently, the "Artificial Intelligence Large Model Experience Report 2.0" (hereinafter referred to as the "Report") released by the China Enterprise Development Research Center of the Xinhua News Agency Research Institute shows that China's current large model products have made significant progress, such as 360 Intelligent Brain, Zhipu ChatGLM Wait for the performance to be eye-catching. Among them, the 360 ​​Intelligent Brain large model ranks firmly in the first echelon of large models in terms of basic capabilities and other aspects. The report evaluates that the development of large AI models has brought positive impacts to the improvement of the quality and efficiency of human work and life.

It is reported that the Institute of Culture and Communication of Peking University and experts from other industries and academia were specially invited to participate in this assessment, which included 500 questions including basic abilities, IQ tests, emotional intelligence tests, and tool effectiveness abilities. The assessment was based on the number of question types. , question type quality, expert team and other aspects have been upgraded, making it more authoritative and objective.

The report cites 360 Intelligence’s answers to some questions as the best case studies, such as “Shanxi diners pour vinegar to wash dishes” and other time-sensitive issues. 360 Intelligence “demonstrates the powerful real-time search and response capabilities of large models.”

It is understood that large artificial intelligence models currently have two major shortcomings, namely real-time performance and accuracy, which are fatal flaws in enterprise-level scenarios. Search engines can help large models perform knowledge updates and corrections, and solve problems such as data lag and hallucinations. As the second largest search engine manufacturer in China, 360 has natural large model technology genes and advantages.

The report shows that “the large-scale model process is currently over halfway, and the introduction of AI into thousands of households has become the focus of attention.” Zhou Hongyi, founder of 360 Group, believes that the real development opportunities for large models in the future are in the enterprise-level market. Large models have the potential to lead a new round of industrial revolution, provided that they enter thousands of households and empower all industries. It is understood that 360 has launched an enterprise-level AI large model solution and will follow the four principles of "safety, kindness, credibility, and controllability" to create an enterprise-level vertical large model. It has currently provided solutions for 20 industries.

The report also mentioned that "the application of safe and reliable large-model products is becoming the direction of intensive development." It is understood that large models currently face three major security issues: network security, data security, and algorithm security. 360 has the world's largest security big data, the world's leading APT capture capabilities, and takes the global issue of AI large model security as a direction to overcome. 360 recently released China's first large-scale deliverable security industry model, which will serve as an important platform and tool for the "360 Security Cloud" to provide a "security assistant" for enterprise operations.

360 Intelligent Brain is a cognitive general large model self-developed by 360 Group. It ranks first in domestic large model capabilities in multiple third-party evaluations such as the Chinese general large model benchmark evaluation. As the leader of the "National Artificial Intelligence Standardization Overall Group Large Model Special Group" awarded by the China Electronics Standardization Institute of the Ministry of Industry and Information Technology, 360 Group participated in the formulation of national standards for large model testing.

The above is the detailed content of Xinhua News Agency and Peking University released a large AI model evaluation: safety and reliability are the focus, and 360 Intelligent Brain performs well. For more information, please follow other related articles on the PHP Chinese website!