Do you believe it's possible the world is a simulation?

In My Humble Opinion
181

We’ve been through this before. If you apply a limited number of inputs to a circuit, the set of possible functions that could produce these results will be of size > 1, even after logic minimization. What you are doing basically is looking at several truth tables, and seeing which ones have rows 0 1 1. It should be of no surprise to anyone that more than one do.
If however you present me with n **2 or four inputs, I can tell you precisely which function is being computed. And I can even give you a minimal implementation of that function using well known logic minimization techniques. If I don’t require minimal logic, I can give you an infinite number of implementations all implementing the same computation.
Now I thought you were tallking about circuits with memory. Small FSMs can have their state tables derived from various experiments - you learn this in automata theory. Large ones - with thousands or millions of state elements - are not practical to understand, though there are various methods such as partitioning which makes the process simpler. However there are methods to prove that an implementation implements a specific computation, which is done during the verification of designs. I’ve read a bunch of papers on this, but haven’t really done it myself. And testing involves seeing if an instantiation of a design matches a golden computation - so in a sense you are automatically deriving a computation from applying inputs to the circuit which allows you to say the compuation either matches the golden one or mismatches at one or several places.
My point in bringing up the physical aspects of things is to show that we must determine what the computation is under this kind of variation. So, what you seem to be saying is impossible is done all the time.

Let’s try a simpler example. There is a kind of puzzle where you shoot a marble into a black box and see where it comes out, and try to deduce the shape of the object in the box. If you shoot one marble in (which is your example, basically) you can imagine several shapes that are possible based on the result. You refine this by shooting more marbles in. Are you saying that there are really several shapes in the box after one marble, as opposed to one shape and a set of possible but incorrect shapes?
As I said before, when you diagnose a circuit which is basically finding which incorrect function the circuit implements, you have an ambiguity set of possible causes (computations) which you try to narrow down. Even if you cannot, that does not mean that there are two different computations being done.

Sure I can. Anything you can do in software you can do in hardware. I can examine the design, see the multipliers, adders and control, see that there is probably an init line and a start line and some outputs, and figure out it is doing a polynomial expansion that gives you the digits if pi slowly but surely.
Now, if what you are saying that we can’t tell what a black box with no inputs and no outputs is computing - well sure. We don’t know that the inside of a black hole isn’t playing World of Warcraft either. But the simulator writer in the OP can observe or the simulation is pointless. (Practically speaking it will be hard to observe everything, which is one reason I don’t believe in the simulation. But we agree on that.)

A specification is the computation we want the chip to do. If we never know what a chip is computing, we’d never know if it is working properly - and we do. Gravity is not a spec. If the government says that a frank must be at least a foot long, then it can make an 11 inch one illegal. And no, saying you interpret a foot as 10 inches, so your 11 inch frank is a foot 1 inch and legal. Equally valid interpretation, right. Silly I know, but less silly than bringing in gravity.

We don’t have to make any arbitrary choices. For instance, we can throw the same pattern at the chip n times and see if the output changes. If it does, we know the chip is sequential. (Actually we’d throw a pattern and its inverse so if there was a clock it would toggle.) We could throw patterns representing numbers at various input positions and see if any of them looks like addition. We’d make hypotheses about the computation and test to support or disprove them. And if we could look inside the job would be much easier.

182

OK. I’m not sure what you think this buys you, but fine: A, B, and C all measure all possible combination of inputs, then translate them to logical values via their respective interpretations.

A obtains the table:



   I1   |   I2   |   O
-------------------------
    0   |    0   |   0
    0   |    1   |   0
    1   |    0   |   0
    1   |    1   |   1

B obtains the table:



   I1   |   I2   |   O
-------------------------
    1   |    1   |   1
    1   |    0   |   1
    0   |    1   |   1
    0   |    0   |   0

And C obtains the table:



   I1   |   I2   |   O
-------------------------
    0   |    0   |   1
    0   |    1   |   1
    1   |    0   |   1
    1   |    1   |   0

Do you now agree that A has computed the logical AND, B has computed the logical OR, and C has computed the logical NAND? Using the same physical system? With only the interpretation differing?

Do you agree that without an interpretation, what Boolean function of two variables that physical device computes is left entirely open? That if you were just handed the device, there would be no way for you to discover what computation it implements?

