English(EN) If AIs are trained on content produced by themselves (or other AIs), they can eventually "collapse," losing the ability to identify edge cases, reinforcing thei

AI模型可能因训练数据包含AI生成内容而退化，专家正在研究

作者 PulseAugur 编辑部 · [1 个来源] · 2026-06-19 08:20

一种理论认为，如果AI模型使用其他AI生成的数据进行训练，它们可能会进入一种“崩溃”状态。这种现象可能导致其性能下降，使它们失去识别异常输入的能力，并最终产生无意义的输出。专家们正在研究大型语言模型出现这种退化的可能性。 AI

影响这一理论上的担忧凸显了AI发展中潜在的长期风险以及仔细管理数据的重要性。

排序理由该条目讨论了关于AI训练数据和模型退化的理论性担忧，引用了专家意见，而非具体事件。

在 Mastodon — mastodon.social 阅读 →

Mastodon

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] · 2026-06-19 08:20

If AIs are trained on content produced by themselves (or other AIs), they can eventually "collapse," losing the ability to identify edge cases, reinforcing thei

If AIs are trained on content produced by themselves (or other AIs), they can eventually "collapse," losing the ability to identify edge cases, reinforcing their own errors, and finally spouting gibberish. I spoke to some experts to see if LLMs are actually degrading. # TechNews …

链接 techopedia.com/ai-model-collapse-theory-r…

报道来源 [1]

If AIs are trained on content produced by themselves (or other AIs), they can eventually "collapse," losing the ability to identify edge cases, reinforcing thei

相关实体

相关话题