一系列 Mastodon 帖子讨论了大型语言模型（LLMs）作为工具的使用，以及它们对人类作者身份和理解的不同影响。一位用户担心过度依赖 LLMs 可能会削弱人性，而另一位用户则指出 LLMs 生成的内容超出了人类的理解范围。Lorenzo Stoakes 的一段话表明，如果人类完全理解贡献，使用 LLMs 的担忧就会减少，并将其比作编译器。Linus 被引用为提倡不对工具使用设定严格规则，但 C++ 被提及为一个例外，这导致了放弃“辅助生成”标签以保持对人类作者身份的关注的决定。 AI
影响 探讨了 LLMs 整合到创意和技术工作流程中的哲学和实际影响。
排序理由 该集群由关于 LLMs 使用的观点文章和讨论组成，而不是特定的事件或发布。
在 Mastodon — mastodon.social 阅读 →
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
of an alcoholic. I use tools like text-to-speech to review how my writing sounds. I use spell check, etc. But... overuse of such tools takes away our humanity. My way of expressing myself, grammatical errors and all, is who I am. We must learn to love ourselves and expand outward…
I remain confident someone does and with enough energy I too could understand the details. By design LLMs generate tokens beyond the ability of anyone to understand. The vast amount of computation combined with data and probability pushes understanding beyond our reach. Matthew's…
Lorenzo Stoakes advised "don't send stuff you don't understand." Thus if each contribution is fully understood by the humans involved we need not be overly concerned a LLM may have been used in the process. Okay, fair enough. Are LLMs merely tools like an editor or compiler? This…
Being cautious about making moral judgements on tool use by others has merit. Linus makes this comment: 'So no. We're not going to make stupid rules that say "you can't use tools for X".' But of course Linux patches do follow rules. As far as I'm aware, C++ has landed in "you can…
For the moment I have decided to avoid engaging with LLMs. However, my thoughts on LLMs continues to evolve. I found this post by Linus on dropping "Assisted-by" interesting: https:// lkml.org/lkml/2026/10/13/3242 Linus sees LLMs a tool akin to a compiler or editor. Matthew Brost…