Three major achievements released!This large model industry event was held in Chengdu

Mondo Education Updated on 2024-01-31

Promote the integration of data and reality, and promote the implementation of the industry!The 2023 Artificial Intelligence Large Model Benchmark Science and Technology Innovation Development Conference and the Midwest Digital Economy Conference were held in Chengdu

On December 28, the "2023 Artificial Intelligence Large Model Benchmarking Science and Technology Innovation Development Conference and Midwest Digital Economy Conference" with the theme of "Big Model, Big Future" was successfully held in Chengdu High-tech Zone.

The conference was hosted by Chengdu Media Group, undertaken by Chengdu Business Daily, co-organized by Red Star News, Red Star Capital Bureau, Sichuan Big Data Development Alliance, Sichuan Artificial Intelligence Society, and strategic cooperation with Chixing Venture Capital. Experts and scholars from the Chinese Academy of Sciences, Tsinghua University, Peking University, Tianjin University, University of Macau, etc., authoritative institutions such as the Artificial Intelligence Research Center of the Chinese Academy of Information and Communications Technology, China Software Evaluation Center, and more than 80 leading enterprises in the field of artificial intelligence large models gathered together to jointly develop new formats, new models and new paths for the integration of data and reality in the era of large models.

This conference not only invited a number of industry leaders, experts and scholars to share their experience and cutting-edge thinking, but also achieved three major achievements: building the OpenEval platform, establishing a large model application innovation collaboration center, and releasing the "2023 Artificial Intelligence Large Model Benchmark Test".

On December 28th, the "Product Innovation and Application Innovation of Artificial Intelligence Large Model" roadshow was also successfully held in Chengdu, 6 large model teams participated in the project roadshow activities, and showed the different landing scenarios of large models, revealing the diversified innovation trend of large models in thousands of industries, and investors such as China Commercial Youth Investment, New Micro Capital, Jingya Capital, Lighting Bourne, Dingshen Capital, Galileo Capital and other investors will also be invited to participate in the event. This roadshow opened up the path of communication and cooperation between the industry and the capital side for Chengdu enterprises.

The three major achievements were released at the meeting

Jointly help the development of large models

Mu Tao, Chairman of Chengdu Media Group, Secretary of the Party Committee and Chairman of Chengdu Media Industry Group, said in his opening speech that cutting-edge technologies need to be transformed into products or applications in order to give full play to their value. It is hoped that this conference can become a medium, bridge and link for communication in the field of China's digital economy, and a source of innovation and development in the field of large models in China.

The 2023 Artificial Intelligence Large Model Benchmark Science and Technology Innovation Development Conference and Midwest Digital Economy Conference is a grand event for the large model industry, which not only brings together leading enterprises, experts, scholars and authoritative institutions, but also condenses the collective wisdom of the industry and forms three major achievements.

Results: The openeval platform promotes the coordinated development of Chinese large model capabilities and security

Before the official holding of the conference, the conference invited two authoritative research institutions, the Artificial Intelligence Research Center of the China Academy of Information and Communications Technology and the China Software Evaluation Center, as well as experts and scholars from major universities to form a large model benchmark evaluation expert committee (hereinafter referred to as the "special committee") to conduct diversified evaluation of domestic large models.

Among them, the members of the Large Model Benchmark Expert Committee include Wei Kai, Artificial Intelligence Research Center, China Academy of Information and Communications Technology, Liang Bin, China Software Evaluation Center (Software and Integrated Circuit Promotion Center of the Ministry of Industry and Information Technology), Xiong Deyi, Department of Intelligence and Computing, Tianjin University, Zan Hongying, School of Computer and Artificial Intelligence, Zhengzhou University, Lin Zhouhan, John Hopcroft Computer Science Center, Shanghai Jiao Tong University, Wang Benyou, School of Data Science, Chinese University of Hong Kong, Shenzhen, He Junxian, Department of Computer Science, Hong Kong University of Science and Technology, Huang Hui, Faculty of Science and Technology, University of Macau.

Before the evaluation of the large model, the members of the special committee developed an automatic test platform for the evaluation of the large model, and named it "openeval", which means open evaluation and endless prospects for the development of the large model.

As the result of this conference, the openeval platform is committed to becoming a multi-dimensional, comprehensive and open evaluation platform for Chinese large model evaluation, exploring cutting-edge model evaluation methods, establishing large model evaluation standards and protocols, building high-quality large model evaluation benchmark data, comprehensively evaluating large model knowledge, capabilities, alignment and security levels, and publishing detailed evaluation reports to promote the coordinated development of Chinese large model capabilities and security.

