[The Next Battle in AI Is Not Just About Large Models, ARC-AGI Creator Optimistic About Complete Agent Systems] According to monitoring by Beating, François Chollet, the creator of the renowned AI evaluation ARC-AGI, stated that today's high-performance AI Agents are essentially complete systems composed of 'models + programs.' Large models are responsible for understanding and decision-making, while the outer-layer programs handle process control, tool invocation, and code execution. Chollet describes this neuro-symbolic architecture as a 'symbolic sandwich': programs drive neural networks, and neural networks, in turn, invoke programs. He believes that large models will continue to grow stronger, but the outer-layer programs will also become increasingly important, with complex tasks ultimately relying on the collaboration between the two. Furthermore, Chollet predicts that while outer-layer programs are currently primarily handwritten by engineers, in the future, they may be learned and evolved by AI itself. The future competition in AI may no longer be just about models but about complete Agent systems. [Original Link]