Can the"World Simulator"Evolve into General Artificial Intelligence?
Open AI's universal visual model,Sora,with its powerful text-to-video capabilities,has gained significant authority in depicting the contemporary world.However,the underlying large language model raises issues regarding the legitimacy of world representation,as well as cognitive risks and societal consequences that require a combined philosophical and technological critique.The paper analyzes multimodal large language model technologies,exploring the technical essence and application challenges of text-to-video technology.Additionally,it examines the technical rationality behind large language models and their impact on reshaping user value perceptions.The study reveals an epistemological paradox in the transition from technological generality to knowledge axioms.Generative artificial intelligence based on deep learning cannot truly comprehend the operational mechanisms of the world.It only processes logical relationships between information units and semantics.Moreover,existing generative large models still have shortcomings in counterfactual reasoning and compliance,falling short of achieving comprehensive understanding and representation of the world.Thus,as virtual-reality integration and human-machine fusion become new era challenges,understanding the relationship between virtual and real becomes a critical dimension to examine.This not only concerns the allocation of social resources but also affects human self-reconstruction and our understanding of the world.