The newsletter discusses a growing shift in artificial intelligence research and investment toward “world models,” systems designed to understand and interact with physical and spatial environments, beyond text prediction. Startups such as World Labs and AMI Labs are attracting significant funding with technologies that generate explorable 3D worlds and support spatial reasoning, signaling investor interest in AI that operates meaningfully in real-world contexts. At the World Economic Forum in Davos, leading researchers including Yann LeCun and Demis Hassabis expressed skepticism that large language models (LLMs) alone can achieve artificial general intelligence (AGI), arguing instead that true AGI requires models capable of learning about space, physical interaction and causal dynamics. While world models could represent a breakthrough toward embodied intelligence, the article also cautions that imbuing AI with deep physical understanding poses societal and safety risks if misused.

Source:
https://www.thedeepview.com/newsletter/will-world-models-beat-llms-on-the-path-to-agi