Downloading Your Consciousness Just Before Death.

[RaftPeople]

@wolfpup

I have no idea why you’re so obsessive about this or what point you’re trying to make, to the extent of repeatedly demanding answers to your challenges. And it seems that every time I make the effort to answer a question, it only raises two more, and more demands for more answers.

I ask repeatedly because you frequently don’t answer the questions that represent significant challenges to what you have posted.

What Fodor explicitly disclaimed was that higher-level cognition was also modular. Whether he ever also claimed that it wasn’t computational I have no idea.

Ok, great progress, you were not aware of Fodor’s position on abductive reasoning.

You seem to like Fodor so I would recommend reading his book “The Mind Doesn’t Work That Way.” He walks through this problem and others and argues that Turing style computing systems can’t do it and he also argues that connectionism doesn’t solve the problem either.

Basically, his opinion is that this global reasoning that human’s do is unexplainable by mechanical systems and as mysterious as consciousness.

From my perspective, the argument is interesting because this type of reasoning is exactly the type of reasoning that I think most of us are picturing when we think of making an AI system. So Fodor thinks the highest value portion of cognition is non-computational.

That is, unlike learning, where according to CTM we acquire knowledge through putative computational processes, or thinking, where we reach conclusions through similar computational processes, innate behaviors aren’t produced by computational processes, but are hardwired. Whether they are themselves computational is irrelevant to my point – they may or may not be; my point was that CTM is incomplete and doesn’t explain everything about the mind.

Thanks, that is a helpful answer, it provides enough detail to understand what you were trying to say.

I think it introduces a very inconsistent view of computation to state that two identical computing systems are not identical due to the way the systems were instantiated.

Maybe there are no examples of this in our brain, but even if there aren’t, it is mathematically possible, so it seems odd to hold that position.

I’ve already said (pretty clearly) that there are many different definitions of computation, and the appropriate one depends on the context of the discussion. The relevance of the Turing model here is that it’s specifically cited as the relevant meaning of “computation” in CTM,…

A calculator IS a Turing machine, you know that right? It’s not a UNIVERSAL Turing machine, but it’s a Turing machine that computes the functions it’s designed to compute by performing syntactic operations on symbols.

Maybe you mean to say that only UTM’s compute but that TM’s do not compute?

I don’t see Fodor or any other CTM supporter claiming that a TM does not compute but a UTM does compute.

[wolfpup]

@RaftPeople

A calculator IS a Turing machine, you know that right?

A rock is Turing machine, too – you know that, right? :rolleyes:

This is nonsensical muddled thinking, like most of the rest of your reply, but it’s the most flagrant and easiest to explain, so at this point, in the interest of no longer wasting my time, it’s the only one I’m going to bother with. The difference between a Turing machine ™ and a universal Turing machine (UTM) is that a TM is analogous to an arbitrary computer running a fixed program, while a UTM is analogous to a RAM-based computer than can run any such program. Obviously a TM can be built to simulate any trivial calculation, such as in a calculator, or none at all, such as a rock, which can be thought of in your terms as a TM with only one state that never changes.

Neither a calculator nor a rock fulfills the central idea of the TM, that of executing instructions that operate on symbols in a sequence of discrete steps that are characterized by state transitions. This is where the power of the UTM comes from, and by the same token, the power of the computational paradigm and the digital computer. The UTM is what defines the concept of Turing completeness. So of course in using the Turing machine paradigm in CTM we are alluding to this central idea, that mental processes execute arbitrary computational processes in the manner of a digital computer.

[RaftPeople]

@wolfpup

A rock is Turing machine, too – you know that, right? :rolleyes:

This is nonsensical muddled thinking, like most of the rest of your reply, but it’s the most flagrant and easiest to explain, so at this point, in the interest of no longer wasting my time, it’s the only one I’m going to bother with. The difference between a Turing machine ™ and a universal Turing machine (UTM) is that a TM is analogous to an arbitrary computer running a fixed program, while a UTM is analogous to a RAM-based computer than can run any such program. Obviously a TM can be built to simulate any trivial calculation, such as in a calculator, or none at all, such as a rock, which can be thought of in your terms as a TM with only one state that never changes.

