Re: Symbol Grounding Problem (attn: Vend)

Don Geddis wrote:
Stephen Harris <cyberguard-1048@xxxxxxxxx> wrote on Tue, 08 May 2007:
The original Church-Turing Thesis does not support Computationalism.
http://plato.stanford.edu/entries/church-turing/
This details distortions made by Philosophy of Mind philosophers.

Well, yes, that's strictly true.

But the surprising result of the Church-Turing work was that every model
of (physically realizable) mechanical computation that anyone had (or has)
every come up with, always winds up being equivalent (at best) to the very
simple Turing Machines.


I don't know if you mean this summary,

"In summary: (1) Every effectively calculable function that has been investigated in this respect has turned out to be computable by
Turing machine. (2) All known methods or operations for obtaining new effectively calculable functions from given effectively calculable functions are paralleled by methods for constructing new Turing
machines from given Turing machines. (3) All attempts to give an
exact analysis of the intuitive notion of an effectively calculable function have turned out to be equivalent in the sense that each analysis offered has been proved to pick out the same class of functions, namely those that are computable by Turing machine."

SH: or reasons to believe the Church-Turing thesis which still does
not support Computationalism, or if you mean Modest Physical CT.

"Other authors distinguish between CT properly so called and a separate thesis that relates computability to physical systems. This latter thesis has been called thesis M (by Gandy 1980) and Physical CT (e.g., by Pitowsky 1990). Of those who discuss Physical CT, some believe it’s true (e.g., Gandy 1980, Wolfram 1985), whereas others believe it’s false (e.g., Stannett 2003).
In keeping with the canonical view of CT, here I will adopt the distinction between CT properly so called, which pertains to calculability by effective procedure, and Physical CT, which pertains
to a generalized notion of computability by physical systems.

In section 7.2, I argued that CT covers the notion of calculability by effective procedure. If Physical CT is to be broader than CT properly so called, it needs to cover more than computation by effective procedure, namely, it needs to cover any physical process that can be employed for transforming input strings into output strings. For example, Physical CT covers operations on real-valued quantities (as in analog computers), continuous dynamical processes operating on strings (as in certain kinds of connectionist computing and molecular computing), and quantum processes (as in quantum computing).
Using this generalized notion of computation, we can formulate a broader version of CT that applies to all computing mechanisms:

(Modest Physical CT) A function is computable by a mechanism if and only if it is Turing-computable.

Modest Physical CT is believed by most computability theorists and computer scientists. To assess Modest Physical CT, we need to discuss different categories of computing mechanisms and the functions they can compute." ...

"So there are good reasons to believe Modest Physical CT. All computing mechanisms that have been physically built compute only functions that are Turing-computable and not others. It is important to understand the exact scope of Modest Physical CT. Modest Physical CT does not entail that everything is a computing mechanism. It only says that if something is a computing mechanism, then the functions it computes are Turing-computable."

SH: The last part of the description sounds pretty close to what
you said earlier "surpising result". To finish this off I will quote
what the Bold Physical CT says, which is seemingly less plausible:

"(Bold Physical CT) Any function whose values are generated by a physical system is Turing-computable."

SH: I have the present opinion that this Bold Physical CT means
the same thing as Deutsch's "Church-Turing Principle",
""Every finitely realizable physical system can be perfectly
simulated by a universal model computing machine operating by
finite means."

I want to point out that Deutsch uses the word "simulated" and
doesn't he mean the brain and also a computer with a program
that can pass the Turing Test?

That's at least inspiration for Computationalism, if not actually support for
it. It shows that the bar is set very low for a proposed computational
substrate. Thus if complex computation matters for some process, it is
exceeding easy to construct a physical device which is sufficient to support
it.


Yes, I agree, since I think you mean what I called Modest Physical CT.

That plays in to the whole Computationalism narrative of the evolution of the
brain over biological time.


I saw this mentioned by Dennett to support his idea of the "intentional stance". In particular Dennett is not a intentional realist. I've read
Dennett's use of evolution criticized by Fodor, so I don't know what
the Computationalist narrative entails, but at least in this area, what
evolution means to Computationalism is controversial. Dennett's intentional stance is criticized by other Names. Millikan seems pretty smart to me.

The problem to me is, that there is no standard meaning for what
the tenets of Computationalism are. There are at least 20 major
Philosophy of Mind theorists and I think it is safe to say that 10
of them focus on AI, and you don't find consensus among them. I've
never read an explanation of Computationalism that explains why it
should work and how it should work... the explanations are vague
and inconsistent, like toss in some CT and a good measure of
Functionalism, add some sensors, and then sprinkle in a dollup of
the special secret sauce and there you have it, CyberTuring.

The secret sauce is more technically known as "the right program"
or "the appropriate program". It really does remind me of brewing
magical potions in the witches cauldron with a mythical history
reaching back to Prometheus, seen later as Golems and most recently
as the Frankenstein Monster (FM). The FM has a patina of scientific
authenticity, like the Chinse mechanical devices, but no beef.

You never get told how it can actually happen. I think that if they
actually had the story of how Comp. happens, then the AI community
would not have 50 years of failure to show for their efforts. Just
last year they started Cosy, which was a project of several years
of discovering what was needed to make AI (consciousness/mind)
work. They wouldn't have to do this if there was already some kind
of reasonable plan in place. What shows up to me is philosopher spats.

But you're right: the CTT is not "proof" that Computationalism is correct.

So it is true that many philosophers have misinterpreted the CTT or
conflated it. Because a physical system can be modeled on a computer does
not make the physical system a Turing Machine.

Of course modelling/simulating is not the same as first-class implementation.
There's a significant difference between simulating the weather, vs.
implementing a software calculator (which is _not_ a "simulation" of some
hardware calculator).


