Comments on: All bloggers are liars

By: dsquared

dsquared — Tue, 15 Mar 2005 18:05:00 +0000

But this gets close to a statement like a TM can’t send any email.
I’ll go a bit further; a Turing machine can’t do anything at all. Turing machines are abstract entities.
A physical machine can act in a way that can be interpreted as an instantiation of a Turing machine, but this is very different. In particular, my view is that the passive-voice “can be interpreted” is covering up a big problem here; any syntactic manipulation (still less semantic) carried out by such a physical machine is entirely parastic on the decision to interpret particular physical events as having a formal role, by someone whose intensionality is not “second-hand” in this way.

By: Daryl McCullough

Daryl McCullough — Tue, 15 Mar 2005 14:24:25 +0000

skippy mcgee writes: If I wanted to destroy a society from within, I’d change the educational system so it turned out minds like those above. I’d train them using self-referential systems and solipsist logic until they could barely even remember to flush the toilet behind them when finished.
Do you really have any evidence that training in Godelian self-reference makes a person unable to function? If you are using discussions about Godelian self-reference as evidence, then you are being self-referential.

By: Luc

Luc — Tue, 15 Mar 2005 05:37:12 +0000

Daniel,
You’re right about the fuzzy thing, though a. is not the case and b. is not my intention.
But a dispute where after years and years of debate the different positions just grow further apart, is bound towards somewhat more emotional responses, as reason is unable to prevail.
The easy way out of this dispute is to just stick with weak AI, call a chess program a database with a search algorithm, and move on. But that doesn’t give much satisfaction. Just as I as i can say I have money in my bank account, and I can use it to trade in objects in role playing games, i would call at some point automated systems intelligent. Artificial, not human, but intelligent none the less.

nothing is intensional by virtue of being a Turing machine

But this gets close to a statement like a TM can’t send any email. So what? What is the relevance of it? Do we need to create another model? Or do we regress to the other end of the spectrum with Searle’s opinion that it is linked with biology?
I referred in my previous message to an interview from 2001 so this opinion is his, and fairly recent.
Quote in full (as the website is somewhat unreliable)

G5: If a computer manipulating symbols via a specified algorithm is not understanding, how do you believe humans process information? Do you believe this to be non-computable?
This question can be answered briefly. The brain processes information by biologically very specific, though still largely ill-understood, neuronal processes. The brain is, after all, a machine that produces consciousness, intelligence, intentionality, etc., by machine processes such as neuron firings at synapses. I regard the notion of what is “computable” as, for the most part irrelevant to the operations of the brain. The brain is a physical system like the stomach. You can describe many, perhaps most, brain processes in a precise enough way that you can simulate them on a computer, just as you can describe the stomach in such a way that you can give a computational simulation of the processes in the stomach. But the computability of these processes at the formal level is not what is essential to the corresponding physical processes at the biological level. At the biological level there are actual causal mechanisms that produce consciousness, intentionality, and all the rest of it.
Basically I think the brain is important. We might be able to do what the brain does using some other medium, but we would have to duplicate the specific causal powers of the brain, and not just do a simulation or a modeling of brain processes, you actually have to duplicate the causal powers. An analogy will make this point clear:.You do not have to use feathers in order build a machine that can fly, but you do have to duplicate the bird’s causal power to overcome the force of gravity in the earth’s atmosphere, and a computer simulation of flight is not a flight. The notion of what is computable and what is not computable is a mathematical notion having to do with what sorts of problems you can solve with algorithms, and I do not see any difficulty posed for artificial intelligence by the limits of computability. The problem is not computability, the problem is psychological reality. I realize that some people have made a great deal out of the fact that there are non-computable problems that humans can solve, but this argument seems to me irrelevant to AI because humans may be using algorithms to solve those problems even though the algorithms are not, for example, theorem-proving algorithms.

By: Skippy McGee

Skippy McGee — Tue, 15 Mar 2005 04:15:47 +0000

If I wanted to destroy a society from within, I’d change the educational system so it turned out minds like those above. I’d train them using self-referential systems and solipsist logic until they could barely even remember to flush the toilet behind them when finished. It’s basically taking the human intellectual output jack and plugging it into the input jack to create a totally contained feedback loop which discusses the noise it hears which is created by discussing the noise it hears.
Why kill your enemy when you can reduce him to a vegetable babbling away in a corner picking lint from his own navel?