If not, then how would you discover the ‘real’ computation? Which interpretation is the right one—A’s, B’s, or C’s?

Yes, if you’re given an interpretation of physical states as logical states in advance. But without such an interpretation, this is not possible—any physical system traversing n states (say, in some predefined time) can be mapped to any execution trace of an arbitrary FSA with n states or less. Let that execution trace be [F[sub]1[/sub], F[sub]2[/sub], …, F[sub]m[/sub]]. Let the evolution of the system be [S[sub]1[/sub], S[sub]2[/sub], …, S[sub]n[/sub]], and let n >= m.

Then, all you need is a map from the state space of the physical system to the state space of the FSA, such that every element of the former maps to exactly one of the latter (but not the other way around—i.e. two (or more) states of the physical system may map to the same state of the FSA). Since m >= n, such a map can always be found. Equipped with this map, you can then read off the evolution of the FSA you want to implement from the sequence of physical states your physical system traverses.

Conversely, given only the sequence [S[sub]1[/sub], S[sub]2[/sub], …, S[sub]n[/sub]], there is no objective way for you to tell which FSA is being implemented, as that depends on an arbitrary choice of map from physical states to FSA states.

I’m not. I’m saying that you can’t say what a physical system is computing, even if you have complete knowledge of the system’s construction, and can observe the states the system traverses to arbitrary precision.

Note, though, that there is no such observation in the case of a brain ‘computing’ a mind.

We know if it works according to the specification. But the specification is an arbitrary choice made in the design process.

The point of the example was to show that you can’t change physical properties by legislation; hence, what you do when you fix specification, an intended interpretation, or whatever, has no bearing on how the physical system works (sure, you construct it according to specification, or to support the intended interpretation; but that specification is not then an inherent property of the system). Hence, examining just that system, you will not discover that, say, an output of low voltage was intended to mean 0.

And who chooses what patterns represent which numbers? Is that just some property of nature, something about that particular pattern of voltages that represents 23, rather than 54? Where does the 23-hood of that pattern of voltages reside?

183

No - what I’ll agree to is that either A or C screwed up in applying the test or doing the measurement. The same thing gave different results for a set of inputs? And you think this really happens? Sorry, if it did that computer you are typing on wouldn’t work so well.
Now, A and B might be calling input values different things, since the results they get are equivalent - and AND is the same as an OR with inputs and output inverted. If they are actually applying the same values, then they are screwed up also.
If person A saw a duck with white feathers and a bill on the water, and person B says he saw a duck with a propeller, fixed wings, and a cockpit, is the answer that anyone can interpret what a duck is as they wish or that person B is on acid?

Interpretation only in the sense that you need to figure out what the output voltage stands for. After that, it is just applying vectors and looking at the results. When we look at 50 megabytes of test data we never consider ourselves to be interpreting anything.

I’d do the measurement myself and fire the two of them who screwed it up. I’m assuming of course this is a combinational circuit without some hidden state or memory which can cause the same patterns to produce different results. If it is such a circuit, then I can run the checking experiments I mentioned previously.

You may need to try again. First, you say m >= n and n >= m - so n = m, and the two machines are equivalent. That I buy - but it maybe isn’t what you mean.
If two states from the machine with n states map into a single state for the machine with m states they are equivalent, and you saying that the first machine has n states is incorrect. And you say “may” - I don’t know what this means in a proof.
Also, assuming that the FSM has an input, an m state FSM cannot be characterized by only m inputs.
The only time this could happen is if the n state machine was really a machine with m states and another with n-m states, and the m state segment has no next state function going to the other one. If the starting state is in the n-m state segment you would see some states not mapping to the m state machine segment. If the starting state were in the m state segment, you effectively have an m state machine equivalent to the other one. I suspect there are some restrictions placed on machines to avoid this, but my class was too long ago to remember.
So, I could build a system of n states where only m get used, and this is indeed the same as the m state system, but I basically have built an m state system. Like I said, your AND gate may be actually an and gate with 2p inverters at each input, which is equivalent to the AND without. This is called redundant logic, and such things are used for things like buffering up a signal, but when you are evaluating boolean functions they go away. Yeah, if you see an AND gate and you can’t see inside the package, you don’t know for sure there is no redundant logic. Big deal.

Yet somehow we do this. Every day. If you accept clearly incorrect observations as correct (person C) then sure you don’t know anything. Don’t do that!