At present, the OpenEval platform starts from four dimensions, including knowledge ability, value alignment, security and controllability, and professional fields, each dimension is divided into multiple sub-dimensions, and several datasets are included.

After the end of the conference, the OpenEval platform will continue to operate, and constantly update the evaluation data, innovate the evaluation method of large models, and organize online and offline evaluation activities, through which third-party evaluation institutions and enterprises can apply for and organize new evaluation tasks.

Results: The 2023 AI Large Model Benchmark Test*** builds a panoramic view of the industry

Based on the above-mentioned evaluation results, the OpenEval platform, China Software Evaluation Center, Natural Language Processing Laboratory of Tianjin University, Natural Language Processing Laboratory of Zhengzhou University, Red Star News and the Yangtze River Delta Data Element Circulation Service Platform jointly released the "2023 Artificial Intelligence Large Model Benchmark Test" at this conference

Xiong Deyi, a professor at the Faculty of Intelligence and Computing at Tianjin University and director of the Sino-foreign Joint Research Center for "Language Intelligence and Technology" at Tianjin University, told reporters that the release of * marks the successful completion of the first large-scale comprehensive evaluation of domestic open source and closed-source large models.

** pointed out that in the past year, the ability of the Chinese large model in subject knowledge and mathematical reasoning has been significantly improved, which shows the potential of the large model in subject knowledge mastery, handling complex problems and reasoning.

Based on the above evaluation results, "* has in-depth insight into the development trend of domestic large models, measures the capability boundaries of Chinese large models, identifies potential security risk points, constructs a Chinese large model capability and security panorama, and puts forward clear suggestions and directions for the future technology and industrial development of domestic large models. Xiong Deyi introduced.

At the same time, "* reviewed the development of artificial intelligence in the world in 2023, analyzed the opportunities and challenges faced by large models in the application field from a professional perspective, built a panoramic view for the industry, and explored the progress and breakthroughs of professional technology in the field of artificial intelligence.

It also suggests that large model providers should focus on value alignment while pursuing knowledge and capability improvement. As the capabilities of large models continue to evolve, this issue will become increasingly important.

Results: Establishment of a large-scale model application innovation collaboration center

At this conference, the Large Model Application Innovation Collaboration Center (hereinafter referred to as the "Center") was officially established.

The center will establish a large-scale model industry community in the form of an innovation consortium, and collaboratively guide large-scale model enterprises to communicate and interact with the real industryEstablish a large-scale model innovation service system to help the growth of large-scale model enterprises through various forms such as capital docking, growth tracking and business exchangesBuild a data asset service base to open up the circulation of data elements and the valorization of data assets.

At present, the first batch of members participating in the center are: Natural Language Processing Laboratory of Tianjin University, Red Star News, Chixing Venture Capital, Yuanhe Chenkun, Left Hand Doctor, United Data, 360 Group, iFLYTEK, Cloudwalk Technology, Daguan Data, Jirui Technology, Xiaoduo Technology, Zhongke Wenge, Deli Technology, Shanghai University of Finance and Economics Financial Model Lab Team, School of Data Science, Chinese University of Hong Kong, Shenzhen, University of Macau, School of Computer and Artificial Intelligence, and Foresight Research.

The Large Model Application Innovation Collaboration Center is committed to bringing together universities, scientific research institutions, and related enterprises in the industry to jointly promote the in-depth application and innovation of large models in various vertical fields, and promote industrial upgrading and economic development, including but not limited to trend research, technical exchanges, cross-border cooperation, achievement incubation and other tasks.

Industry leaders, experts and scholars gathered together

Discuss the development of large models and the artificial intelligence industry

At the conference, industry leaders, experts, scholars and academicians shared their cutting-edge thinking in different subdivisions around the two themes of "the development of the large model industry" and "the future of the large model industry".

Shi Yong, Academician of the Academy of Sciences for Developing Countries, Director of the Research Center for Virtual Economy and Data Science of the Chinese Academy of Sciences, and Counselor of the Chinese Academy of Sciences, Sun Maosong, Foreign Academician of the European Academy of Sciences and Executive Vice President of the Institute of Artificial Intelligence of Tsinghua University, and Ding Zhiming, Second-level Researcher of the Institute of Software of the Chinese Academy of Sciences, Director of the Spatio-temporal Data Management and Data Science Research Center, and Foreign Academician of the Russian Academy of Engineering, delivered keynote speeches from different perspectives.

