"Neither VLA nor World Models are the endgame, there will be models unique to the physical world" | Ant Lingbo Shen Yujun @AIGC2026
Ant Group's Lingbo Technology Chief Scientist Shen Yujun believes that current large models, which leverage decades of internet data, are insufficient for the physical world of robotics. He proposes AIGA (AI Generated Action) as the next phase, focusing on generating actions rather than just content, to address the data scarcity in robotics. Shen suggests that both Vision-Language-Action (VLA) and World Models are not the ultimate solutions, predicting a convergence towards a model uniquely suited for the physical world, capable of integrating diverse sensory inputs and predicting future states. AI
IMPACT Predicts a new paradigm for AI in robotics, moving beyond current models to specialized physical world intelligence.