`At 15:42 09/01/04 -0500, Jesse Mazer wrote:`

Bruno Marchal wrote:

I don't think the word "universe" is a basic term. It is a sort or deity for atheist. All my work can be seen as an attempt to mak it more palatable in the comp frame. Tegmark, imo, goes in the right direction, but seems unaware of the difficulties mathematicians discovered when just trying to define the or even a "mathematical universe". Of course tremendous progress has been made (in set theory, in category theory) giving tools to provide some *approximation*, but the big mathematical whole seems really inaccessible. With comp it can be shown (first person) inaccessible, even unnameable ...

"Inaccessible" in what sense? How do you use comp to show this? If this is something you've addressed in a previous post, feel free to just provide a link...

`This is a consequence of Tarski theorem. Do you know it?`

I think I have said this before but I don't find the links (I have to much mentioned

the result by McKinsey and Tarski in Modal logic, so searching the archive

with "tarski" does not help).

Let me explain it briefly. With the platonist assumption being a part

of the comp hyp, we can identify in some way truth and reality (in a very

large sense which does not postulate that reality is necessarily

physical reality). That is "Reality" is identified with the set of all true propositions

in some rich language.

Now Tarski theorem, like Godel's theorem, can be applied to any

"sufficiently rich" theory or to any sound machine. Tarski theorem says that

there is no truth predicate definable in the language of such theories/machines.

Nor is "knowledge" definable for similar reason. So any complete platonist

notion of truth or knowledge cannot be defined in any language used by the machine,

strictly speaking such vast notion of truth is just inaccessible by the machine,

and this despite the fact a machine can build transfinite ladder of better approximation

of it. By way of contrast the notion of "consistency" *is* definable in the language of the

machine, only themachine cannot prove its own consistency (by Godel), but the machine

can express it. Now, with Tarski the machine cannot even express it.

I think I have said this before but I don't find the links (I have to much mentioned

the result by McKinsey and Tarski in Modal logic, so searching the archive

with "tarski" does not help).

Let me explain it briefly. With the platonist assumption being a part

of the comp hyp, we can identify in some way truth and reality (in a very

large sense which does not postulate that reality is necessarily

physical reality). That is "Reality" is identified with the set of all true propositions

in some rich language.

Now Tarski theorem, like Godel's theorem, can be applied to any

"sufficiently rich" theory or to any sound machine. Tarski theorem says that

there is no truth predicate definable in the language of such theories/machines.

Nor is "knowledge" definable for similar reason. So any complete platonist

notion of truth or knowledge cannot be defined in any language used by the machine,

strictly speaking such vast notion of truth is just inaccessible by the machine,

and this despite the fact a machine can build transfinite ladder of better approximation

of it. By way of contrast the notion of "consistency" *is* definable in the language of the

machine, only themachine cannot prove its own consistency (by Godel), but the machine

can express it. Now, with Tarski the machine cannot even express it.

Like Godel's theorem, tarski theorem is a quasi direct consequence of the *diagonalisation lemma":

For any formula A(x), there is a proposition k such that the machine will prove "k <-> A(k)". Note: A(k) is put for the longer A(code of k)

`In case a truth predicate V(x) could be defined in arithmetic or in the machine's`

tong, the machine would be able to define a falsity predicate (as -V(x) ), and

by the diagonalisation lemma, the machine would be able to prove the

"Epimenid sentence" "k <-> -V(k), which is absurd V being a truth predicate.

tong, the machine would be able to define a falsity predicate (as -V(x) ), and

by the diagonalisation lemma, the machine would be able to prove the

"Epimenid sentence" "k <-> -V(k), which is absurd V being a truth predicate.

`Truth, or any "complete description" of reality cannot have a definition, or a name:`

semantical notion like truth or knowledge are undefinable (unnameable).

semantical notion like truth or knowledge are undefinable (unnameable).

Actually we don't really need comp in the sense that these "limitation theorem" applies to much powerful theories or "divine" machine with oracle, ...

`OK?`

`Bruno`