Neither a calculator nor a rock fulfills the central idea of the TM, that of executing instructions that operate on symbols in a sequence of discrete steps that are characterized by state transitions. This is where the power of the UTM comes from, and by the same token, the power of the computational paradigm and the digital computer. The UTM is what defines the concept of Turing completeness. So of course in using the Turing machine paradigm in CTM we are alluding to this central idea, that mental processes execute arbitrary computational processes in the manner of a digital computer.

It’s difficult to have a debate about computationalism with a person that thinks Alan Turing was wrong about computers, computation and Turing Machines.

The question I’ve been wondering throughout this thread is whether you have self-awareness. Do you ever read up on these topics like Turing Machines and reflect back on your position in the posts and identify the inconsistencies?

Do you ever think to yourself “hmmm, maybe I don’t actually understand what I think I do”?

[wolfpup]

@RaftPeople

Do you ever think to yourself “hmmm, maybe I don’t actually understand what I think I do”?

… says the guy who doesn’t seem to understand the difference between a calculator and a stored-program computer. I’m done here.

[RaftPeople]

@wolfpup

… says the guy who doesn’t seem to understand the difference between a calculator and a stored-program computer. I’m done here.

Let’s review the evidence.

My post:

A calculator IS a Turing machine, you know that right? It’s not a UNIVERSAL Turing machine, but it’s a Turing machine that computes the functions it’s designed to compute by performing syntactic operations on symbols.

And you’re conclusion from reading those two sentences is that I don’t seem to understand the difference between a calculator and a UTM?

In summary:
Raftpeople says: a calculator is a TM but it’s not a UTM
wolfpup responds: you don’t seem to understand that a calculator isn’t a UTM

[begbert2]

I’m not sure how much value there is left to squeeze out of this thread, but I do feel it’s worth noting that by my understanding of the definition of a Turing machine, a rock ain’t one. Among other things, a rock lacks the infinite data tape that a proper Turing machine includes as part of its makeup.

There is a difference between “something that implements one or more functions” and “a Turing machine”. The latter is a specific thing.

For the record a calculator is also not a Turing machine. It’s actually impossible to make a real Turing machine in the physical world (infinite tapes being hard to come by), and even a finite approximation of a Turing machine requires it to be able to accept a program composed of user-defined commands as an input. (The initial state of its data tape being the other input.) So a standard ten-key calculator is not a Turing machine. A programmable calculator may include a Turing-equivalent processor, though.

[RaftPeople]

@begbert2

I’m not sure how much value there is left to squeeze out of this thread, but I do feel it’s worth noting that by my understanding of the definition of a Turing machine, a rock ain’t one. Among other things, a rock lacks the infinite data tape that a proper Turing machine includes as part of its makeup.

There is a difference between “something that implements one or more functions” and “a Turing machine”. The latter is a specific thing.

For the record a calculator is also not a Turing machine. It’s actually impossible to make a real Turing machine in the physical world (infinite tapes being hard to come by), and even a finite approximation of a Turing machine requires it to be able to accept a program composed of user-defined commands as an input. (The initial state of its data tape being the other input.) So a standard ten-key calculator is not a Turing machine. A programmable calculator may include a Turing-equivalent processor, though.

A turing machine is just a model to reason about computation. It’s just a theoretical machine that performs syntactic operations on symbols.

A universal turing machine has the capability to simulate all other turing machines, which means it can compute every computable function.

A turing machine that isn’t a universal turing machine can’t compute all computable functions, it can only compute the functions it was designed to compute (like a calculator).

Summary:
There is a distinction between a turing machine and a universal turing machine.

[RaftPeople]

If you read this page you will see some very simple examples of Turing machines that Turing created. Simple single purpose machines, not universal machines.

https://en.wikipedia.org/wiki/Turing_machine_examples#Turing’s_very_first_example

[wolfpup]

@begbert2

I’m not sure how much value there is left to squeeze out of this thread, but I do feel it’s worth noting that by my understanding of the definition of a Turing machine, a rock ain’t one. Among other things, a rock lacks the infinite data tape that a proper Turing machine includes as part of its makeup.

There is a difference between “something that implements one or more functions” and “a Turing machine”. The latter is a specific thing.

