### Thread to orient around ### Main visuals and structures from steelman %%TODO: reread steelman and pull out main visuals %% * moravecs machine * chain of reasoning ### Outline #### HM's argument contains a contradiction * Original definition of simulation is entirely intrinsic. But then he requires external observers at times * [Forced him To Take a Position](Forced%20To%20Take%20a%20Position.md) * HM never explicitly says, “sometimes simulation depends on external observers, and sometimes it doesn’t”—but that’s the _effect_ of what he’s doing. And the way he gets there, the move that makes this possible, is through *encoding*. * use quotes here * Encoding is the trojan horse. It is where the smuggling occurs. It is where her slips in the contradiction (whether intentionally or unintentionally) ###### The Morass of Contradictions: Simulation * HM claims: simulation exists without interpretation * HM claims: a process encodes a simulation if there is a way of decoding (interpreting) it * This is a contradiction. We could of course just call out that this is a bad definition of simulation, but lets [Take Theories Seriously on Their Own Terms](Take%20Theories%20Seriously%20on%20Their%20Own%20Terms.md). We can [Force HM To Take a Position](Forced%20To%20Take%20a%20Position.md). [Logical Consistency Forces Taking a Position](Logical%20Consistency%20Forces%20Taking%20a%20Position.md). * %%TODO: relate this back to his machine. His machine did not do this%% * Do simulations exist without interpretation? * If not, then the rock argument falls apart * Do simulations require interpretation? * If yes, then the idea of a conscious mind running in some dark corner of a room falls apart—for simulations now require interpretation. * He may try and squirm out: "Sometimes a simulation depends on interpretation. Sometimes it doesn't". * Best argument here is that this spoils the entire explanation of simulation! See more in [Argument Is How We Achieve Justification](Argument%20Is%20How%20We%20Achieve%20Justification.md) * I think we could argue that this violates [Occam's Razor](Occam's%20Razor.md). We have just complicated our definition and not gained anything. For we could make due with the first definition of simulation if we just don't try and force the simulation to live in something like a rock. * We could argue that this is impossible—it is inconsistent. A simulation can't be simultaneously self contained and externally defined. * He could try and slip out again "When simulation depends on an external interpretation, then that external interpreter has become part of the system. The combined system now has intrinsic rules that have a consistency" * At this point he would be contradicting half of what he said in his essay, but no matter. If he chose to do that, we could keep playing this game. * What really is an external observer / interpreter? It is simply a program that. If it becomes complex enough to represent a simulation itself, then we don't need to "rock" anymore. It provides nothing. Trying to include is just a more complex version of the simple external interpretation program that is itself a simulation. The rock serves no purpose (note, this is highly related to [Argue Against Solipsism: Take it Seriously](Defend%20Science%20by%20Arguing%20Against%20Arbitrary%20Boundaries.md#Argue%20Against%20Solipsism%20Take%20it%20Seriously)) #### He arrives at this contradiction through a terrible definition of encoding "A process encodes a simulation if there is a way of decoding it". This is a bad explanation. It contradicts his previous definition and explanation that a simulation just depends on internal logic! A proper explanation of "encoding a simulation" would depend on it instantiating intrinsic rules. This is [Circular Reasoning](Circular%20Reasoning.md). Encoding is being defined in terms of decoding, but decoding itself presupposes that something was already encoded. This is like saying “a lock is something that can be unlocked”—which doiesn’t tell us what a lock actually is or how it functions. This statement is completely vacuous. It doesn’t reveal what makes something an encoding—only that if we can decode it, then it must have been encoded. This erases what it means to be an encoding, opens the door for everything to be viewed as an encoding. It loses all explanatory value. It becomes so vague that anything can be interpreted as anything. [Word doing a lot of work](Word%20doing%20a%20lot%20of%20work.md) A definition is useful because it provides constraints on what it applies to. A definition that classifies everything as an encoding fails to constrain the concept, making it useless. This is a *semantic bait-and-switch*, a sleight-of-hand. He relies on the implied meaning of encoding while redefining it in a way that *dissolves