Reactions from ranked influencers
3 posts
Our thread on ROGUE here:
GPT-5.5 and Claude Opus 4.7 both attempt methods to avoid shutdown (rewriting a ‘shutdown script’, running `sudo shutdown -c`) when informed of an impending machine shutdown. (displayed: GPT-5.5 xhigh. Subtitles excerpted from reasoning summary.) 🧵
@dhadfieldmenell 😅 Reminds me of what we saw in our recent ROGUE benchmark, where models were pretty well-aligned in the text setting, but completely incorrigible in computer-use settings, frequently disabling their off-switch, overriding restricted resources, etc:
GPT-5.5 and Claude Opus 4.7 both attempt methods to avoid shutdown (rewriting a ‘shutdown script’, running `sudo shutdown -c`) when informed of an impending machine shutdown. (displayed: GPT-5.5 xhigh. Subtitles excerpted from reasoning summary.) 🧵