English(EN) I am starting to realize why LLM's sometimes seem to hallucinate and come up with answers that are clearly wrong: They have been trained by skimming an internet

大型语言模型因互联网训练数据而模仿人类输出

作者 PulseAugur 编辑部 · [1 个来源] · 2026-06-02 10:43

大型语言模型之所以会产生幻觉或提供不正确答案，可能是因为它们接受了海量互联网数据的训练。这个训练过程涉及“浏览”人类生成的内容，意味着模型本质上是在模仿其训练集，而不是展现真正的智能。其输出反映了它们处理过的数据，而非真正的理解。 AI

影响表明大型语言模型的输出是对训练数据的模仿，而非真正的智能，这会影响用户的信任和期望。

排序理由该条目是一篇讨论大型语言模型训练和输出性质的观点文章。

在 Mastodon — mastodon.social 阅读 →

internet
LLM

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

报道来源 [1]

Mastodon — mastodon.social TIER_1 English(EN) · [email protected] · 2026-06-02 10:43

I am starting to realize why LLM's sometimes seem to hallucinate and come up with answers that are clearly wrong: They have been trained by skimming an internet

I am starting to realize why LLM's sometimes seem to hallucinate and come up with answers that are clearly wrong: They have been trained by skimming an internet full of human output. At best, they are acting just like we are. There is no intelligence, the system is just imitating…

报道来源 [1]

I am starting to realize why LLM's sometimes seem to hallucinate and come up with answers that are clearly wrong: They have been trained by skimming an internet

相关实体

相关话题