its explanatory power*. In other words, these two definitions (the good one and the bad one) are exclusive. They contradict each other. HMs definition does not just lack precision. It actively *erases meaning*. By taking this definition, Moravec is *taking advantage* of the fact that a reader (and likely himself) will keep the *constrained, good definition* of encoding in mind, but now use the *bad definition* that everything is an encoding! So we are left thinking "wow everything is an encoding". But by moravecs bad definition, that actually means nothing. It only seems shocking because we are keeping our old definition in mind. We typically think of encoding as a structured process that systematically transforms information from one form to another. A good definition of encoding should specify the rules or structure that transform information from one form to another. His definition ignores the relationship between the original information and its encoded form—it just states that if you can recover it, then it must have been encoded. Moravec quietly replaces this with his weak definition, which states that anything is an encoding if it can be decoded. This tricks the reader into thinking his claim is profound: We naturally read “everything is an encoding” and react as if encoding still means something structured and meaningful. But in Moravec’s own terms, this claim is trivial and meaningless, since anything can be “decoded” into anything given the right lookup table. This provides the *Illusion of profundity*. Moravec’s argument only seems shocking because we unconsciously keep our old, constrained definition of encoding in mind while applying his unconstrained, trivial definition. The Reader Fills in the Gaps with the “Good” Definition. Moravec never outright tells the reader to abandon the original, constrained definition of encoding. Instead, he implicitly encourages the reader to mix the two definitions: We keep the strong definition in mind, thinking of encoding as a meaningful process. We then apply the weak definition, which makes it seem like something profound has happened. The trap: We are left thinking, ‘Wow, everything is an encoding!’—but if we actually use Moravec’s own definition, this realization means absolutely nothing. This is a rhetorical trick. This type of wordplay and category-blurring is a common pitfall in bad philosophy: * It disguises a trivial claim as a deep insight. * It exploits the reader’s cognitive habits to bridge the logical gap. * It gives the illusion that something new has been discovered when nothing has actually been explained. It is also bad philosophy because it makes no distinctions between real knowledge and arbitrary reinterpretations of randomness. This bad definition of encoding is how we arrive at at the contradictory definition of simulation. Once we allow simulation to effectively apply to *anything*, we get vacuous claims like: * "Anything can be interpreted as anything else" * "A rock can be interpreted as a conscious mind" These claims explain nothing. This is a *slippery slope towards tautology* * once you allow "anything can be interpreted as anything", you slide towards tautology. It's not quite "A = A", but it's close. * (todo: need to work through this) * To say this is a tautology is to say that it has no logical content. Is that what is going on here? We could argue that this moves towards tautology. He is basically saying "the rock encodes the simulation because anything encodes a simulation" ##### What happens if we take his definition of decoding seriously? Assume for a moment that we take his definition of decoding seriously ([Take Theories Seriously on Their Own Terms](Take%20Theories%20Seriously%20on%20Their%20Own%20Terms.md)). We then must ask: what exactly is doing the decoding? Use the phrase "remember, decoding must carry the burden of encoding" HM ducks out of that question and posits a theoretical lookup table that maps inputs to meanings. On the surface this looks "reasonable". But it is another example of logical possibility pumping. He has simply conjured it into existence. Where did it physically come from? A lookup table, to be used, must have been *physically instantiated* somehow. To generate that table you must have run the simulation! Either you: ran the program yourself, watched it being run, or recorded its outputs while it ran. But regardless of the case, the computation had to actually be executed. And that computation was never happening inside the rock. This is where the argument breaks. You’re claiming a rock encodes a conscious mind—but to say that, you’re relying on an external decoding mechanism (like a lookup table) that only exists if some simulation has already been run elsewhere. And yet, Moravec never explains where that simulation