# 前Anthropic研究员警告AI递归自我改进风险

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-09-11 10:10
- AIHOT 分数：35
- AIHOT 链接：https://aihot.news/items/cmtwbmn9o0azerolkr9dqdgq0
- 原文链接：https://x.com/rohanpaul_ai/status/2098232712331182266

## AI 摘要

前Anthropic研究员Jacob Coxon在NBC News采访中表示，AI递归自我改进（RSI）是最大风险信号：一旦AI能让自己变得更聪明，整个过程可能快到无人能控制。他指出AI已大幅改变研究岗位工作方式，呼吁公众关注RSI是否正在发生，并推动对其程度保持极度透明。

## 正文

Jacob Coxon on NBC News (ex-Anthropic+Open AI researcher, who resigend) talking about Recursive self-Improvement being the greatest risk.

NBC News: "What are some of the telltale that Americans, should start looking for? What are the warning signs that tell us it has gotten out of control?"

Jacob Coxon : "The main sign you should be looking for is if the AI can start making itself better, because at that point, the thing just spirals completely out of control.

AI can do many things, right? One of the things it can do is help with AI research. A lot of our job is now delegated to AIs that go and help us. Our job has changed dramatically over the last two years as the AIs have become better.

Once the AI can make itself smarter, the whole thing could go so fast that no one has any chance to control it.

I think being aware of whether that has happened is important. People call it RSI, recursive self-improvement. Being aware of whether RSI is happening, and encouraging extreme transparency about the extent of it, is critical."

----

From "NBC News" YouTube channel, (full video link in comment)