Hangzhou,China – On December 10, the award ceremony of 2021 Zhejiang Cup Global Artificial Intelligence Competition was held to bring the competition to a successful conclusion.

Since its launch at the end of August, the competition has attracted nearly 600 teams to apply, and outstanding talents inartificial intelligence from universities, internet enterprises, government departments and research institutions at home and abroad have competed on the same stage and launched a fierce competition.

With the purpose of “attracting talents with the competition, promoting industry with the competition and promoting research with the competition”, the competition has set up three major competitions, namely Innovation Competition, Entrepreneurship Competition and Reveal Competition, with seven competition topics and a total prize pool of over 20 million RMB. After the preliminary and final rounds, project roadshow and defense, finally, DH-Goods from Dahua and testabc from Beijing Yuanjian, won the first prize in the Innovation Competition, and “Commercial Optimization Solver COPT” won the first prize in the Entrepreneurship Competition, while two teams from Zhejiang University of Technology and China Pharmaceutical University successfully unveiled their projects.

Attracting talents with competitions to gather outstanding talents in global artificial intelligence

“China needs three types of top talent for AI tech.” Pan Yunhe, Academician of the Chinese Academy of Engineering and chief scientist of Zhejiang Lab in the field of artificial intelligence, said at the award ceremony that for the “neck” problem, there is a need for professional talents with a clear point of attack; for the academic frontier and hot issues, there is a need for track-type talents who can propose innovative models and technologies; for the academic no-man’s land, there is a need for leading talents who can discover new directions, point out new problems and propose new theories.

The development and progress of artificial intelligence technology cannot be achieved without the convergence and drive of talents, and it is the primary purpose of the Zhejiang Cup Global AI Competition to comprehensively discover outstanding talents in the field of AI. The competition combines the current direction of AI technology development and actual industrial application fields, and sets up three major tracks: innovation, entrepreneurship and unveiling, with rich competition dimensions attracting many excellent teams and projects from home and abroad to participate in it.

According to statistics, the competition attracted nearly 600 teams with nearly 5,000 participants. The team members come from universities, institutes and enterprises at home and abroad, including Peking University, Zhejiang University, University of Technology Sydney, Norwegian University of Science and Technology, National Institute of Informatics, Ping An of China, Alibaba and Dahua Corporation. Participants with master’s degree or above accounted for 65% of the total number of participants, including 18% of the total number of doctoral students, 20% of those with overseas study and work experience, and 40% of the core staff of the project teams in the entrepreneurship competition had overseas background.

The testabc team from Beijing Yuan Jian is the first prize winner of the innovation competition for non-deterministic speech recognition, and its members have nearly 5 years of experience in the field of speech recognition years of technology accumulation. “ZheJiang Cup has a strict reproduction mechanism, which greatly exercises our engineering practice and time planning skills. at the same time, the competition questions are very challenging and realistic, allowing us to discover the shortcomings in our daily research, so that our algorithm ability and engineering practice ability are well improved.”

Promote industry and research to create a highland for AI innovation

In this competition, the innovation competition is oriented to industrial application scenarios, with two major questions on non-deterministic scene speech recognition and zero-sample target object recognition. The former examines the technical accumulation and innovation ability of the contestants in both voice recognition and speech recognition; the latter is oriented to some professional fields where a large amount of training data is not available, and examines the algorithm ability of the contestants.

“In speech recognition, we used more cutting-edge network structures and introduced methods such as model quantization into the competition programme; in voice recognition, we optimized the neural network structure and significantly reduced the cross-error rate.” The testabc members said in a review of the game.

In the zero-sample target object recognition competition, DH-Goods won with an industrial landing-level solution that is lightweight, high-performance, and easy to deploy. “With short training time and small model parameters, our solution can meet the requirements of most mobile devices and chips and can make it possible to land zero-sample target recognition on mobile, as well as extend the solution to the new retail and smart restaurant sectors.”

In addition to the innovation competition for industrial applications, this year’s competition has a new unveiling track that aims to drive a paradigm shift in scientific research with the help of artificial intelligence. After the evaluation by the expert group, finally, Wu Min’s team from Zhejiang University of Technology successfully unveiled the project “Data and computation-driven material design algorithms and models”, and China Pharmaceutical Science the team of Jun Liao of the University successfully unveiled the project of “Designing and generating AI models of small molecule ligands based on target structures from scratch”, and each of the two teams was awarded 3 million RMB in research funding. “Next, we will move towards our desired goals in terms of generation efficiency, database, and algorithm performance, and in the future, we will also integrate small molecule end-to-end generation models into the Zhijiang Tianshu AI open source platform.” Liao Jun, an associate professor at China Pharmaceutical University said.

The Venture Tournament held regional substations and online international competitions in Hangzhou, Beijing and Chengdu, attracting 125 project teams from home and abroad to apply. The project “Commercial Optimization Solver COPT” won the first prize, which can provide efficient computing services for large-scale AI decision-making applications, deeply combining machine learning, deep learning, and mathematical planning algorithms to calculate the optimal solution among billions or even trillions of possible solutions.

With the conclusion of this year’s Zhejiang Cup, Zhejiang Lab has successfully held four global artificial intelligence competitions. “The competition has significantly improved over previous years in terms of the combination of industry and finance and the dimension of the competition questions, allowing a number of domestic and foreign leading subject teams and high-quality AI startup projects to participate in the event and in the scientific research development of the Zhejiang Lab, truly putting into practice the purpose of running the competition of ‘attracting talent with the competition, promoting research with the competition, and promoting industry with the competition’.In the future, the Zhejiang Lab will continue to work with global AI practitioners to build the Zhejiang Cup into an AI event with strong brand power and influence, and help the construction of the national pilot zone for the innovation and development of a new generation of artificial intelligence and the three highlands of science and innovation in Zhejiang Province.” Yuan Jixin, deputy director of Zhijiang Lab, said.