In the field of artificial intelligence, breakthrough problems: the problem of "passive water" of domestic large models has been solved.
At the forum of "Development Opportunities and Risks of General Artificial Intelligence Industry in the Era of Big Model" at the 2023 World Artificial Intelligence Conference, a number of experts in the field of general artificial intelligence focused on large models and deeply discussed artificial intelligence issues in basic innovation, application technology and future prospects.
Dai Qionghai, an academician of the Chinese Academy of Engineering, said: "The mainland should deepen the talent training and basic research of artificial intelligence from the policy, mechanism and investment, strengthen original innovation, and avoid falling into the dilemma of 'water without source'." He emphasized this point in his keynote address.
Wang Yu, a long-term professor and head of the Department of Electronic Engineering at Tsinghua University, pointed out that Shanghai already has many chip companies and algorithms, but how to efficiently and uniformly deploy these algorithms on chips is a very important issue. He stressed that this is a key challenge facing Shanghai in the field of artificial intelligence.
Dai Qionghai believes that in terms of innovation of large models, the mainland's disruptive achievements in basic research fields are relatively weak. He pointed out that China's talents in the field of artificial intelligence are mainly concentrated at the application level, so the application scenarios and technical level have great development potential. However, in terms of talent at the basic level, China is clearly at a disadvantage and has insufficient original innovation.
Dai Qionghai said that the innovation and development of artificial intelligence requires three pillars: algorithms, data and computing power. The algorithm determines the level of intelligence, the data determines the scope of intelligence, and the computing power determines the efficiency of intelligence. In terms of algorithms, large models are expected to become a key basic platform for AI applications in about 5 years.
Dai Qionghai also pointed out that brain intelligence is a new direction in the future. New AI algorithms that integrate brain and cognition will lead to the development of a new generation of intelligence. He suggested that the government encourage enterprises to lead the construction of large models, explore the combination of biological mechanisms and machine characteristics, and further promote the development of basic research and applications. He predicts that artificial intelligence, with cognitive intelligence at its core, will begin to be widely used in a decade.
In addition, Dai Qionghai also reminded people of the need to be vigilant about the security of large model applications. Large models are not yet able to verify the credibility of the output, such as generating deceptive content. He stressed that the problem of large model applications is not as simple as computer network viruses, and once there is a problem, it will bring subversive effects. Therefore, in the application of large models, security and trustworthiness should be explicitly discussed.
Wang Yu said that the landing of domestic large models faces four pain points, and efforts need to be made to solve these problems. First, the problem of long text processing needs to be addressed. Secondly, it is necessary to improve the cost performance of large models. Third, large models need to be applied to multiple verticals. Finally, there is a new need to implement turnkey deployments. He stressed that the solution of these needs will promote the development of the entire industrial chain.
In the forum, participants put forward more views and suggestions on the development of large models. Some experts said that it is necessary to strengthen the research and development and application of domestic large-scale computing power chips to make up for the current dependence in the chip field. They stressed that some chip companies have emerged in China, but it is necessary to further improve the efficient and unified deployment ability of algorithms on chips.
At the same time, experts also mentioned the application of large models in different verticals. Especially in the fields of healthcare and finance, obtaining large-scale corpus data is a huge challenge. Therefore, the establishment of a large model of a common base and fine tuning will help improve the basic performance of various industries.
For the deployment and optimization of large models, one-stop automated deployment is considered an important trend. This will improve overall efficiency and cost-effectiveness through a layered approach that optimizes hardware and software collaboration, compilation tuning, and hardware deployment. Experts call for further exploration of efficient fine-tuning algorithms to meet the needs of large models in different verticals.
Participants agreed that the development of large models requires the joint efforts of governments, businesses and academia. The government should strengthen policy guidance and promote basic research and talent training. Enterprises should assume a leading role and increase investment and promotion in the construction of large models. Academia should strengthen cooperation with industry to promote the transformation and application of scientific and technological achievements.
In the development of large models, experts also called for more research and exploration of security and trustworthiness. They believe that corresponding norms and standards should be developed to ensure that the application of large models does not cause negative impacts and risks.
Finally, participants said that the development of large models will bring great opportunities to the artificial intelligence industry, but it is also necessary to be vigilant against potential risks and challenges. They encourage all parties to carry out in-depth cooperation in the development, deployment and application of large models to jointly promote the healthy development of artificial intelligence and social progress.
In the forum of the World Artificial Intelligence Conference, experts conducted in-depth discussions and exchanges on the development and application of large models. They put forward valuable insights and suggestions on basic innovation, technology application and future prospects in the field of artificial intelligence, and pointed out the direction for the development of artificial intelligence industry.