In addition, Liu Xunzhe, Deputy Director of the Data Space Technology and System Research Center of Peking University and Deputy Director of the Institute of Software Research of Peking University, Li Sun, Senior Business Director of the Platform Department of the Artificial Intelligence Research Center of the China Academy of Information and Communications Technology, and Huang Hui, Associate Professor of the Faculty of Science and Technology of the University of Macau and Director of the Laboratory of Natural Language Processing and Sino-Portuguese Machine Translation at the University of Macau, also delivered wonderful speeches.

It is worth mentioning that Li Sun pre-released the "Practical Guide to Large Model Benchmarking" to be launched by the China Academy of Information and Communications Technology in the speech session.

At the same time, on behalf of the four guests of industry, academia, research and capital, Wang Jipeng, senior partner of Oriza Chenkun, Qin Qiangzi, secretary general of Sichuan Big Data Development Alliance, Li Jianping, chief scientist of the Ministry of Public Security, professor and doctoral supervisor of University of Electronic Science and Technology of China, and Xie Zhiyu, chief architect of Koala Youran Technology, held a roundtable discussion under the chairmanship of Guo Haoran, founding partner of Chixing Venture Capital, on the topic of "Opportunities and Challenges of Economic Development Empowered by Large Models".

The conference not only aims to share industry experience and cutting-edge thinking, but also aims to build a communication platform for leading departments, experts and scholars, start-ups and large enterprises and investment institutions in the field of artificial intelligence.

The results of this conference are fruitful and bright, and it also provides experience cases and effective paths for the future development of large models and artificial intelligence industries, which will help accelerate the implementation of various applications, thereby empowering thousands of industries, accelerating the transformation and upgrading of traditional industries, and promoting the deep integration of the digital economy and the real economy.

Domestic large models are in full bloom

For the first time, excellent cases of general + vertical large models were evaluated on different tracks

In fact, with the rapid development of large models represented by "ChatGPT", the curtain of the global "artificial intelligence" technology competition has also been lifted, which has triggered a new round of artificial intelligence application boom.

According to incomplete statistics, more than 100 artificial intelligence models have been released in China, and the outside world has always paid close attention to the development, capabilities, growth space and market potential of each large model.

In order to find out the development of the large model industry, the conference invited two authoritative research institutions, the Artificial Intelligence Research Center of the China Academy of Information and Communications Technology and the China Software Evaluation Center, as well as experts and scholars from major universities to form a large model benchmark evaluation expert committee to conduct diversified evaluation of domestic large models, and to check the pulse and guide the direction for the future development of large models.

Zan Hongying, a professor and doctoral supervisor of the School of Computer and Artificial Intelligence of Zhengzhou University, a member of the special committee, told reporters, "The evaluation standard can help us overtake in corners on the large model track." Research and application of large models need to rely on this standard, just like the baton of the college entrance examination falls on the first, and the national education will catch up in which direction. ”

The reporter learned that before the conference, the special committee innovatively divided the large model into two tracks: "general large model" and "vertical large model". For the general large model, the two dimensions of knowledge ability and value alignment are focused on the evaluation, while the subjective evaluation method is adopted for the vertical large model field, and more emphasis is placed on the feasibility of commercial value realization.

At this conference, a total of 5 large models performed in the 6 selection details of the "general large model", which showed the possibility that large models can change people's lives more widely and deeplyIn addition, 13 large models were selected as excellent cases of "vertical large models", which let the outside world see the potential of large models to land in various industries (see the end of the article for details).

On the whole, this evaluation takes "authority" and "academic" as the basic tone, establishes a fair, just and multi-perspective standard, shows the diversity of large models, can promote the orderly development of large models in the future, and creates a good industry ecology.

P.SList of excellent cases of innovative application of vertical large models:

First Prize: Daguan Data, Xiaoduo Technology, Zhongke Wenge.

Second Prize: Jirui Technology, Koala Leisurely, Touch Elephant Technology, Weimob, Left Hand Doctor.

Third Prize: Beijing Jiaotong University Transportation Model Team, Deli Technology, Xilingli Technology, Shanghai University of Finance and Economics Financial Model Laboratory Team, Shengshi Junlian.

Chengdu Business Daily-Chengdu released reporter |Yang Peiwen Chengdu release editor |KirbyScan the code**

Like and watch **Know

Related Pages