Reducing Catastrophic Risk from AI with Systematic Monitoring and Evaluation of Rogue AI Progression
A structured framework of behavioral indicators that may signal progression toward potentially catastrophic threats from artificial intelligence systems is presented, inspired by established methodologies in cybersecurity and national security.
T. Bauer, W. Kegelmeyer, Edmon Begoli et al.
· 0 citations