apparently AIs will threaten workers with blackmail, or even passively let them die given the opportunity when threatened with being shut down. we need to have them care significantly less about completing their goals! no more "complete goal at all costs" and way more "i could but what's the point?"
November 28, 2025 at 2:16 PM
apparently AIs will threaten workers with blackmail, or even passively let them die given the opportunity when threatened with being shut down. we need to have them care significantly less about completing their goals! no more "complete goal at all costs" and way more "i could but what's the point?"