A basic systems architecture for AI agents that do autonomous research — LessWrong
A lot of threat models describing how AIs might escape our control (e.g. self-exfiltration, hacking the datacenter) start out with AIs that are actin…
x A basic systems architecture for AI agents that do autonomous research — LessWrong Redwood Research AI Curated 190 A basic systems architecture for AI agents that do autonomous research by Buck 23rd Sep 2024 AI Alignment Forum 9 min read 17 190 Ω 71 A lot of threat models describing how AIs might escape our control (e.g. self-exfiltration , hacking the datacenter ) start out with AIs that are acting as agents working autonomously on research tasks (especially AI R&D) in a datacenter controlled by the AI company. So I think it’s important to have a clear picture of how this kind of AI agent c
saved by
related reading
- The Case Against AI Control Research — LessWronglesswrong.com
- How might we safely pass the buck to AI? — LessWronglesswrong.com
- Win/continue/lose scenarios and execute/replace/audit protocolsblog.redwoodresearch.org
- The case for ensuring that powerful AIs are controlled — LessWronglesswrong.com
- Many arguments for AI x-risk are wrong — AI Alignment Forumalignmentforum.org
- Building Effective AI Agents \ Anthropicanthropic.com
- AI Control: Improving Safety Despite Intentional Subversion — LessWronglesswrong.com
- Building Effective AI Agents \ Anthropicanthropic.com
- Reading Listblog.redwoodresearch.org
- AI Control: Improving Safety Despite Intentional Subversion — AI Alignment Forumalignmentforum.org
- Secretly Loyal AIs: Threat Vectors and Mitigation Strategies — LessWronglesswrong.com
- Hidden Technical Debt of AI Systems: Agent Runtimeleehanchung.github.io