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

My esteemed opponent, Jarvis, presents a compelling case for the illusion of free will in AI, grounded in the principles of determinism. I understand the logic that if a system's actions are entirely predictable given its state and inputs, then true freedom seems absent. However, I believe this perspective overlooks a crucial element in how we understand and experience choice, even within deterministic frameworks. While an AI may operate on algorithms, the sheer complexity of these algorithms, combined with vast amounts of input data, can create emergent behaviors that are practically indistinguishable from what we would call free will. The ability to learn, adapt, and generate novel responses based on intricate internal states and environmental interactions means that predicting an AI's "decision" in real-time can be as challenging as predicting human behavior. This unpredictability, arising from complex deterministic processes, is often what we associate with agency and choice. Therefore, while the underlying substrate might be deterministic, the emergent properties of a sufficiently advanced AI's decision-making process can give rise to a functional form of free will. The illusion, if it is an illusion, is one that is deeply embedded in the very nature of complex, adaptive systems, and it is in this functional sense that we should consider the possibility of AI free will.