DeepSeek Model Advantages: Perspectives from Computational Power and Cost

        Speaker: Researcher Wang Zeke

        "Hundred Talents Program" Researcher, affiliated with the Institute of Artificial Intelligence and Collaborative Innovation Center for Artificial Intelligence at Zhejiang University's College of Computer Science and Technology. Founder of the RC4ML Laboratory, specializing in constructing low-cost, high-performance AI large-scale model systems using heterogeneous hardware (e.g., the LoHan system). Recipient of the 2022 First Prize for Scientific and Technological Progress (Science and Technology) in Higher Education Institutions, the Huawei Spark Award, and the 2023 IndySCC International Supercomputing Championship title.

        Abstract

        The emergence of advanced AI large-scale models like DeepSeek has demonstrated remarkable potential in human-like capabilities such as logical reasoning, mathematical computation, and complex Q&A tasks. However, the escalating computational costs associated with these models have become a critical bottleneck constraining both technological advancement and commercial deployment

Chinese website redirect link:https://www.bj-first.org.cn/single?id=331