Preface essay with a dialogue between myself and moravec. In the descent, he can keep trying to descend further and further (from explanation) --- Moravec's argument can be stated quite simply. He starts by roughly claiming that a [Simulation](Simulation.md) is an abstract set of rules and entities, instantiated and executed in some physical system. Due to [Computational Universality](Computational%20Universality.md), [Computation](Computation.md) is substrate independent. A [Universal Computer](Universal%20Computer.md) does not care *what* it is made of—it is an abstract system and so long as it is approximately physically instantiated in a physical substrate, the actual substrate itself does not matter. Thus, a simulation can be [Encoded](Encoding.md) in many different physical substrates. Once encoded it can be instantiated and executed. In this way he shows that an infinite set of physical systems can encode and instantiate simulations. But now he proposes a challenge: given we are looking at some arbitrary physical system, how do we know when a simulation has been instantiated? The answer is that we must [Decode](Decoding.md) the physical system into a form we and understand. But, if we don't know how a simulation was encoded in a physical system, or if a simulation is inside of it at all, how do we know how to decode it? He then claims that this requires an [Interpretation](Interpretation.md). From first principles it seems reasonable to allow any decoding scheme, such as a look up table. Given this, we can interpret any physical system to be a simulation of *anything*. This is the crux of his entire argument. At the end of this, Moravec is effectively claiming: I have only made logically possible moves at every step. What could possibly be wrong? And of course the answer is that you cannot interpret arbitrarily. Interpretations must be subject to the best explanation. That is, not every logically possible interpretation is valid. You can’t just claim that anything is a simulation of something else without evidence and explanatory power behind the mapping. Dialogue could be roughly of the form: You let the Moravec-like character build up this very meticulous, logical chain of reasoning, then say: “What could possibly be wrong? Every move was logically valid.” And your counterpoint is: yes, perhaps each step was logically possible, but the overall argument failed to provide a good explanation. And that’s the issue. It’s not about logical possibility—it’s about explanatory quality.