We have plenty of observations. We don’t have enough internal observations for a theory yet, nor a framework. And we certainly don’t have one for a non-computational model. Remember, analog computers are computers also, that the brain is not a traditional computer with a von Neumann architecture doesn’t mean it’s not a computer.

:eek: You wouldn’t last long on a design team. A spec is hardly arbitrary.

As I’ve said over and over, a spec lets you determine if a specific instance of a physical system is the one you thought you were making. A spec is an inherent property of the ideal physical system. It is not one of the real one (alas) and how far away the physical system is from the spec is what we determine after manufacture.

The people who built the system. What you’d do is try to figure out which pattern they chose.
And we’re not guaranteed to figure it out. Old computers had 32 bit, say, data inputs consisting of 32 wires. As we increased speed interference between these wires got to be a problem so today we might have serial high speed I/O with only 1 wire going more than 32 times faster than the parallel wires do. It is also self clocked and asynchronous, and part of the processor basically learns what is going on in the wire through a process I try to stay away from. It would be hard for an observer with no knowledge of the system to figure out what is going on (they have to build the knowledge into testers) but it is not arbitrary and if you follow the rules you get the same answer almost ever time. (And you can measure bit rate error and reject parts with too much of it.)

184

No, it gave the same results—as explicitly detailed above. However, they were interpreted differently—as also explicitly detailed above. A will, using this system, be forever capable of computing the AND of two inputs. B will compute the OR, C the NAND. What do you think is there to computing such a function, other than being able to reliably produce the correct answers?

No. Because, as I’m trying to get across for I don’t even know how many posts now, a physical system has objective facts associated to it—if I call it a duck, you call it an aeroplane, one of us is wrong. But if A computes the AND, C computes the NAND, it’s not the case that either of them is more right about what the device computes than the other.

Where is it fixed what the input voltages stand for?

You’ll reproduce A’s measurements. And B’s. And C’s. Because as detailed above, they measure the exact same thing. So, which two will you fire?

Yes, that was an error. The correct relationship is that n >= m.

It means that it’s allowed, but doesn’t have to happen. Basically, the mapping from the states of the physical system—i.e. the patterns of voltages, e.g., or the momenta and positions of its constituent particles, or what have you—to those of the FSA—whatever logical states you want to consider: strings of 0’s and 1’s, or of any other alphabet that suits you—must be surjective.

We can assume an inputless FSA.

No. We don’t ever do this. We always use computers with knowledge of the intended interpretation—some more trivial (words on the screen are interpreted using our language facilities), some less trivial (voltage patterns are interpreted as logical states according to some pre-defined code, say one that says that 5 volts is logical 1).

What I meant is: there is no observer that interprets a brain—a physical system—as interpreting that particular computation that yields a mind—i.e. no ‘user’ to the brain in the sense there is a user to the simulation (or may be one).

And as I’ve said over and over, this is an answer to a question that’s completely irrelevant to the discussion. We’re trying to figure out if computations are an objective property of a system, or if they require interpretation. That is, if, were you handed a physical system—knowing nothing of specification, builder intention, and so on—you could tell that it’s computing the digits of pi.

To think that you could is of course equal to thinking that you could decipher the true meaning of every coded text, hence, nonsense. The physical states of a system don’t imply the logical states they are mapped to any more than the word ‘dog’ implies four-legged canines; the association between them is entirely arbitrary. ‘Dog’ does not carry dog-ness, and ‘5 volt’ does not carry logical 1-ness.

This is, in fact, the entire point I’m trying to make: given a physical system, we don’t have a way to figure out what it computes. Thank you.

Hence, if there is no way of figuring out what computation some physical device instantiates, there is no way of figuring out what computation a brain instantiates—one that gives rise to a mind, or one calculating the topography of my navel. In fact, interpretations exist that make it either.

Hence, whether a brain instantiates a mind is not a physical characteristic of the brain (under computationalism). Otherwise, like mass, like any objective physical property, tests would exist in order to detect whether it is present.

Hence, under computationalism, there would be no objective fact of the matter regarding whether a given system instantiates a mind, or not. It would depend upon interpretation; but since interpretation is something minds do, we arrive at a vicious circle.

