科技爱好者周刊#291：AI 没有护城河
English
This article from Issue 291 of "Tech Enthusiast Weekly" titled "AI Has No Moat" was published on March 1, 2024. The core argument is derived from an alleged report written by a Google internal researcher, which points out that even giants like Google and OpenAI struggle to build "moats" - persistent competitive advantages and monopoly profits - in the field of AI.
The main arguments include:
- Open Source Core Technologies Most generative AI core technologies are open source, accessible to anyone, making it difficult for exclusive technologies to become key advantages.
- High Model Substitutability Although AI models from different companies vary in strength, their core functions are similar and easily substitutable. New features are quickly caught up by other companies.
- Competitiveness Depends on Computing Power and Data The core competitiveness of AI lies in the richness of computing power and training data, both of which depend on financial strength.
- High Mobility of Researchers AI researchers change jobs frequently, and many come from academia where they can freely publish their findings, which不利于 companies to conserve core technologies.
In conclusion, the article argues that no aspect of the AI industry has formed effective moats. The companies that will ultimately survive are likely to be those with the richest resources, while smaller companies rushing in will find it even harder to succeed.
中文
这是《科技爱好者周刊》第291期题为"AI 没有护城河"的文章，于2024年3月1日发布。
这篇文章的核心观点源于一份据称是谷歌内部研究员撰写的报告，该报告指出，即使是谷歌和OpenAI这样的巨头，在AI领域也难以构建"护城河"，即难以形成持久的竞争优势和垄断利润。
文章主要论点包括：
- 核心技术开源 生成式AI的核心技术大部分是开源的，任何人都可以获取和使用，这使得独家技术难以成为关键优势。
- 模型替代性强 各公司的AI模型虽然有强弱之分，但核心功能类似，容易被替代，新功能也很快会被其他公司追赶。
- 竞争力依赖算力和语料 AI的核心竞争力在于算力和训练语料的丰富程度，这两者都取决于经济实力。
- 研究人员流动性高 AI科学家跳槽频繁，且许多研究人员来自学术界，可以自由发表成果，这不利于公司保守核心技术。
综上所述，文章认为AI行业的各个方面都没有形成有效的护城河，最终能存活下来的公司很可能是资源最丰富的那些，而小公司匆忙入局则更难获得成功。