Künstliche Intelligenz, KI, Artificial Intelligence, AI, Quantencomputer, Robotik, ChatGPT, Grok, Transhumanismus

Demis Hassabis has highlighted a key limitation in current AI: language can describe reality, but cannot fully capture it. Hassabis, who has long studied “World Models,” notes that while language models learn much from human text—gaining knowledge of physics, psychology, culture, tools, and causality—text remains only a compressed trace of direct experience. Language can relay that a cup has fallen, yet it fails to encode the weight, grip, balance, friction, timing, and other physical aspects involved. World Models are designed to address these gaps by learning the underlying structure of physical reality—including persistence, movement, and feedback through actions. Language models rely on written information, but World Models aim to understand reality itself before it is put into words. The distinction lies in moving from knowledge based solely on explanation to knowledge grounded in real-world outcomes. 📰 @aipost