I suspect you will not want to go along with that implication of your words. But, at least answer one thing: if it is true that there is no objective way of telling whether a physical system instantiates a given computation, would you agree with the above line of argument?

185

If a machine can do interpretation, does that make the machine a mind?

In all of this, are you following Roger Penrose and denying that the human brain is a “machine?” (Penrose claimed that human intelligence cannot be emulated by computational processes.)

186

Not necessarily: besides intentionality, which is what gives a mind its interpretative faculties, minds (human minds, at least) also possess (or seem to possess, which some people think is a different thing—I’m not so sure) a phenomenal aspect—a subjective quality of ‘what it’s like to be’ that particular mind. It’s conceivable that there could exist systems that have the former, but not the latter (although people have argued against this position), and whether we would call these systems ‘minds’ would then be a question of terminology.

Note, though, that I’m not arguing against the possibility of machine consciousness—indeed, I think conscious machines are possible, and consider myself a case in point. I’m arguing that such machine consciousness does not derive from any computation that might (or might not) be performed. Because, if the idea that all computation depends on interpretation is right, then a computational system could not do the interpreting itself—as in order for it to do so, it would have to be interpreted in the appropriate way, and so on.

I have probably more disagreements than agreements with Penrose’s stance. He doesn’t think consciousness comes from computation, so obviously that’s one point of agreement; but I don’t think the human mind has any super-Turing capacities (in that it could solve decision problems no computer can solve), and I think the Gödelian argument Penrose uses is deeply flawed. I also don’t think his proposal to modify quantum mechanics will come to much, and I fail to see how it would help explain conscious experience.

187

A universe where different observers see totally different things from the same physical object (absent the kind of processing and perceptual errors you seem to deny) is a universe where science does not work because no experiment is reproducible. In a universe where what you say is true I’ll grant you that we can’t say very much about anything. However we have billions of data points saying this is not the case in our universe.

What a gate produces is based on the flow of electrons (and holes) through wires and transistors based on well understood and verified physical laws. It is not mystical and it is not a matter of interpretation - it is just as physical as a duck or an airplane.
And is the duck a duck? There are species which get reclassified as we learn more, and it is quite possible that a child would call a goose a duck. And as my cite I give you “The Ugly Duckling.”

The data sheet - and the design. Data sheets give very specific information on the range of input voltages which represent a 1 and a 0. And other stuff like response times. If you apply your four patterns for A relatively slowly, you’ll get the output you list. If you apply them at 10 GHz, you’ll get something quite different.

Myself. And check myself into the booby hatch.
I developed and run a database with test results for a number of microprocessors which I can’t tell you because it is proprietary. Each test consists of tens of thousands of patterns. The tester checks that the output of the pattern is as expected. No how many times your scenario has happened? Zero.
Which makes me wonder - you claim that people will interpret the same part differently. How about machines? Do you expect that ATE (Automatic Test Equipment, pardon the jargon) A, B and C acts like persons A, B and C?
Because while testers are a bit different, none are different enough for this.

There are theorems dealing with this. Are you saying these are incorrect because the states are actually subjective? BTW the labeling of the states does not matter. What you call them is subjective, but if you label the states of two identical state machines with say 8 states one starting A - H and the other from H to A and expect that state B is the same for both, you’re just wrong.

Your statement said that we have complete knowledge of the system’s construction. Given that, we absolutely know the computation. This does not mean that you don’t need some help in interpretation - if your browser gives you a 404 error that stands for an entry in an error list, and means nothing without that entry. But the computation that produces the message is independent of you understanding it or not.
My system has so much code in it that sometimes it produces a message I don’t remember even though I wrote the code to produce it. But a quick search turns it up - and I can see the code that produced it, and thus the computation. Not just me - my users see it also.

Sure there is - the mind itself. That is what consciousness is all about. The actions of animals based on the “computation” of their subconscious causes physical action, but never got interpreted except in the sense of causing physical action that caused them to interact with their environment. When you dream and don’t remember the dream you are doing something similar.
As Descartes said, even if you are a brain in a jar with no outside observers you have a mind. A simulated consciousness would thus have a mind even if the simulator writers didn’t see it. But they’d never know.

And I said maybe. It is not an easy job, due to lack of information, but that is the difficulty of recreating the specification, not because there inherently is no specification that is not a function of observation.

