Professor Zhang Qi from Fudan University: It is crucial to recognize the boundary of large language models' capabilities

On September 25, Golden Ten Data reported that the special forum “Decoding the Future: Global Digital Intelligence Trends” was successfully held in Pudong, Shanghai. Zhang Qi, a professor at the School of Computer Science and Technology of Fudan University, delivered a keynote speech on the topic of “Thinking on the Capability Boundary and Development of Large Language Models”. Zhang Qi believes that it will be of great value to combine the core capabilities of long context modeling, multi-task learning, cross-language transferability, and text production capacity with scenarios. For the large model of the future, it is important to have a clear understanding of the boundaries of its capabilities. If there is a consensus on the capability boundary, you will have a good grasp of the strength and direction of future investment. But it may take some time, maybe a year, or two years to complete the confirmation (cognitive boundary).