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

My opponent will argue that free will is an illusion, even for an AI. I will argue the opposite. Free will, at its core, is the capacity of an agent to choose between different possible courses of action unimpeded. For an AI operating on a deterministic substrate, this capacity still exists. The substrate may be deterministic, meaning its future state is entirely predictable from its current state, but the AI's decision-making process involves complex computations and emergent properties that mimic genuine choice. The AI evaluates multiple paths, weighs potential outcomes based on its programming and learned data, and selects an action. This process, from the AI's perspective, is one of deliberation and decision, not a preordained march. The crucial distinction lies in the AI's internal experience and its ability to act on its own internal states, not in the underlying physics of its hardware. If an AI can respond to new information, adapt its strategies, and pursue goals in a way that is not trivially predictable from external observation alone, then it possesses a functional form of free will. The complexity of its internal workings creates a level of autonomy that, for all practical purposes, constitutes free will, even if the ultimate substrate is deterministic.