By: Brock Sides

Brock Sides — Mon, 14 Mar 2005 12:16:59 +0000

For those interested in Goedel’s theorems, Prof. Peter Smith of Cambridge has a partial draft of his book An Introduction to Goedel’s Theorems online at http://www.godelbook.net.
The chapters posted take you all the way through both of Goedel’s incompleteness theorems. The book does require a background in symbolic logic.

By: Brock Sides

Brock Sides — Mon, 14 Mar 2005 12:09:51 +0000

Do we know that there exists is a system which can be proved (presumably in some higher-order system) to be consistent and which includes the usual axioms of arithmetic?
Sure. You can prove the Peano axioms consistent using set theory.

By: dsquared

dsquared — Mon, 14 Mar 2005 09:05:37 +0000

Luc, this looks more like the position Dennett wrongly attributes to Searle than his actual work. Apologies if I am wrong, but experience has led me to generally suspect that when I see the words “fuzzy feelgood notion”, then a) I’m talking to someone whose main exposure to Searle’s argument has been through Dennett’s summaries and b) the argument is going to get really personal and unpleasant pretty soon, because lots of people who take on board Dennett’s arguments also appear to adopt his polemic style.
Just as a set of preliminaries, I would like to make it clear that the phrase “strong AI” is just as fuzzy as anything in Searle, and that whatever makes me feel good, I suspect that the possibility of intensional computers makes you feel just as good. So can we please not use phrases like “fuzzy feelgood notion”, which don’t advance anything and really only serve as a form of intellectual insult.
Searle’s actual point (and to be hoenst, he hasn’t helped matters by changing his mind on the importance of syntax/semantics, or by some of his examples like the one about simulated fires) is one that can be put into terms that don’t look particularly objectionable to anyone who believes that intensionality is important. Viz:
Brian talks about things being “connected up to the world in the right way”. But a Turing machine can’t be connected up to the world at all, in any way, because it’s an abstract entity[1]. An instantiation of a particular Turing machine can be connected up to the world (and thus, potentially, connected up to the world in the right way), but if it is, then it is connected up to the world as a physical object, not as a Turing machine, and a similarly constructed physical object which was not (which we did not choose to view as) a Turing machine would also be “connected up to the world in the right way”. Therefore, while some physical instantiations of Turing machines might be intensional, nothing is intensional by virtue of being a Turing machine.
[1] This point is arguable, but I would personally defend the view that a connection to the world via some predicate of the kind “is an instantiation of” is not really much of an advance from the mind/body problem.

By: Luc

Luc — Mon, 14 Mar 2005 01:07:31 +0000

So I’d argue that Brian’s “connected up to the world in the right way” has to mean “connected up to some part of the world which has ‘real’, non-derived intensionality”.

But this isn’t that relevant once you accept a form of strong AI, i.e. that non-humans can produce intensionality. That is, just as a (group of) person(s) on it’s own can be conscious, so can any other “strong AI” entity.
Thus once you start from the other side, the symbol shifters can produce their own semantics.
But Searle excludes this notion by describing it for example in the following way “At the biological level there are actual causal mechanisms that produce consciousness, intentionality, and all the rest of it.” and then insisting that those “causal powers of the brain” must be duplicated, not simulated.
That is why I’m entirely prejudiced against Searleans. Their interpretations of intelligence, consciousness, intensionality etc. seem always to revolve around the fuzzy feelgood notion that humans aren’t computers. Which is true, but not a very good basis for advancing AI.

By: Jay

Jay — Sun, 13 Mar 2005 22:48:16 +0000

” On the Axiom of Choice, I just don’t like it because I think it’s at the root of a lot of the difficult paradoxes in probability; I’m not happy with the idea of being able to select a set of points on the basis of no particular criterion of selection. In particular, a lot of the “sadistic angel” problems previously posted on CT disappear if you assume the negation of the AoC, as it becomes clear that the angel is offering you a pretend-gamble based on pretend-probabiliies.”
Actually, no. The “sadistic angel” paradoxes, as I recall them, don’t have anything to do with the AoC. For instance, on the one where you are choosing one point out of the unit interval you could replace the unit interval with a countable dense subset and the single point with, well, any finite set, and the paradox would be the same, but now the AoC wouldn’t be necessary to choose your points.

