###### A Retreat From Explanation We have seen that Moravec's argument on is nonsense *on it's own terms*. There can hardly be a more lethal blow than that. But, as devastating a counter argument as it is, it still does not explain *why* it felt so hard to argue against. For that we must look a bit more closely at just how he *constructed* this argument. I used the word *construct* very deliberately. Arguments are constructed, they are built up, they are created. They have a *structure*. Table stakes required that this structure must be intrinsically consistent and free of contradictions—which we already saw Moravec's argument did not meet that basic requirement. But arguments also have *consequences*—they imply certain things. And it is this area that Moravec went off the rails. Let's unravel the thread of his argument again, this time look specifically at it's quality as an explanation. He started by saying that a simulation consists of a set of abstract rules that are then instantiated in some physical substrate—so far, so good. He then must answer *how* a simulation gets instantiated in a physical substrate. He references the concept of *encoding*, which has well defined meaning in this context. But instead of simply using that meaning, he skirts this definition and states that "something is palpably an encoding if it can be decoded". This is fine, but notice that he has moved the explanatory burden to rest entirely upon the concept of decoding now. In terms of argument structure, this is a now a *load bearing node*. This should raise a little alarm in our heads when we read it. Not because it is an illegal move—he is welcome to define terms however he likes—but because he is *erasing* part of the explanatory content of encoding. Semantic bait and switch. It is critical to keep in mind that ideas and explanations must solve some *problem*. That is their entire purpose. So we must evaluate Moravec's argument through the lens of what problems does it solve, and how well does it solve them? * Chops off consequences * this requires the understanding of the construction of a tree based argument [ChatGPT](https://chatgpt.com/share/e/682915a9-4b80-8006-8313-cf1522fea08c)