芬兰软件工程师 Kristian Polso 于2026年11月8日发表博客文章,正式宣布将在职业工作中停止使用大语言模型(LLM)和 AI 代理 1。尽管拥有十年机器学习工程经验,Polso 并非因无知而排斥该技术,而是认为其带来的隐患超过了收益 1。他指出,多项研究包括微软内部调查均显示,LLM 的使用导致了“认知债务”的积累以及用户的过度依赖 1。此外,他认为虽然 AI 在解决特定技术卡点时能提供短暂帮助,但这种依赖性最终弊大于利,因此主张回归传统的人工编码方式 1。
除了对开发者个人能力的影响,Polso 还列举了代码质量、环境及隐私层面的多重顾虑 1。他批评 LLM 生成的代码往往看似完整却包含逻辑错误,并存在严重的可维护性问题 1。在环境方面,他提及谷歌曾在芬兰未经许可砍伐数百公顷森林以建设数据中心,引发争议 1。同时,他对数据隐私表示担忧,指出 OpenAI、Anthropic 等美国科技公司受《云法案》约束,可能侵犯欧盟公民依据 GDPR 享有的数据隐私权利 1。
Kristian Polso, a software engineer with ten years of experience in machine learning, announced on November 8, 2026, that he would cease using Large Language Models (LLMs) and AI agents in his professional work 1. The decision stems from concerns regarding cognitive degradation, poor code quality, environmental impact, and privacy risks associated with US-based tech companies, rather than a lack of technical understanding 1.
Polso cited research indicating that LLM usage leads to the accumulation of "cognitive debt" and excessive user dependency, referencing an internal Microsoft survey as evidence 1. He criticized the technology for generating code that appears complete but frequently contains logical errors and poses significant maintenance challenges 1. Although he acknowledged that AI offers brief assistance when resolving specific obstacles, he concluded that the resulting dependency and hidden dangers outweigh these benefits, prompting a return to traditional manual coding 1.
The engineer also raised geopolitical and ethical objections, pointing to Google’s controversial clearing of hundreds of hectares of forest in Finland without permission for data center construction 1. Furthermore, Polso expressed concern over potential violations of EU citizen data privacy under the GDPR, noting that US providers such as OpenAI and Anthropic are subject to the Cloud Act 1.
评论
还没有评论,欢迎留下第一条。