State Department Report Warns of AI Risks, Proposes New Safeguards
-
Report commissioned by US State Dept warns rapidly evolving AI could pose "catastrophic risk" to humanity, recommends limiting compute power for training AI models
-
Report advises US govt must act "quickly and decisively" on AI safety measures or risk "extinction-level threat to human species"
-
Recommends US agency set upper limit on compute power allowed for training any single AI model, require govt permission above certain threshold
-
Also advises making it illegal to open-source or reveal inner workings of powerful AI models to prevent losing control
-
Report authors argue current AI safety measures "inadequate" and govt oversight needed, but critics say recommendations amount to overreach that could stifle innovation