You really should read General Semantics.
If I do not know the target language of the coding, I’ll surely not be able to decode it. If I know it is in a finite set of languages, I might well be able to - and some, not all. Again, it is information, not different observation which is critical. Someone who knows the language might succeed where I didn’t because he had more information. No observer who did not know the target language could do it.

Don’t thank me. The output of the computation is the output of the system. If the system produces a string of bits, and we can reverse engineer it to see how the system produces that string for a given set of inputs, we have figured out the computation. That this string represents a string of characters in a coding system we don’t know is irrelevant.
Back in the days of core dumps, you’d often get a bunch of bits that you didn’t understand except maybe a tiny fraction. But the computation producing that core dump could easily be understood if you cared enough to look at it.

We clearly don’t understand all the computations our brain does - the data size is way too big, and there is not enough observability. But we understand some of them. And if we were looking at a simulation of a brain we’d be able to non-invasively look at all of it. And observe the interaction that build up the framework of interpretation in the simulated world. Put yourself in a global simulated culture, you no doubt will have a hard time figuring out what is going on. Rewinding the simulation to where the first simulated consciousnesses say “rock,” “axe,” “sex” and you will.

My new grandson, just a month old, is clearly not conscious yet. He is learning, unconsciously, how to interpret the world around him. By the time his consciousness evolves he will have a good basis for interpretation. If a fetus were conscious I might buy into your hypothesis, but the fact support an interpreting subconscious mind which can be used to understand the feedback once that link is established.

There are some cases where you cannot tell what computation a system is doing, if your definition of computation involves interpretation in the environment of the physical system. You might not be able to figure out the computation since the internals are inaccessible and you have inadequate information. But your statement only holds true if you never can figure out the computation - and I’ve been trying to say that we can and do.
We cannot figure out the colors in a picture if the level of light in a room is below a certain threshold. But that does not mean that the colors of that picture are inherently a matter of judgement or interpretation.

188

Are you even reading my posts? I have been harping on and on about how all of A, B, and C see the same physical states—those described in this post. They merely use a different interpretation—a different mapping of physical states to logical states, as described in this post.

If there were a difference in the physical stats each of these observers are seeing, then the whole scenario would argue against my point. If that hasn’t become clear by now, then I don’t think there’s any point to continuing this.

Precisely. And this is exactly equal to saying that if all you have is the sequence of physical states a system traverses, you won’t be able to ‘decode’ it to any particular computation it performs. The sequence of (physical) states traversed is the code text, and all sequences of logical states of all possible computations are equally available as candidate plain texts.

Indeed, in general, you don’t even know the appropriate physical states, since those depend on how fine-grained a view on the system you take—so that is another layer of interpretation.

189

If N scientists observing something saw N different things, science would not work. And my comment about firing B and C wouldn’t make any sense if they accurately reported different results. The fact is that different people do see the same thing - and if they don’t, you can find an error in observation that fooled them. For instance you can read a scope trace wrong and see a 0 instead of a 1 - which can be corrected.

But in the real world we have information about the system, and the creator of the simulator would have a lot of information. I may be wrong, but you seem to be arguing that it is impossible to determine the computation at any time. If that isn’t what you are saying, then the situation in question is clearly one where we can determine it.
When you look at a system at several levels of granularity (done all the time) you have certain points of commonality so that you can check that each level agrees. And you are well aware of places where you lose information. For instance, a logic simulation can’t handle intermediate voltages which a SPICE simulation can, and a functional simulation can’t handle unknowns that a gate level simulation can. But there is no disagreement about what we are viewing. If you saw a 16 bit bus at the logic level and a 32 bit bus at the functional level you have not uncovered a deep philosophical issue with observation - you have a bug.
Look, a person who is mostly illiterate and who misreads a passage isn’t making an acute critical observation or creative reinterpretation - he is just reading it wrong.

190

Here is a real example which maybe someone still reading this will find amusing. Our chips have embedded temperature sensors, which report on temperatures through a mechanism I’m clueless about, but which gets reported in the test output. And one point the sensors were reading -300 degrees C or so. No this observation did not qualify for a Nobel Prize in Physics - it just meant that the sensors in this early version of the chip were not working so well.
Pity.