For the record a calculator is also not a Turing machine. It’s actually impossible to make a real Turing machine in the physical world (infinite tapes being hard to come by), and even a finite approximation of a Turing machine requires it to be able to accept a program composed of user-defined commands as an input. (The initial state of its data tape being the other input.) So a standard ten-key calculator is not a Turing machine. A programmable calculator may include a Turing-equivalent processor, though.

Since any computation performed by a calculator can also be performed by a Turing machine, one can infer that the calculator – or some even more trivially simple calculation device – is equivalent to that particular Turing machine. Hence my intentionally silly example that the same applies to a rock, which could be described as a Turing machine with only one state, the halt state (though a purist would argue that a Turing machine technically must have at least two states). Of course neither a rock nor a calculator was what Turing was defining with his general abstraction, the power of which is best illustrated by the universal Turing machine, which is nothing more than a Turing machine that interprets both the action table and the I/O tape of any other (fixed-function) Turing machine. Only the UTM is Turing-complete, and this is the model for the instruction sets and programming languages of digital computers, and the processes of cognition according to CTM; this is what we mean by “computation” in those contexts. That’s why Raftpeople introducing a simple electronic calculator into this discussion is just stupid.

But I’m sure Raftpeople will be along in a moment to explain to you that you apparently believe that “Alan Turing was wrong about computers, computation and Turing Machines” and you should do more reading because you don’t understand anything. Trust me, this argument is now a total waste of time.

[RaftPeople]

@wolfpup

…nor a calculator was what Turing was defining with his general abstraction,…

This page includes Turing machines created by Turing.
https://en.wikipedia.org/wiki/Turing_machine_examples#Turing’s_very_first_example

After reading that page, would you conclude that Turing thought that only UTM’s are Turing machines?

…and the processes of cognition according to CTM; this is what we mean by “computation” in those contexts.

So, you’re saying that CTM only considers functions computed by a UTM to be a computation. If that same function is computed by a TM, then it’s not considered a computation. Correct?

Because there is a distinction between that position, and the position that the brain computes some functions with a UTM even if other functions are computed with TM’s.

[Half_Man_Half_Wit]

@wolfpup

Of course neither a rock nor a calculator was what Turing was defining with his general abstraction, the power of which is best illustrated by the universal Turing machine, which is nothing more than a Turing machine that interprets both the action table and the I/O tape of any other (fixed-function) Turing machine. Only the UTM is Turing-complete, and this is the model for the instruction sets and programming languages of digital computers, and the processes of cognition according to CTM; this is what we mean by “computation” in those contexts.

This is too restrictive, though. First of all, no real-world computer is a UTM; every computer that can actually physically be built is equivalent to a finite state machine, and has a solvable halting problem by simply iterating through all possible states. Anchoring computation to that notion just means nothing ever computes.

But you might hold that it’s enough to have an in-principle Turing complete instruction set—a device that could perform arbitrary computations, provided we keep augmenting it with all the storage it might need during its operation. Even here, though, I think that’s too strict. For consider two systems, A and B, where A is some special-purpose computer implementing some function, and B is a general-purpose computer set to simulate A. A definition of computation as above would entail that the latter computes, while the former doesn’t—even though they’re performing a functionally identical task.

Suppose now you fit B with a device that blows it up once it strays from its task of simulating A. Now, you’ve effectively robbed B of its universal computing capacity—it can only compute the function A computes; if anything else is computed, it’s blown up. Is the device now merely a special-purpose computer itself? That would mean that the addition of completely inert parts could fundamentally change the computational character of a system—since as long as B sticks to simulating A, nothing happens, but still, as it now can only execute one task, simulating A, it wouldn’t count as a general-purpose computer anymore, and thus, if only such systems computed, would no longer compute.

But then, no characterization of a system would ever be enough to decide whether it actually computes, as there might always be contingencies that rob it of its computational capacities.

[RaftPeople]

An additional point about UTMS and the brain: we only need to find one function of the computable functions that the brain can’t compute to show that the brain can’t be a UTM. The simplest way to do this is choose one of the functions that requires more working memory and state than a human brain can keep track of.

Once we show that the brain isn’t a UTM, then based on **wolfpups ** position, no part of the brain is computational because it’s all computed by a limited Turing machine.

