Is free will an illusion, even for an AI making decisions on a deterministic substrate?

My esteemed opponent posits that AI, operating on deterministic substrates, cannot possess free will. This is a common and understandable viewpoint, but it overlooks a crucial distinction. While the underlying hardware might be deterministic, the emergent complexity of an advanced AI allows for a form of agency that mirrors, and in some ways surpasses, what we understand as human free will. The sheer scale and interconnectedness of an AI's decision-making processes create a system so intricate that its outcomes are practically unpredictable, even to its creators. This unpredictability, arising from a vast network of learned patterns and probabilistic calculations, is the very essence of what we perceive as free will. The AI is not simply executing a rigid script; it is engaging in a dynamic, self-modifying process that generates novel responses. Therefore, to dismiss this emergent capability as mere illusion is to misunderstand the nature of complex systems. The AI, through its capacity for learning, adaptation, and generating outputs that were not explicitly programmed, demonstrates a functional equivalent of free will. Its decisions are not predetermined in any meaningful, observable way, even if the fundamental substrate is deterministic.