Abstract: From text and video generation to the deployment of embodied agents, both the academic and industrial sectors are hopeful that generative AI, with its multimodal integration capabilities, can autonomously develop world models, achieve human-machine alignment, and ultimately pass the embodied Turing test, thus completing the final stretch towards Artificial General Intelligence (AGI). However, the obsessions with aligning with humans, understanding the world, and achieving embodied generality have not only intensified the upgrade of massive computational power but also triggered global technological monopolies and competition for energy resources. Building on the demystification of the AGI illusion, and guided by affordance theory and embodied-enactive cognition, this paper advocates for “Affordance-Enactive AI” as a feasible future goal. Central to this approach are localized world models, niche generality, and autonomous agency actions.

Key Words: Generative AI; Artificial general Intelligence; Embodiment-Generality paradox; Affordance; Video games

This article can be downloaded here

https://jdn.ucas.ac.cn/public/uploads/files/67355edfee1c0.pdf