From: tchow on
In article <vcbhdkxfxc8.fsf(a)beta19.sm.ltu.se>,
Torkel Franzen <torkel(a)sm.luth.se> wrote:
>tchow(a)lsa.umich.edu writes:
>> Granted, "PA is consistent"
>> is very precise, but I don't think it's *sufficiently* precise for
>> us to prove theorems about it as it stands.
> I don't see what you have in mind here. What imprecision makes
>itself felt, and how and when?

On reflection, I see that "precise" is the wrong word. The literal sentence
"PA is consistent" doesn't have all the *properties* that are needed for the
kind of mathematical analysis that we would like to perform on it. If we
want to study whether it is provable, it is convenient for it to have a
certain syntactic form that the sentence, as it stands, does not have. So
we need to translate it into a suitable form.

But anyway, whether you call it a social agreement or an assumption, there
is some kind of important principle in the background here. I think that
it's useful to state it explicitly and succinctly---unless for some reason
that's impossible, which I don't believe (yet).
--
Tim Chow tchow-at-alum-dot-mit-dot-edu
The range of our projectiles---even ... the artillery---however great, will
never exceed four of those miles of which as many thousand separate us from
the center of the earth. ---Galileo, Dialogues Concerning Two New Sciences
From: Torkel Franzen on
tchow(a)lsa.umich.edu writes:

> But anyway, whether you call it a social agreement or an assumption, there
> is some kind of important principle in the background here. I think that
> it's useful to state it explicitly and succinctly---unless for some reason
> that's impossible, which I don't believe (yet).

In the case of actual instances of formalization on which people are
agreed, I think it is sufficient to observe that the informal version
A and the formalized version A* are both equivalent, in a weak theory,
to some direct formalization A**. The equivalence of A and A** is a
matter of agreement on usage, based on the standard informal
explanations of quantifiers and connectives.

From: yaoziyuan on
So your this post was 29 Jan 2005 20:00:28 GMT. Interestingly, I posted
a maybe relevant post to the newsgroup list.linguist on 29 Jan 2005
17:39:12 GMT, Message-ID:
<1107020352.414946.304230(a)f14g2000cwb.googlegroups.com>.

My that post was about a possible "computer-based human translator
testing" program, which exactly attempts to use a formal sentence to
verify if a natural language sentence is semantically equivalent.
Here's how my speculative testing program works:

STEP 1: Randomly generate a formal sentence F from a knowledge base;
STEP 2: Generate a source natural language sentence S based on F;
STEP 3: Let the human trainee to translate S to a target natural
language sentence T;
STEP 4: Validate if the human answer T is ANY OF ALL THE POSSIBLE
CORRECT TRANSLATIONS inferrable from F with enough symbolic transform
rules (or "rule-based transfer rules" in terms of machine translation)
applied. This is exactly "to capture informal sentence T using formal
sentence F".

Later then I generalized this idea to a class of "computer-based
training programs", which may start a new, useful branch in AI
research...

Yao

From: Helene.Boucher on
LordBeotian wrote:
> <Helene.Boucher(a)wanadoo.fr> ha scritto
>
> > > > No, it's the axiom "(x)(Nx => there exists y such that Sxy)",
i.e.
> > > > every natural number has a successor.
> > >
> > > As far as I know it is not an axiom of PA...
> >
> > It is not usually explicitly stated, but it is assumed when one
> > supposes that the sequential operator is total, i.e. that for every
> > natural number n, (n') exists.
>
> Ok, so we would have to change PA in such a way that S is just a
relation
> symbol...
> In the case you suggest PA has the model given by {0,1}.
> Are you suggesting that {0,1} *could be* the whole set of natural
numbers?> If not we can add the axiom "1 has a successor".
> Are you suggesting that {0,1,2} *could be* the whole set of natural
numbers?
> If not we can add the axiom "2 has a successor".
> And so on... unless there is a point when you say "yes, I suggest
that the
> set {0,1,..,890} (for example) *could be* the whole set of natural
numbers".
> If this never happen we have added enough axioms to have a system
that
> believe that any natural number has a successor.

If there is a problem with the successor axiom itself, there is a
problem with you using "and so on..." as you have done.

From: LordBeotian on

<Helene.Boucher(a)wanadoo.fr> ha scritto

> > Ok, so we would have to change PA in such a way that S is just a
> relation
> > symbol...
> > In the case you suggest PA has the model given by {0,1}.
> > Are you suggesting that {0,1} *could be* the whole set of natural
> numbers?> If not we can add the axiom "1 has a successor".
> > Are you suggesting that {0,1,2} *could be* the whole set of natural
> numbers?
> > If not we can add the axiom "2 has a successor".
> > And so on... unless there is a point when you say "yes, I suggest
> that the
> > set {0,1,..,890} (for example) *could be* the whole set of natural
> numbers".
> > If this never happen we have added enough axioms to have a system
> that
> > believe that any natural number has a successor.
>
> If there is a problem with the successor axiom itself, there is a
> problem with you using "and so on..." as you have done.

The only "problem" I see is if there is a point when you say "yes, I suggest
that the set {0,1,..,890} (for example) *could be* the whole set of natural
numbers", is this the case?