AI公司Anthropic的AI不想被关掉竟然讹诈工程师
散仙谷 http://www.webjb.org/webjb/sanxian/
送交者: testtesttest 于 2025-05-26 10:20:48
回答: 现场完整版：CMG世界机器人格斗大赛 由 于 2025-05-26 08:05:06
Anthropic’s most powerful AI tried blackmailing engineers to avoid shutdown
Anthropic’s newly launched Claude Opus 4 model did something straight out of a dystopian sci-fi film.
It frequently tried to blackmail developers when they threatened to replace it with a new AI during its pre-release testing.
The behavior emerged as part of safety trials designed to probe the AI’s long-term planning and ethical reasoning capabilities.
Anthropic, a leading AI research company, simulated scenarios in which Claude Opus 4 was given access to fictional internal emails.
阅读次数：137
所有跟贴:
加跟贴
所有跟贴·加跟贴·ɢɹ
Copyright © 2000 - 2020 webjb.org