From cognition in the wild to cognition in silico: flourishing cultural diversity in AI

There has been unprecedented progress in artificial intelligence (AI), yet most current AI systems are trained on data from Western, industrialised societies. This narrow focus limits the generalisability and adaptability of AI models, which fail to reflect the full spectrum of human cognition and culture. Moreover, heavy dependence on pre-existing data overlooks how learning occurs in the wild, through direct interaction and feedback from the environment, and constrains innovation. Here, we integrate insights from anthropology, cognitive science, and computer science to argue for a broader, ecologically grounded approach to AI development. We examine examples from traditional cultures, particularly extant hunter-gatherers, whose modes of living reflect the ecological and social conditions under which human cognition evolved. These communities offer underexplored but highly relevant perspectives on core challenges in AI, from multimodal sensory perception to embodied and experiential learning and collective intelligence. These comparisons not only offer conceptual tools for AI design but also invite reflection on the reverse question: what can building AI teach us about human cognition, especially when inspired by ecologically diverse strategies? To develop embodied AI systems capable of engaging meaningfully with the world’s sensory and interactive richness, we must look to those who have long done so.