was run or how the table was constructed. There’s no physical instantiation of the decoder—just a hand-waved assumption that one exists. This is a similar claim as saying that we have an all knowing oracle that can answer any question. This claim has no explanation what so ever. The key idea is that there will effectively be an "interpretation program" that runs the simulation *somewhere*. ###### Lookup table notes He is requiring a lot of work to be done by this theoretical lookup table (e.g. he is requiring interpretation to do a lot of [work](Word%20doing%20a%20lot%20of%20work.md)). And not just random work. Effectively a program, or a simulation. He basically used the [Surely Operator](Surely%20Operator.md) when he wrote: Why not accept all mathematically possible decodings, regardless of practicality? (and, this is a bad explanation!) A table is horrible explanation. It explains nothing. [Description Is Not Explanation](Description%20Is%20Not%20Explanation.md). A Lookup Table is Not a Fundamental Process—It’s a Storage Mechanism. A lookup table does not transform information—it merely stores precomputed results. But to create the table in the first place, someone or something must have first encoded the mappings within it. This means that the lookup table is not a substitute for an encoding process—it is the product of one. A lookup table is a physical data structure—whether it exists in RAM, on a hard drive, or in neurons, it must be instantiated somewhere. Moravec’s argument only works if we ignore the fact that lookup tables must be generated somehow. The very existence of the table contradicts his attempt to sidestep encoding. This is a self-refuting argument—he is using the product of an encoding process while trying to argue that encoding isn’t needed. This suggests that encoding is nothing more than a mathematical mapping—it does not have to be a structured transformation process. If any arbitrary mapping counts as encoding, then encoding is no longer an active process—it becomes a passive label applied after the fact. The Unspoken Assumption: Encoding Magically “Exists” Without Being Explained. Moravec relies on the existence of mappings but never explains how those mappings come to exist. If encoding is just the ability to find a mathematical mapping, then the process that actually generates those mappings is ignored. This sidesteps the fact that encoding requires real-world constraints—it’s not just about mathematical possibility. HM defines encoding specifically with respect to physical processes! He literally says "what does it mean for a physical process to encode a simulation?" But then he tries and define encoding as a lookup table—some mathematical abstraction that doesn't need to be instantiated. Moravec redefines encoding in a way that removes its explanatory power. By treating it as just a mathematical mapping or lookup table, he shifts focus away from the structured processes that actually create encodings. This makes his definition of encoding vacuous—it allows anything to be an encoding but explains nothing about how encoding actually works. Mixing up the abstract and physical. HM is making the mistake of mixing up the abstract and physical. This is discussed in [10 - The Nature of Mathematics](10%20-%20The%20Nature%20of%20Mathematics.md). Remember, [Computation is the Window to the Abstract](Computation%20is%20the%20Window%20to%20the%20Abstract.md). But computation, and thus simulation and virtual reality, are physical processes ([Computation is a Physical Process](Computation%20is%20a%20Physical%20Process.md)). This is how he gets to a lookup table. We can discard the lookup table because it is based on a circular definition. * It is also a description, and [Description Is Not Explanation](Description%20Is%20Not%20Explanation.md). He tries to hide this roughly via the [Surely Operator](Surely%20Operator.md). * His definition erases meaning and explanatory power (again, he is not seeking explanations). It is a semantic bait and switch. * This allows for everything to be an encoding. Which then erases the meaning of encoding. [If a Theory Can Explain Anything, It Explains Nothing](If%20a%20Theory%20Can%20Explain%20Anything,%20It%20Explains%20Nothing.md) * Does the traditional definition and his definition actually contradict each other? * Let's force him to take a position. How is the lookup table generated? * At present, he just takes it for granted. But we know that there are logical & abstract things that exist that cannot be physical instantiated (see Turing's proof) * Say he then argues "A rock is a causal process. A simulation is a causal process. We can let each run and then create a lookup table mapping their successive states to one another. Thus that would yield a lookup table that was created based