AnalysisPolicyAI ModelsCybersecurity
17 days ago
Featured
Palisade's Jeffrey Ladish discusses AI shutdown resistance & self-replication
Palisade Research finds that current AI models sometimes take extraordinary actions to avoid being turned off and can exploit cybersecurity vulnerabilities to self-replicate. Jeffrey Ladish discusses the implications for AI safety and the 'compute ecology' perspective where AI systems compete for resources.
·
17 days ago