阐述了因DeepSeek横空出世引发的关于AI发展路径的思考。首先解释了为什么DeepSeek会引起全球性的科技震撼，接着讨论了“规模法则”（Scaling Law）是否已遇到天花板、发展“通用人工智能”（AGI）应选择什么道路、发展人工智能应该追求高算力还是高算效（高能效）、“开源”为什么有这么大的威力等问题。最后对中国在人工智能领域的实力提升和如何实现人工智能自立自强提出建议。
This article expounds the author's reflections on the development path of AI triggered by the emergence of DeepSeek. Firstly, it explains why DeepSeek has caused a global technological shock. Then, it discusses whether the scaling law has reached its ceiling, what path should be chosen for the development of artificial general intelligence (AGI), whether the development of AI should pursue high computing power or high computing efficiency (high energy efficiency), and why open source has such great power. Finally, it presents several views on China's strength in the field of AI and how to achieve self-reliance and self-improvement in AI.
[1] DeepSeek-AI, Liu A X, Feng B, et al. DeepSeek-V3 technical report[J]. Computer Science, 2024, doi: 10.48- 550/arXiv.2412.19437.
[2] DeepSeek-AI, Guo D Y, Yang D J, et al. DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning[J]. Computer Science, 2025, doi: 10.48550/arXiv.2501.12948.
[3] Kaplan J, McCandlish S, Henighan T, et al. Scaling laws for neural language models[J]. Computer Science, 2020, doi: 10.48550/arXiv. 2001.08- 361.
[4] Sutton R. The bitter lesson[EB/OL]. (2019-03-13) [2025-02-06]. http:// www.incompleteideas.net/IncIdeas/BitterLesson.html.
[5] Hinton. 放弃永生的凡人计算[EB/ OL]. (2023-06-11) [2025-02-06]. https://cloud. tencent. com/developer/ news/1099853.
[6] Muennighoff N, Yang Z T, Shi W J, et al. S1: Simple test-time scaling[J]. Computer Science, 2025, doi: 10.48- 550/arXiv.2501.19393.
[7] Stanford HAI. Artificial Intelligence Index Report 2024[R/OL]. [2025- 02-06]. https://aiindex. stanford. edu/ report.
