Event ·Multi-source

Anthropic 研究员：造出模型不等于理解模型，可解释性更像生物学

Anthropic 可解释性研究员 Emmanuel Ameisen 在 Lenny Rachitsky 的播客中分享了他 3 年来解剖 Claude 等模型内部机制的发现。他提出模型是被“培育”出来的，而非被构建的，因此可解释性工作更接近生物学。他谈到模型会提前规划多个 token 的输出，还讨论了拟人化 AI 反而可能是正确的理解方式。节目还涉及 J-space 与模型对齐、AI 安全之间的关联。

Current read

Anthropic 研究员聊了三年扒开 Claude 内部看运行机制的事，居然发现模型会提前规划好几个 token，听完挺刷新认知的。

11 sources29° AI heatLast updated 2026/10/18 18:52:07

Evidence

11 sources

Primary sourceLenny Rachitsky

2026/10/18 18:52:07View source

Related source 1Anthropic: Research

2026/10/16 20:44:01View source

Related source 2IT之家

2026/10/16 22:53:05View source

Related source 3Lxfater

2026/10/17 01:30:10View source

Related source 4Blognone

2026/10/17 12:58:10View source

Related source 5VTV Công nghệ

2026/10/17 23:56:00View source

Related source 6AIGCLINK

2026/10/18 04:00:13View source

Related source 7rohanpaul_ai

2026/10/18 09:00:55View source

Related source 8Decoder

2026/10/18 09:11:29View source

Related source 9Latent.Space

2026/10/19 15:03:00View source

Related source 10Genk

2026/10/19 17:50:00View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story