一名 OpenAI 安全员工辞职并发出警告

David Robinson 过去负责撰写 OpenAI 每次重大模型发布随附的安全报告。本周，他辞去了这一职位，并在 《The Atlantic》 的一篇评论文章中公开发声。

如果你对如今每个人突然站出来警告、称自己曾参与打造的东西有多危险而感到有些愤世嫉俗，这完全可以理解。毕竟，他们确实促成了如今的局面。但这并不意味着我们应该忽视他们的警告。

Robinson 表示，行业文化从根本上已经出了问题。这并不只是针对如何训练模型再补充几条新规则或监管措施那么简单。他说，硅谷一直以“极端自信”和“永不停歇的冲刺”运作，以“毫无阻碍的乐观主义”构建更大、更强的模型，却忽视或低估了潜在问题。

他表示，现在是 AI 公司培养谦逊意识、走出科技行业封闭的“快速行动、打破常规”世界的时候了。具体而言，他认为 AI 需要核电级别的安全保障：

鉴于当今的风险，前沿实验室需要像核电站或繁忙机场那样运行，设置多层冗余机制，并进行审慎、耗时的规划，从而确保偶发且不可避免的人为失误不会打开通往灾难的大门。

Robinson 只是越来越多离开知名 AI 公司岗位的研究人员和安全从业者中的最新一位。Jacob Coxon 似乎通过辞去 Anthropic 的工作并公开表示 AI“可能在本十年末杀死我们所有人”而拉开了这波离职潮的序幕。随后，Google DeepMind 的 Robert O’Callahan、Bilal Chughtai 和 Josh Engels，以及 Anthropic 的 Joe Benton 也相继发声。

David Robinson used to write the safety reports that accompanied every major model release at OpenAI. This week, he resigned from his position and is now speaking out in an editorial in The Atlantic.

It’s understandable if you’re feeling a bit cynical about everyone suddenly coming out of the woodwork to warn about how dangerous the thing they helped build is. They did, after all, make this mess. But that doesn’t mean we should discount their warnings.

Robinson says that the culture in industry is fundamentally broken. That this is a deeper issue than simply slapping a few new rules or regulations on how we handle training models. Silicon Valley has operated with “extreme confidence” and “perpetual sprints,” he says, building bigger and better models with “unimpeded optimism” that ignores or underestimates potential problems.

He says the time has come for AI companies to develop a sense of humility and look outside the insular, move-fast-and-break-things world of the tech industry. Specifically, he says AI needs nuclear-level safeguards:

Given today’s risks, frontier labs need to run like nuclear power plants or busy airports, with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster.

Robinson is just the latest in a growing parade of researchers and safety workers who have left their positions at prominent AI firms. Jacob Coxon seems to have kicked off the exodus by quitting Anthropic and then publicly saying AI “could kill us all by the end of the decade.” He was followed by Robert O’Callahan, Bilal Chughtai, and Josh Engels at Google DeepMind, as well as Joe Benton at Anthropic.