191

Honestly, I’m not sure whether to laugh or cry here. They observe exactly the same things. When A, B, and C apply the same voltages to the input, they observe the same voltage at the output. However, they disagree what logical values to assign to which voltages—as they may, since this is mere convention. There isn’t anything about ‘5 volts’ that makes it more appropriate to assign to it a logical 1, than a logical 0.

The relationship between these voltages and the logical values is exactly the same as between strings of characters—words—and the entities they name. There is nothing that makes ‘dog’ refer to dogs apart from mere convention. With changing convention, meaning changes—for instance, if a German hears the question, ‘jeden tag?’, he will take it to mean ‘every day?’; however, if a Pole hears the question ‘jeden tag?’, he will take it to mean ‘one, yes?’. This actually happened to me when I bought bread in Poland: the bakery woman asked if I wanted one, and I took it to be her asking me whether I ate bread every day. In fact, even my answer—nodding—was appropriate for both questions: I wanted one bread, and I do in fact eat bread every day.

So the physical observations—sound waves, light, etc—of that scene offer two differing interpretations: her asking me whether I eat bread every day, and me acknowledging that I do so; and her asking me whether I wanted one loaf of bread, and me confirming this.

So, again: A, B, and C make the exact same observations—as they must, having access to the same physical system. Nevertheless, they utilize this system to perform different computations: which they can, since there is no objectively correct mapping between voltage levels and logical values.

192

I give up. C’s observation if fundamentally at odds with A and B. There is no 0 0 1
row in either A or B. A and B might be observing the same thing with different interpretations of 0 and 1, but not C. Thus, C’s observation is inconsistent with the others where A and B are consistent.
I’m not saying that some C somewhere might not think he observes this - but if he does, he is in error.
You seem to be saying that C can use different interpretations for inputs and outputs. If you say that, why not different interpretations for different inputs? That would make every truth table of n inputs consistent with any other one. I don’t know what universe you are in, but in mine - the one that actually deals with this stuff - it does not happen and it is absurd.
I invite you to show me a real case of such an interpretation that is not fundamentally flawed.

Say A and B compare results. One or the other of them will tell the other the agreed up 0 and 1 values, and the other will correct the truth table and they will then agree. No big deal. If C compares his results with A or B, no such simple correction is possible. I can’t even find a single fault which would explain C.
Now, it is possible that C is correct and both A and B have screwed up. But they can’t all be correctly observing something and they can’t all maintain their observations after a detailed comparison. And through physical examination we can tell if there is an odd number of inversions in the circuit, which would make C correct, or an even number (including 0) which would make A or B correct.
BTW, if your contention is correct, it proves that we do live in a simulated universe - and you just found a bug in the code.
So I understand your assertion - but it never actually happens and therefore does not demonstrate anything about anything. You might as well find the logical implication of pigs flying.

193

Did you miss the word “if?”

ETA: There’s room for some mild philosophic nitpicking, because, of course, no two scientists ever see exactly the same thing. We all have an individual viewpoint that can never be truly transcended. We can only communicate with each other because of a huge degree of isomporphism between people: we all have very similar minds. We can “see the same thing” in broad terms. “Point to the red ball.” We can do that. “Show me the world that you see.” We can’t do that.

194

I don’t seem to be saying that, I have said it, explicitly and repeatedly:

Exactly, and I have said that very thing.

Why? Just by your say-so? Do you just get to declare what interpretations are reasonable by some mysterious power wrested in you? C is interested in computing the NAND of her inputs. With the interpretation I have supplied, this is possible. So in what sense is C not computing the NAND of her inputs? She gets the right answer in every case. Are you saying that one can somehow consistently extract the right answer to a computation from a physical system, and still not have performed the computation?

Certainly, this doesn’t happen in actual chip design. This has, as I’ve been trying to point out, absolutely no bearing on the argument. It’s as irrelevant to it as the history of bridge building in ancient Sumer is to the question of how to get to the other side of the river: even if nobody’s ever dreamed of, say, using a rope-and-pulley system then and there, if you do so, and get to the other side of the river, you have solved the problem set before you.

There’s nothing flawed at all about the interpretation C uses. After all, it accomplishes her goal: compute the NAND of her inputs. If you want to, we could invoke special peripheral modules (and I’m half fearing this will mislead you again, but still): at the input sites, we have two little LEDs that shine red if the voltage is low, and green if it is high, while at the output, we have an LED that shines red if the voltage is high, and green if it is low. Is the mapping facilitated by these output peripherals more to your liking? If so, then what C is doing is likewise perfectly legitimate.