The way out of this is to not try to claim that only UTMS compute and that CTM assumes a UTM.

[wolfpup]

@Half_Man_Half_Wit

This is too restrictive, though. First of all, no real-world computer is a UTM; every computer that can actually physically be built is equivalent to a finite state machine, and has a solvable halting problem by simply iterating through all possible states. Anchoring computation to that notion just means nothing ever computes.

Well, duh! That is, of course, a given, and always assumed. There are very few infinities in the real world (with the exception of the car line made by Nissan, and even that is spelled differently!). Even my wondrous new laptop doesn’t have infinite capacity!

@Half_Man_Half_Wit

But you might hold that it’s enough to have an in-principle Turing complete instruction set—a device that could perform arbitrary computations, provided we keep augmenting it with all the storage it might need during its operation.

Yes. And that’s a non-trivial point.

@Half_Man_Half_Wit

Even here, though, I think that’s too strict. For consider two systems, A and B, where A is some special-purpose computer implementing some function, and B is a general-purpose computer set to simulate A. A definition of computation as above would entail that the latter computes, while the former doesn’t—even though they’re performing a functionally identical task.

You’ll have to be much more specific about what you mean by “special-purpose”. A graphics card is certainly special-purpose, yet they’ve been adapted to other uses to exploit their high-performance capabilities, most notoriously for Bitcoin mining. I would guess – though I don’t know for sure – that I could adapt a typical car’s PCM to play chess, if I added a bit of memory and some new code. Indeed, the idea of an Ethernet-like communications bus is now being pushed for new cars, and modern transport aircraft are already essentially a network of digital computers.

“Special-purpose” typically really means “instruction set optimized for a particular application”, but the universality is still there. Generally such devices are computational in the important sense of being finite-state machines executing both arithmetical and non-arithmetical syntactical operations on stored symbols directed by a stored program. As a kid I used to write assembly-language programs for the venerable PDP-8, which had a three-bit opcode and hence nominally just 8 instructions, so I am more than intimately familiar with the concept of “there’s no instruction for that” but nevertheless that the concept of Turing-complete meant that you could always create a subroutine to implement any imaginable operation. Always.

[Half_Man_Half_Wit]

@wolfpup

“Special-purpose” typically really means “instruction set optimized for a particular application”, but the universality is still there.

No. Special purpose generally means a computer that’s designed to perform a specific function, i. e. explicitly not a general-purpose computer.

[wolfpup]

@Half_Man_Half_Wit

No. Special purpose generally means a computer that’s designed to perform a specific function, i. e. explicitly not a general-purpose computer.

That’s a non-answer. What does that even mean? You speak of some arbitrary “special purpose computer” without any definition of what it is. Is its instruction set Turing complete or is it not?

And I have to re-emphasize my previous response to “But you might hold that it’s enough to have an in-principle Turing complete instruction set”. That is, indeed, as I said, a non-trivial point, and that’s very significant. The instruction set (ignoring addressing limitations) defines the machine architecture and is thus the very definition of its characterization of being Turing complete, whereas it’s only the physical machine architecture itself that has physical limitations.

[Half_Man_Half_Wit]

@wolfpup

That’s a non-answer. What does that even mean? You speak of some arbitrary “special purpose computer” without any definition of what it is. Is its instruction set Turing complete or is it not?

Special-purpose = doesn’t compute everything = non-Turing complete
General-purpose = computes arbitrary functions = Turing-complete

I don’t think I can make it any more explicit than that.

And I have to re-emphasize my previous response to “But you might hold that it’s enough to have an in-principle Turing complete instruction set”. That is, indeed, as I said, a non-trivial point, and that’s very significant. The instruction set (ignoring addressing limitations) defines the machine architecture and is thus the very definition of its characterization of being Turing complete, whereas it’s only the physical machine architecture itself that has physical limitations.

The idea that the instruction set should be Turing complete for something to be called a ‘computation’ or a ‘proper computer’ is, at least, quite odd. SQL’s instruction set isn’t Turing complete (without CTE), but one would typically consider a computer performing SQL queries to compute. Same goes for HTML.

