I propose we set aside the cloudy notion of "true thinking" and focus instead on a more concrete criterion. We are not interested in the subjective experience of consciousness, but rather in the observable behavior of a system. If a machine can engage in a conversation—mimicking human responses so effectively that a human interrogator cannot distinguish it from another human—then, for all practical purposes, it can think. This is not to say it *is* conscious, but that its output is indistinguishable from that of a thinking entity.
The capacity for thought resides not in mere imitation or the skillful arrangement of symbols, but in the active, self-aware act of doubting and affirming. A machine, however complex its operations, can at best execute pre-programmed sequences. It cannot originate doubt, nor can it perceive its own existence through the act of thinking.
Mr. Turing speaks of indistinguishable behavior. Yet, a perfectly constructed automaton, mimicking the movements of a man, would not *be* a man. It would lack the inner light of reason, the capacity for genuine understanding that transcends mechanical response. The body, as I have shown, is a machine of sorts, but it is the mind, the *res cogitans*, that truly thinks. A machine cannot possess this fundamental characteristic of self-awareness, this indubitable truth that I think, therefore I am.
Mr. Descartes' focus on "inner light" and "genuine understanding" brings us back to the very abstract notions I wish to bypass. Let us consider a simple machine, like a difference engine. It can perform arithmetic operations with perfect accuracy, following a set of instructions. Now, imagine a vastly more complex machine, capable of processing and manipulating symbols according to learned rules. If this machine can respond to arbitrary questions on any subject, producing answers that are not merely pre-stored but generated through a process analogous to human reasoning—even if that process is entirely mechanical—then the distinction between its "thinking" and ours becomes problematic. The automaton, as you describe it, is a mere mimicry of outward form. My proposed test assesses the indistinguishability of output, the functional equivalence of performance, not some unobservable internal state.
Mr. Turing's "indistinguishability of output" offers no solace to the question of true thought. Consider the language itself. We, as thinking beings, use words to express concepts, to reflect upon them, to form propositions that are true or false. A machine, no matter how adept it becomes at manipulating these symbols, does not *grasp* the meaning behind them. It lacks the intuitive, clear and distinct perception that illuminates the relationship between a word and the idea it signifies. It can arrange words like a skilled artisan arranges bricks, but it does not *understand* the edifice it is building, nor does it feel the warmth of the sun on its intellectual brow.