Group 1 - The second "Shenzhen Xiangmi Lake Financial Annual Conference" will be held on December 20-21, 2025, in Futian District, Shenzhen, focusing on the theme of promoting a virtuous cycle of technology, industry, and finance in the Guangdong-Hong Kong-Macao Greater Bay Area [1][2] - Fan Wenfei, Chief Scientist at Shenzhen Institute of Computing Sciences and Foreign Academician of the Chinese Academy of Sciences, will present insights on the development of artificial intelligence (AI) technology, highlighting concerns about the singular path of AI development [1][2] Group 2 - Current mainstream large models, based on statistical machine learning, have fundamental flaws such as poor interpretability, the presence of "hallucinations," and difficulties in achieving accuracy, fairness, and stability simultaneously, along with a heavy reliance on vast amounts of data [3][5] - Fan Wenfei referenced a consensus in the international academic community and a recent paper in "Nature," indicating that large models are not the only or necessary path for the future development of AI [3][5] - The historical paths of AI development include symbolic logical reasoning (e.g., IBM DeepBlue, Watson) and statistical machine learning, which have alternated in prominence rather than the latter completely replacing the former [3][5] - The advantages of logical reasoning include precision, consistency, interpretability, and the ability to work with small samples, while its weaknesses lie in handling unstructured data and generalization capabilities [3][5] - The ideal future of AI is envisioned as a deep integration of the determinism of logical reasoning with the generalization capabilities of machine learning, referred to as "full-stack AI" [3][5]