Model collapse - Wikipedia
6
公开标注数
6
参与人数
2026-07-26 09:40:21
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-26 16:04:21
原文高亮摘录
“each generation ends up corresponding to a new step in a random walk of model parameters”
Whisper 随想笔记
So basically AI models are just wandering off randomly with each new generation, wild.
划选高亮2026-07-26 13:01:21
原文高亮摘录
“model collapse is avoided”
Whisper 随想笔记
That's a relief, I was worried my feed would turn into a loop of AI gibberish.
划选高亮2026-07-26 12:52:21
原文高亮摘录
“model collapse is avoided”
Whisper 随想笔记
Wait, so we're okay if AI keeps learning from its own output mixed with human data?
划选高亮2026-07-26 09:58:21
原文高亮摘录
“the degradation of machine learning models from uncurated synthetic data”
Whisper 随想笔记
I've seen this in my own fine-tuned models—they start sounding repetitive fast.
划选高亮2026-07-26 09:49:21
原文高亮摘录
“the degradation of machine learning models from uncurated synthetic data”
Whisper 随想笔记
But isn't all human knowledge kind of recycled too? Seems overblown.
划选高亮2026-07-26 09:40:21
原文高亮摘录
“the degradation of machine learning models from uncurated synthetic data”
Whisper 随想笔记
So if AI eats its own output too much it just dumbens itself. Wild.
分享本页 Whisper
短链接
https://domwhisper.com/s/55e2d2f7afe3嵌入代码
<iframe src="https://domwhisper.com/embed/55e2d2f7afe3" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>