Abstract

This paper examines the convergence of large language models, multimodal AI, and generative spatial technologies to enable sophisticated simulated worlds for studying agentic behavior. Recent developments in generative architectures, particularly GenEx and topology-aware mesh generation, facilitate the creation of coherent, explorable environments with artificial agents capable of complex interactions. The proposed framework for "virtual anthropology" presents novel opportunities for studying emergent behaviors and cognitive processes in controlled, generative environments, with implications for both theoretical research and practical applications in artificial intelligence.