The distinction rests on the difference between what a thing does and what a thing is. A wax apple can do some of what a real apple does — sit in a bowl, reflect light, function as a visual referent for 'apple.' It cannot do what a real apple is for: nourish. The capacities of wax apples overlap with real apples in certain dimensions (visual appearance) and diverge categorically in others (biological function). Midgley's insight is that the overlap can be significant, even striking, without erasing the categorical divergence.
Applied to AI, the wax apple distinction cuts through a great deal of confusion. Large language models produce sentences that look like the products of thinking. The sentences are grammatically correct, contextually appropriate, rhetorically effective. They land with the weight of meaning. But the process that produced them is a statistical prediction over token sequences, not an act of understanding performed by a conscious being. The outputs may be indistinguishable from human-produced sentences. The processes that produced them are categorically different. And the categorical difference is what determines the moral significance of the output.
The standard objection is that the distinction is either mysterious or empty. If we cannot tell the difference between the wax apple and the real apple from the outside, why does the difference matter? Midgley's response was characteristically robust: the inability of a casual observer to tell the difference is not evidence that there is no difference. The real apple has properties — it nourishes, it decays, it reproduces — that the wax apple lacks. The real apple is embedded in living systems — trees, ecosystems, seasonal cycles — that the wax apple is not. A thorough examination reveals the difference immediately. The casual observation misses it because the casual observation is not looking for it.
The wax apple image also captures something important about the direction of concern. We do not worry that wax apples might secretly be nourishing us. We worry that they might be mistaken for real apples in contexts where the difference matters — when someone is hungry, when biology class is trying to teach what apples are, when children are forming their sense of what fruit is. Similarly, we should not worry that AI outputs might secretly contain understanding. We should worry that they might be mistaken for the products of understanding in contexts where the difference matters — when we are trying to learn, when we are trying to make decisions, when children are forming their sense of what thinking is.
The wax apple example appears in Midgley's later work, particularly in contexts where she was explaining to general audiences the difference between behavioral resemblance and experiential identity. Its charm lies in the everyday familiarity of the distinction — nobody has ever confused a wax apple for a real one in any consequential way, yet the philosophical point it illustrates is precisely the point the AI discourse has failed to grasp.
Surface vs. substance. The look of a thing and the being of a thing are different — and the difference becomes visible only when you ask what the thing is for.
Categorical divergence. Two things can resemble each other in important respects while differing in category — not merely in degree but in kind.
Casual observation misses the point. The inability to distinguish at a glance is not evidence of identity — it is evidence that one has not looked carefully.
Process vs. product. What a thing does and what produced it are distinct questions, and the moral significance of a thing often resides in the second.