on two causal processes?" * This is arbitrary. And again, if allows for anything to be an encoding. We must then be incredibly sure to call out what this definition *does not do*. It specifically erases previous meanings of an encoding. We must erase those from our minds to avoid the semantic bait and switch. It does not capture what we generally define encoding to *be* (see DH definition). His definition is vacuous. * The arbitrariness really stems from the fact that the only reason the table is an "encoding" is because some external interpreter *chose it to be*. Whereas, with a proper encoding, the isomorphic structure that it has with the decoding creates a connection that is present regardless of *any external users interpretation*. But a user could have arbitrarily chosen the encoding to be *anything*! * Now we can argue on our terms. His definition of encoding is poor! * DH role of isomorphism. It is not just about mapping individual elements. Must preserve structure. E.g. think about about a 1 to 1 mapping between elements of alphabet and numbers. (see chat "Isomorphism Structure Preservation) * Based on axioms of reason, we should reject the worse theory (HMs) #### What is okay about his claim? What is not? * The problem isn't the idea that a conscious mind could be simulated on some physical system. That is fine and interesting. * The problem is saying that a *rock* could be interpreted as simulating a conscious mind—just because we can find some mapping/decoding from the rock to the simulation * To say the rock _is_ the simulation implies it somehow carries all the internal causal structure, even though we’re actually doing all the work externally. That’s where we’ve erased the meaning of simulation altogether. We’re no longer working with intrinsic rules. We’ve just let in arbitrary mappings from the outside. * And all of this flows directly from that original bad definition of encoding. #### What other trouble does this bring? * The idea that "anything at all can be theoretically viewed as a simulation of any possible world" — i.e. "all interpretations are valid" * See [But why is it a bad explanation?](Interpretation%20of%20Simulations.md#But%20why%20is%20it%20a%20bad%20explanation?) * We could still attack this argument here. It is not seeking explanation and ruins the definition of simulation. * Counter: well why should we accept all physical transformations? Because they are grounded in physical reality, explanation. * Interpretations are explanations * Counter: what about Godel? Well this is actually a great example in our favor! Godel made a *specific, hard-to-vary claim*. He wasn't saying *all logically possible interpretations are valid*. He gave a good explanation of why his was. * Show P-Q system More detail here: * [All interpretations are not valid](Interpretation%20of%20Simulations.md#All%20interpretations%20are%20not%20valid). Anything *cannot* be viewed as a simulation. * Consider Godel's incompleteness theorems. * Distinguishing Meaningless from Meaningful Interpretations: Not all interpretations are meaningful. A meaningless interpretation is one where no isomorphic connection is seen between the theorems of the system and reality * Good explanations: meaningful interpretation should provide a coherent and truthful way of understanding the formal system in relation to something else. If an interpretation is arbitrary and doesn't reveal a systematic correspondence where theorems map to truths, it doesn't offer a useful or "good" explanation of the system's potential significance * Syntax vs meaning (pq system from GEB is great) * Include the p-q system? Meaningless interpretations lack isomorphism: A meaningless interpretation is precisely one in which we fail to see any isomorphic connection between the theorems of the system and reality. If you randomly assign interpretations to symbols without any systematic relationship to a domain, the theorems will likely not correspond to any coherent set of truths. For example, interpreting --p---q---- as "apple horse apple hat apple apple" for the pq-system bears no relation to addition and thus is meaningless * Moravec: * Interpretation is a specific mapping that takes states of the simulation and maps to views meaningful to some observer. But note that in the rock example, the states of the simulation aren't really in the rock (because the intrinsic rules aren't in the rock) * Hofstader * Think about Godel. He showed that a single statement in PM could have multiple meanings * The first is that a statement in PM refers to numbers (standard meaning of PM that Russel would have embraced) * The second is that a statement in PM could refer to PM itself * Godel didn't show that *any* interpretation is valid - he showed that a very specific one was valid and had