The 2024 World Artificial Intelligence Conference was grandly held in Shanghai, bringing together top academic and industrial representatives from the world to discuss the future development direction of artificial intelligence. The core topics of this conference focus on the practical application of AI technology. Experts unanimously believe that AI has shifted from the theoretical research stage to a critical period of how to implement and generate value in actual scenarios.
Baidu founder Robin Li emphasized at the conference that the AI era should not blindly pursue "super applications", but should focus on "super capable" applications that can truly bring gain to the industry. He specifically pointed out that agents will become an important direction for future AI applications, and search technology may become the core entrance to agent distribution.
Ant Group Chairman Jing Xiandong raised three major challenges from the perspective of industry implementation: the lack of domain knowledge, the difficulty of complex decision-making, and the gap between dialogue interaction and effective collaboration. He believes that through the deep connection of professional agents, these problems can be gradually solved and AI technology can be promoted to bring intergenerational upgrades of services like the Internet.
Xu Li, CEO of SenseTime, further pointed out that the widespread application of AI requires breakthroughs in three bottlenecks: high-quality data, smooth interactive experience and technology controllability. He believes that only by making breakthroughs in these aspects can AI truly enter a "super moment" and bring changes to all walks of life.
In the development direction of large models, Zhang Peng, CEO of Zhipu AI, proposed that multimodal capabilities will become the core breakthrough of large models. This capability will bring AI closer to the way humans solve problems in the real world, thereby enhancing their application value. Yan Junjie, founder of MiniMax, emphasized that improving the accuracy of the model is the key to the implementation of the application, and the goal is to reduce the error rate from the current 30%-40% to the single digits.
Regarding the debate between open source and closed source models, Li Yanhong said that although open source models are valuable in specific scenarios such as academic research, closed source models have more advantages in commercial competition. This view sparked widespread discussion among participants.
The security and ethics of AI have also become important issues at the conference. Zhou Bowen, director of the Shanghai Artificial Intelligence Laboratory, pointed out that the current investment in AI security is far from enough, and only 1% of resources are used for alignment or security considerations. He called on the industry to strengthen research and investment in AI security.
Turing Award winner Yao Qizhi analyzed the risks brought by AI from a more macro perspective, including the expansion of network risks, the potential subversion of social structure and existing risks. He stressed that while promoting the development of AI, it is necessary to find a balance between controlling risks and maintaining innovation potential.
In terms of industrial transformation and opportunities, Huawei Cloud CEO Zhang Ping'an pointed out that AI innovation cannot be separated from the upgrade of computing power infrastructure, especially the release of AI computing power demand for end-side hardware to the cloud. Meng Pu, chairman of Qualcomm China, predicts that transferring 20% of generative AI workloads to the terminal side can save US$16 billion in computing resource costs by 2028. He believes that the close integration of terminals and cloud will promote the scale expansion of generative AI.
Alibaba Cloud founder Wang Jian pointed out from the perspective of the industry structure that although large companies may have more advantages in the development of AI, this does not mean tolerance. He believes that new AI giants will emerge in the future, and some existing large companies may also achieve rebirth through AI technology.
This conference fully reflects the transformation of the AI industry from theoretical research to practical application. How to make AI technology implemented in all walks of life and generate actual value has become the focus of attention of all parties. At the same time, issues such as AI's security and ethics have also received unprecedented attention, and the industry is working hard to find a balance between development and risk control.