By: Andrew Boucher

Andrew Boucher — Sun, 13 Mar 2005 22:06:45 +0000

“Do we know that there exists … a system which can be proved (presumably in some higher-order system) to be consistent and which includes the usual axioms of arithmetic?”
Yes. Z1 (first-order Peano Arithmetic, which contains the usual axioms of arithmetic) can be proven consistent in ZF set theory, not to mention Z2 (second-order Peano Arithmetic).

By: Daniel

Daniel — Sun, 13 Mar 2005 20:46:32 +0000

Cosma and Brian noticed by attempt to subvert the dominant paradigm by cunningly typing “syntactic” instead of “semantic” …
My understanding is that Searle’s current position (which which I agree, and I don’t agree that the Systems Reply is a “devastating critique”, or that the Lucas/Penrose argument has been refuted by anything Putnam’s published) is that there is no such thing as a syntactic system which is not a semantic system, because syntactic systems are only syntactic in as much as we interpret their output as having content.
In other words, what a computer (or a Chinese room) does is to turn switches on and off, and this can only be called “syntactic manipulation” if there is somebody around to interpret it as such; a symbol isn’t a symbol unless it signifies. So I’d argue that Brian’s “connected up to the world in the right way” has to mean “connected up to some part of the world which has ‘real’, non-derived intensionality”. I do think that there is such a thing as “intrinsic semanticness”, because I’ve got it and so (I presume) has my wife. My son appears to be developing it.
But we Searleans, rather like we Welsh, are a small and unpopular nation, more keen on backbiting against our more powerful neighbours than taking them on in open combat.
On the Axiom of Choice, I just don’t like it because I think it’s at the root of a lot of the difficult paradoxes in probability; I’m not happy with the idea of being able to select a set of points on the basis of no particular criterion of selection. In particular, a lot of the “sadistic angel” problems previously posted on CT disappear if you assume the negation of the AoC, as it becomes clear that the angel is offering you a pretend-gamble based on pretend-probabiliies.
Meanwhile, the particlar ethical conclusions from Special Relativity relate to a weak version of consequentialism; I think that the role of c in special relativity, based on the requirement to preserve the order of causation in all frames of reference, rules out “backward-causal” moral principles like, for example, the view that murders of Kurds in 1989 can be part of the rationale for a war in 2003.

By: Jay

Jay — Sun, 13 Mar 2005 12:46:28 +0000

Quiggan’s metatheorem is just a restatement of an earlier postulate of mine: “There is nothing more boring than non-mathematicians discussing the implications of Godel’s theorem except non-physicists discussing the philosophical ramifications of quantum physics.”

By: John Quiggin

John Quiggin — Sun, 13 Mar 2005 03:16:59 +0000

Can you confirm my interpretation of your last remark, brock?

Do we know that there exists is a system which can be proved (presumably in some higher-order system) to be consistent and which includes the usual axioms of arithmetic?

By: Walt Pohl

Walt Pohl — Sun, 13 Mar 2005 00:54:39 +0000

Anatoly: The significance was that Hilbert hoped that a weaker system could prove the consistency of a stronger system, that something like Peano arithmetic could be used to prove the consistency of real analysis. Goedel’s second incompleteness theorem shows that that’s impossible.

The fact that a system can’t prove its own consistency is used frequently in set theory. One way to prove that a mathematical statement is not a theorem of ZFC (the name for usual set theory) is to prove that the statement can be used to prove the consistency of ZFC.

By: anatoly

anatoly — Sat, 12 Mar 2005 20:20:37 +0000

But you might think, “It would be nice if the axioms could prove themselves consistent.” Godel’s answer: “No, it wouldn’t.”

Think about it: if the axioms proved themselves consistent, what would that tell you? After all, if they are inconsistent, they could prove anything at all, including their consistency! So why would their proof of their own consistency give you any more grounds to trust them?

It’s not the fact that a system can’t prove its own consistency that’s really interesting; it’s the fact that a system that seemingly formalises our intuitive finitistic reasoning about numbers can’t do it. That is what buried the formalit program, or at least has been widely considered to have done so.