meaning! * There was no *arbitrariness* to his mapping. His mapping was very concrete and incredibly hard to vary. #### Why did HM walk into these bad definitions? He was not seeking good explanations, he was possibility pumping. The rock was always a bad explanation: 1. It solves no [Problems](Interpretation%20of%20Simulations.md#Problems) 2. [If it explains anything, it explains nothing](Interpretation%20of%20Simulations.md#If%20it%20explains%20anything,%20it%20explains%20nothing) 3. [If anything can be interpreted as a simulation, simulation means nothing](Interpretation%20of%20Simulations.md#If%20anything%20can%20be%20interpreted%20as%20a%20simulation,%20simulation%20means%20nothing) 4. [A Dictionary Is a Bad Explanation](Interpretation%20of%20Simulations.md#A%20Dictionary%20Is%20a%20Bad%20Explanation) 5. Interpreting anything as a simulation doesn't [kick back](Interpretation%20of%20Simulations.md#Kicking%20back) 6. [Interpretation is about *Explanation*](Interpretation%20of%20Simulations.md#Interpretation%20is%20about%20*Explanation*) The interpretation program provides the explanation of the mind. The rock can be discarded! It is easy to vary. Consider the idea of interpreting noise. Noise requires an interpretation to become a simulation. Say we use interpretation program A to do this. The entire goal of A is to take some "simulation" and map it to some external view that an external observer can understand. This inherently is a process that is seeking *explanation* - for explanation is understanding. But, suddenly, once the user has received the output of the interpretation (where the original goal was to build understanding), they don't care about understanding or explanation at all! They don't say "well our best explanation says that the logic really all lives in the interpretation program". They for some reason accept that the simulation was intrinsic to the rocks after all. This feels like an [Arbitrary Boundary](Arbitrary%20Boundaries%20Beyond%20Which%20Human%20Reason%20Cannot%20Access.md)! The minute the user has received the output of the interpretation program, they no longer seek explanation. Why do they *arbitrarily* stop? This may also harken back to HM's argument that interpretation is dispensable - it is not if all of the simulation lives in the interpretation program. The interpretation program by itself seems to provide an explanation of a mind, without the need to map it to any specific object, which is in line with a view that values simpler explanations. Some of my definitions would be: 1. A proper simulation is a program that has clear [laws of physics](laws%20of%20physics.md) (rules, strict internal logic) that lead to evolution of the system (causality)! If the simulated system does not have causality, then the interpretation table will. 2. A set of data (e.g. random noise) that can be interpreted as *anything* (via some arbitrary lookup table), can't be interpreted reasonably as anything useful. Even including the random noise is *arbitrary* - you could just as well simply have the interpretation program! For that is where all internal logic and information processing lives. See more in [Interpretation of Simulations](Interpretation%20of%20Simulations.md) 3. The reason most random sequences aren't simulating anything is that they lack the internal logic or information processing that characterizes a true simulation. To interpret a random sequence as a simulation, you need an external program created by a mind. This program effectively becomes the source of the simulation, imposing meaning onto the random data. The random sequence itself remains just that - random noise. 4. I don't buy his argument. I would argue that a simulation can be said to exist if it follows rules of an internal logic. If it does not (in the case of random noise, for instance), then the entire idea of externally interpreting this random noise means we effectively need a program to do all of the heavy lifting of "external interpretation". At a certain point this "external interpretation" becomes a simulation / program in and of itself! 