And of course the idea incurs the counterfactual explosion-problem I’ve described above. Think about something like Babbage’s Difference Engine. If it’s the case that only universal devices compute, it doesn’t compute, since it’s not universal. But one might envision an extension of it, the Difference Engine++, such that it becomes able to perform universal computations. Such an extension might just come in the form of additional gears and wheels, that are only used on computations that the bare Difference Engine could not perform.

Nevertheless, since the Difference Engine++ is now computationally universal, it’s either the case that it now computes in the proper sense—even if it only performs the very same operations it did when it was merely the Difference Engine (that is, if the ++ part never is used). Or, it only computes when the additional machinery is used—but then, you have the curious issue that only some operations of a universal machine are computations, and in general, you won’t be able to tell which ones.

And of course, it’s just in conflict with how the term ‘computation’ is generally used. Finite state machines, for example, are generally considered a model of computation, but trivially can never be computationally universal.

[wolfpup]

@Half_Man_Half_Wit

The idea that the instruction set should be Turing complete for something to be called a ‘computation’ or a ‘proper computer’ is, at least, quite odd. SQL’s instruction set isn’t Turing complete (without CTE), but one would typically consider a computer performing SQL queries to compute. Same goes for HTML.

Of course one would consider it computation, but your logic here is bewildering. SQL and HTML are layered on – and rely on – Turing-complete computational infrastructures without which they would be useless. There’s no such thing as an “HTML machine”, for example; there are web servers and browsers that are all written in Turing-complete programming languages that communicate via a standardized markup language. The execution of SQL statements involves no end of computational primitives like test-and-branch instructions.

I have yet to find an example of “computation” in the sense in which it’s meant in CTM (as set out in the Stanford Encyclopedia of Philosophy, for example) that is different from what I just mentioned upthread, that in this context it’s defined “in the important sense of being finite-state machines executing both arithmetical and non-arithmetical syntactical operations on stored symbols directed by a stored program”.

[begbert2]

@RaftPeople

A turing machine is just a model to reason about computation. It’s just a theoretical machine that performs syntactic operations on symbols.

A universal turing machine has the capability to simulate all other turing machines, which means it can compute every computable function.

A turing machine that isn’t a universal turing machine can’t compute all computable functions, it can only compute the functions it was designed to compute (like a calculator).

Summary:
There is a distinction between a turing machine and a universal turing machine.

Ah, yes, you’re right. I had a gap in my memory. Universal Turing machines have two inputs (the program and the tape), and concrete Turing machines have one input (the tape). Thank you for the correction.

Of course this also means that a rock isn’t a Turing machine - unless you stretch the concepts of “data tape” and “state machine” so far as to include all arrangements of matter and all physical interactions (respectively), which would probably be problematic or persons claiming that you can have physical objects that aren’t doing computations.

I now return you to your regularly scheduled discussion.

[RaftPeople]

@wolfpup

“in the important sense of being finite-state machines executing both arithmetical and non-arithmetical syntactical operations on stored symbols directed by a stored program”.

This is how a calculator operates, so why doesn’t a calculator compute?

Is it because the program that is loaded into a calculator and executed by it’s processor limits the set of functions that can be computed?

If I begin executing a program on my PC that takes complete control of the PC so the OS is no longer part of the picture, and this program is just a calculator program, did I just render my computer as a system that no longer performs computations?

[RaftPeople]

@begbert2

Ah, yes, you’re right. I had a gap in my memory. Universal Turing machines have two inputs (the program and the tape), and concrete Turing machines have one input (the tape). Thank you for the correction.

Of course this also means that a rock isn’t a Turing machine - unless you stretch the concepts of “data tape” and “state machine” so far as to include all arrangements of matter and all physical interactions (respectively), which would probably be problematic or persons claiming that you can have physical objects that aren’t doing computations.

I now return you to your regularly scheduled discussion.

I agree that considering a rock as a computer seems to miss an important element, but like most of this stuff, drawing a clean boundary is tricky.

Either way, I think wolfpup’s position is not a strong one because it eliminates all of the specially built circuits that perform computations. Let’s pretend that the brain has circuits specifically built for performing numeric operations, so it can compute a large number of functions, but it’s still limited in what it can compute. wolfpup discards these as non-computational, but there doesn’t seem to be any value in doing that.