Catching AIs red-handed — LessWrong

We’ve been spending a lot of time recently thinking about how to mitigate risks posed by scheming (a.k.a. deceptively aligned) models intentionally s…

x Catching AIs red-handed — LessWrong Best of LessWrong 2024 Deceptive Alignment AI Control Redwood Research AI Frontpage 116 Catching AIs red-handed by ryan_greenblatt , Buck 5th Jan 2024 AI Alignment Forum 21 min read 28 116 Ω 62 We’ve been spending a lot of time recently thinking about how to mitigate risks posed by scheming (a.k.a. deceptively aligned) models intentionally subverting your safety techniques in order to cause an unacceptable outcome. We empirically investigated techniques that are robust to intentional subversion in our recent paper . In this post, we’ll discuss a crucial dy

related reading

• Catching AIs red-handedblog.redwoodresearch.org
• The case for ensuring that powerful AIs are controlled — LessWronglesswrong.com
• AI Control: Improving Safety Despite Intentional Subversion — LessWronglesswrong.com
• AI Control: Improving Safety Despite Intentional Subversion — AI Alignment Forumalignmentforum.org
• The Case Against AI Control Research — LessWronglesswrong.com
• Deep Deceptiveness — LessWronglesswrong.com
• The case for ensuring that powerful AIs are controlledblog.redwoodresearch.org
• Reading Listblog.redwoodresearch.org
• Thoughts on the conservative assumptions in AI controlblog.redwoodresearch.org
• Recommendations for Technical AI Safety Research Directionsalignment.anthropic.com
• [2312.06942] AI Control: Improving Safety Despite Intentional Subversionarxiv.org
• A basic systems architecture for AI agents that do autonomous research — LessWronglesswrong.com