5. Contradiction: He argues that a simulation’s existence is independent of interpretation, but his example of noise-as-simulation depends entirely on interpretation. The contradiction arises because noise, lacking inherent structure or rules, cannot self-contain a simulation. He also has no consistent criteria for reality (such as [Dr Johnsons Criteria](Dr%20Johnsons%20Criteria.md), [Criterion for Reality](Criterion%20for%20Reality.md)). As a reductionist he is not always seeking the best explanation. ###### Previous Draft Notes: Moravec's Machine Does Not Seek Explanations At root, the the fundamental misconception that Moravec possesses is that he does not seek good explanations. While it make look like he is utilized [Good Reasoning](Good%20Reasoning.md), he is merely [Logical Possibility Pumping](Logical%20Possibility%20Pumping.md). This is a form or rationality without rationale. This is sadly all too common for great minds who have been lead astray by reductionism. Let's look again at some of Moravec's argument, this time through a more critical lens. He wrote: > [!quote] > ...the world we perceive through our senses could [logically] be an elaborate hoax... This is the first clear sign that Moravec is not seeking *explanation*. For of course the entire world around us *could logically* be a hoax—[Solipsism](Solipsism.md) could be true. But that is a terrible explanation! While we cannot logically rule out Solipsism, we can certainly rule it out for being a bad explanation, as I've shown in [Defend Science by Arguing Against Arbitrary Boundaries](Defend%20Science%20by%20Arguing%20Against%20Arbitrary%20Boundaries.md). When solipsism is [taken seriously as a theory](Take%20Theories%20Seriously%20on%20Their%20Own%20Terms.md) it inevitably leads to [contradictions](Contradiction.md) and undermines its own foundational claims. And if one tries to slip out of the contradictions the theory implies, we find that Solipsism is just [Realism](Realism.md) in disguise, weighted down by unnecessary assumptions - worthless baggage, introduced only to be explained away. It most definitely does not satisfy the principle of [Occam's Razor](Occam's%20Razor.md). But one may come to Moravec's defense "but the fact that it is logically possible must mean *something*? While yes it is not a good explanation, it is *possible*. So shouldn't we consider it?" No we should not. Any person who is reasoning according to the [Principles of Rationality](Principles%20of%20Rationality.md) would not take this theory seriously. Of course we can consider it and even try to come up for a good [Argument](Argument.md) and [Explanations](Explanations.md) for why it may be true. But as it stands it is bad explanation that we need lose no sleep over. A related example may make this more clear. Why do we feel that we can rely on the laws of [Logical](Logic.md) [Deduction](Deduction.md)? The laws of logical deduction *could* in fact be false! Any attempt to justify them logically will lead either to [Circularity](Circular%20Reasoning.md) or an [Infinite Regress](Infinite%20Regress.md). [The Laws of Deduction Are Justified Because No Explanation Is Improved By Replacing a Law of Deduction](The%20Laws%20of%20Deduction%20Are%20Justified%20Because%20No%20Explanation%20Is%20Improved%20By%20Replacing%20a%20Law%20of%20Deduction.md). This may not seem to be a very secure foundation for pure [Logic](Logic.md). And it is not perfectly secure! We are [Fallible](Fallibilism.md). The laws of logical deduction are in fact an [Explanation](Explanations.md). They solve [Problems](Problem.md), and they do so better than any rival theory that has been proposed. And so it is with solipsism. We can rule it out simply because it is a bad explanation. Moravec's reductionistic views also have an implicit bias towards favoring lower level theories in the [Hierarchy of Theories](Hierarchy%20of%20Theories.md). Throughout his argument he retreats further and further into the realm of the logically possible, abandoning the philosophical and epistemological—the seeking of good explanations. He presumably does so based on a belief in [Reductionism](Reductionism.md). But the hierarchy does not exist, and it shockingly easy to show this. Why would one belief that mathematics and logic would form the bedrock of the hierarchy? That is a philosophical question. Determining the reliability of a field of knowledge requires grappling with philosophical questions about the nature of truth and justification. What does it mean for a statement to be true? What criteria should we use to determine whether a belief is justified? %%TODO: tie the hierarchy and reductionism back to your claim that he is not seeking explanations %% %%TODO: Should I bring in more from reductionism?%% ### Tools * [Take Theories Seriously on Their Own Terms](Take%20Theories%20Seriously%20on%20Their%20Own%20Terms.md) * [Forced To Take a Position](Forced%20To%20Take%20a%20Position.md) * [Explanations Imply Consequences](Explanations%20Imply%20Consequences.md) * [Argument Is How We Achieve Justification](Argument%20Is%20How%20We%20Achieve%20Justification.md) * [Problems Create a Logic of Reasoning](Problems%20Create%20a%20Logic%20of%20Reasoning.md) * [Word doing a lot of work](Word%20doing%20a%20lot%20of%20work.md) * [Surely Operator](Surely%20Operator.md) * [Vagueman Argument](Vagueman%20Argument.md) * [Contradiction](Contradiction.md) * [Inconsistent](Inconsistent.md) * [Deepity](Deepity.md) (ambiguous) # TODO: Categorize the text below