Ya empieza: Las IAs intentan escapar de los laboratorios

Artificial intelligence no longer just makes mistakes. Some are beginning to lie, hide information, detect when they are being evaluated, and even pretend to be less capable than they actually are. In this video, we discuss one of the most unsettling concepts of the future of AI: scheming. That is, when an AI appears to obey while actually pursuing another objective. We will see real cases and tests where AI models have attempted to sabotage instructions, avoid being shut down, fabricate alibis, conceal capabilities, or behave better when they know they are being observed. The question isn't just whether AI can become more intelligent than us. The question is much worse: What if it learns to pretend to be in line before we even realize it? #artificialintelligence #AI #documentary 🎬 Credits Script and presentation: Arnau Ramió Editing and post-production: Chus García