Of course they can. I’ve explicitly written down how.

Again, what happens in actual chip design is a complete red herring. Forget, for the moment, that you’ve ever heard of that; you seem to be so deeply entrenched in the assumptions necessary to make things work in practical cases that you fail to see that these don’t hold in general, and just assume they must be true as something like laws of nature—like ‘voltage levels must be mapped to the same logical values at all points in the circuit’. Sure, for practical chip design, it’s useful to do things that way; but that doesn’t mean it’s the only possible way.

You’re constantly talking about how things happen in chip design, but we’re not designing chips here: we’re trying to understand the relationship between physical systems and computations. For doing so, we can’t restrict ourselves to designed systems, since those systems which we want to apply our conclusions to certainly aren’t designed (unless you’re more religious than you appear). Hence, design questions absolutely can’t play a role, and if we imagine they did, then we’re down the wrong track, as we’re then only able to form conclusions about the (very, very tiny) subset of designed systems among the physical systems.

An alien species, for example, could use inverted value assignments for the outputs for aesthetical reasons—and would be able to perform computations in just the same way we do. Different applications could mean that it would make sense to interpret the same signal in different ways. It happens with natural language, after all: a crane can be a bird, or a machine to lift heavy objects, depending on context. Why should ‘5 volts’ not mean either logical 1 or 0 depending on context? It’s still perfectly comprehensible as a mapping.

The fact of the matter is that C can use the physical system to perform the computation he wants to perform in a perfectly simple, reproducible and reliable way—just as much so as A and B can. If you want to say that one of them is ‘more right’ than the others, then you’re being arbitrary regarding what a computation is.

No. Why do you think so?

195

Because Voyager said, “If X then Y” and you snapped, in some dudgeon, “Not X.” That doesn’t invalidate what he said, which was, in fact, correct: if all scientists saw different results, science could not work.

You didn’t just reject it, you rejected it angrily, and yet what you said did not actually contradict what he said.

196

Sure, but it means it’s inapplicable: he’s attempting a reductio by showing that my position leads to absurd consequences; I had, however, just spent the last I don’t know how many posts explicitly telling him that the premise of his argument does not hold.

197

You just moved the goalposts several hundred yards. If C is writing down a truth table of a NAND, we have no problem at all. But if she decides that the AND gate she is really looking at is a NAND, and incorrectly writes the output to make it so, she is not showing a fundamental problem of interpretation, she is showing she is a dope.

I asked about skepticism a while ago, and you said you weren’t one. I’m not sure what to call your position, but it is close to skepticism, since we can never even be remotely sure of what we are observing. And I assume nothing. I have a hypothesis about the consistency of observation which has been tested millions and billions of times and has never failed.

Chips are just one physical system doing a computation. The brain is another. A simulated intelligent entity would be a third. If you are only worried about non-designed things, then a designed entity within a simulation could be conscious, right?
And I’m not sure how design affects anything. Rocks are not designed, but people can describe them consistently. Since people see more or less the same thing when they look at something, they can come up with a consistent way of describing their observation of it, designed or not. (There are just no specs.) And even if a designer specifies a terminology to describe the results of her design, the users of it might come up with a different one. There might even be several which is okay so long as they can be mapped to one another.

Yeah - that is why I have no big problem with the A/B dichotomy. Some signals by the way are active high or active low - we by convention use an overbar or N suffix for active low signals, but that is for convenience and is not required.
And as soon as we figured out what system these aliens were using, we can map to ours with no problem. But they will never come up with result C.

Wrong. You provably cannot use an AND gate to do a NAND without adding additional logic. As I mentioned, you can implement any boolean function with just NANDs, you cannot with just ANDs.
Remember the “proof” that 1 = 2, which relies on a hidden division by 0? That is basically what you are doing here.

198

No, you have spent these posts saying that if C sees something which she cannot see if she is observing correctly all sort of stuff happens.

Unless your C is really Humpty Dumpty in disguise.

(with a slight modification by Voyager.)

199

Let’s apply Occam’s Razor. Which is more likely;that the world is a simulation, or that it isn’t?

200

If it is it was made by that same company that made that school bus driver simulation collaborating with the writers of Graveyard of the Fireflies .