Well I'm not sure that I followed that analogy, maybe you mean
simulation versus implementation of a brain/mind on a computer
and computer program mentioned below. If you implement the
weather on the computer program does it get wet yet? Maybe no,
it doesn't come with properties. But maybe you are saying that
if you implement the "right program" then consciousness/mind as
a property instantiates, or just mind if consciousness brings
in to many hard to define complications.

Moreover, I want to make the small point that not every computational system
"is" a Turing Machine. The claim is that the physical computational systems
have no more computational power than a Turing Machine would have. But
Turing Machines are constructed with a specific architecture. Your ordinary
desktop PC "is not" a Turing Machine (since it is built using a different
architecture). But its computations are indeed equivalent to the
computations of some Turing Machine. The PC is no more computationally
powerful than (some) Turing Machine.

So we get to Computationalism. Which is not that the brain can be modelled
(imperfectly) by a computer. Nor that the brain "is" a computer. But rather
that the brain is a signal-processing device, which is performing
computations within the class of computations also performed by Turing
Machines. While not _being_ a Turing Machine itself, the brain is equivalent
to some Turing Machine.


Well, I may have written the brain "is" a computer or quoted it.
I'll take a second to check if Putnam wrote that [yep I found it].

"Computational functionalism was proposed as a solution to the mind-body problem (Putnam 1967b), and became very influential. As a result, many philosophers became convinced that computationalism is a consequence of this popular solution to the mind-body problem.

This specific version of functionalism, from now on called computational functionalism, entails the computationalist thesis that the brain is a computing mechanism.2

Putnam believed that the human brain is a probabilistic TM, which he sometimes called probabilistic automaton.15 He commented on computationalism (about the brain) as early as 1958, in a long paper on reductionism, which he co-authored with Paul Oppenheim before any of his papers on minds and machines. In that paper, Oppenheim and Putnam address the reduction of mental phenomena, such as “learning, intelligence, and perception,” to cellular activity, and specifically to the activity of networks of neurons (Oppenheim and Putnam 1958, pp. 18-19).
Oppenheim and Putnam argue that Turing’s analysis of computation "naturally" leads to hypothesize that the brain is a TM: ...

In 1961, Putnam writes as follows: “I would suggest that there are many considerations which point to the idea that a Turing machine plus random elements is a reasonable model for the human brain” (Putnam 1961)."

In a later paper he points out that, if the human brain is such a device, then any physical realization of the same TM program has the same psychology that humans have:
Putnam:
"[I]f the human brain is a ‘probabilistic automaton,’ then any robot with the same ‘machine table’ [i.e., TM program] will be psychologically isomorphic to a human being. If the human brain is simply a neural net with a certain program, as in the theory of Pitts and McCulloch, then a robot whose ‘brain’ was a similar net, only constructed of flip-flops rather than neurons, would have exactly the same psychology as a human (Putnam 1964, pp. 394-395)."

(Comp.) The functions whose values are generated by brains are
Turing-computable. [Comp. = Computationalism]

Right.

Comp. is claiming that the brain is Turing-computable and that does mean
the same thing as the brain can be approximated by a PC which uses
Turing-computable functions = algorithms.

I think this is slightly misstated. The point is not to "approximate" the
brain by a simulation, in the way that a weather simulation approximates the
real weather. If that were the only possible implementation of a human-level
mind, then Computationalism as a philosophical claim would be false.


Ok, approximate doesn't stretch to include "the right program instantiates a mind".

Computationalism says your previous thing, that the brain is actually doing
computations which are Turing-computable. And thus some regular digital
computer (although probably not your PC) could _implement_ (not "simulate")
the same computations. (Which implies the running software would have a
"mind" as well.)

-- Don

I can agree with what you say. But, I also think that I use "simulate"
because AI authors use "simulate", but I won't research it unless pressed. I try to usually say instantiate.
"[W]e have good reason to believe that the laws of physics are computable, so that we at least ought to be able to simulate human behavior computationally (Chalmers 1996a, p. 313)."

I was going to post this to "What Compuationalism Means" because
the nebulous nature of the Comp. theory and the various presentations
by AI philosophers don't produce a coherent plausible structure.

Regards,
Stephen
.

Relevant Pages

  • Re: Penrose vs the Robot
    ... then is it true that it is STRONGER than a Turing machine? ... by a universal model computing machine operating by finite means." ... "Quantum mechanical measurements on a physical system are ... "Penrose argues that the nondeterminism of physics ...
    (sci.logic)
  • Re: A dilemma
    ... Not knowing your math or computing background I'm not sure what to point in your direction though. ... A hypercomputer working with real numbers could theoretically calculate things that a UTM could only ever approximate whilst taking all the time and space in the universe to do so. ... If the brain is an ARNN, and consciousness is an epiphenomenon of that hardware, then you'll never build one out of Lego, no matter how much you have. ... Computers need only classical physics to operate which is a subset of the physics the rest of the universe uses, including our brains. ...
    (rec.arts.sf.composition)
  • Quantum Computer Algorithms
    ... internal state (where these operations behave as in classical physics). ... time differed from computing device to computing device...) ... Our current best quantum mechanical ... which produce the same result as classical algorithms but ...
    (sci.physics.research)
  • Re: Trying ti install FreeBSD 5.1 on Sun Ultra 10
    ... In the computing world, nobody cares, even if they do. ... > And it seems that physics in the high-schools in the US has pretty ... what can you expect from brain damaged software on brain ... and hardware builders that never seem to learn from history, ...
    (comp.unix.bsd.freebsd.misc)
  • Re: BOOK for Learning AND teaching the Foundation Certificate
    ... I studied some physics at college but ended up doing computing. ... or 802.11 - and this got me interested in radio theory. ...
    (uk.radio.amateur)
Loading