On 03/10/2026 06:45 AM, olcott wrote:My system does nothing like this. Understanding my perspectiveOn 3/10/2026 7:03 AM, Tristan Wibberley wrote:That sort of approach after the "Berkeley school" ofOn 09/03/2026 12:42, olcott wrote:>On 3/9/2026 4:45 AM, Tristan Wibberley wrote:>On 09/03/2026 07:46, Mikko wrote:>On 08/03/2026 15:12, olcott wrote:>>It proven to be very useful.>
Very useful tools can be very harmful if used carelessly. For example
a knife is useful when you cut bread or wood but harful if you happen
to cut your hand. Likewise an AI that can answer questions although
sometimes incorrectly is useful if you can filter out the incorrect
answers but may be harmful if you fail to filter out one incorrect
answer.
>
One should expect to fail in the long term. An LLM, naively
administered
to be more acceptable with more elapse of time, will fool you
eventually
by the criterion.
>
This is impossible when one only accepts answers
that are grounded in key quotes of foundational
peer reviewed papers in the field. When one does
this then the these quotes can be cited as the
basis ignoring everything that the LLM said.
That's naively true but is typically interpreted roughly the same as
"This is impossible when one only accepts answers that are grounded in
key quotes of foundational peer reviewed papers in the field and one is
not fooled wrt. what those quotes are at the time one makes one's
judgement."
>
The former judgement might not be possible at any time.
>
It has been dead obvious to me that the body of knowledge
expressed in language can be fully expressed as relations
between finite strings as a knowledge ontology acyclic
directed graph of knowledge semantic tautology for decades.
Now because of LLMs I have the conventional terms of the art
to explain all of the details of this within the various
aspects of proof theoretic semantics.
>
attempting to eliminate either all constants or all
variables from the model of the theory, while making
requires understanding alternative various formal foundations of
semantics in linguistics. Few people well versed in the philosophical
underpinnings of the foundations of math would have much experience
with this. Very few of these would have any deep understanding of
alternative philosophical foundations.
for a quick sort of arithmetization then for computing,"William T. Parry, Entailment Logics" redefines ¬ ∧ ∨ →
has sort of eliminated itself from being "the body of
the body of knowledge", since you got "material implication"
there so it's broken.
so that that conventional paradoxes do not arise.
For example A → (A ∨ B) is a logical fallacy.
It's fair to make for tableau for calculi the logical,My system takes relevance logic to its maximum extreme
and even expedient or convenient, if it's not a _modal_
logic and a _relevance_ logic, then it's _quasi-modal_,
at best, and calling that complete is false, or wrong.
fully mapping every nuance of every sense meaning of
every natural or formal language term to the complete
set of relations exhaustively defining the complete
semantic meaning of this term.
The key concepts of "monotonicity" and "entailment"--
in what you have there as "see rule 1 + last wins"
or "proof by contradiction", is not "constructivist",
either. I.e., monotonicity and entailment are
violated by quasi-modal ir-relevance logic, which
makes for _abuse_ of language.
It does make a great lie machine where that's stupid,
though, including claims of never being wrong.
It's still wrong though, or, "that ain't right".
Copyright 2026 Olcott<br><br>
My 28 year goal has been to make <br>
"true on the basis of meaning expressed in language"<br>
reliably computable for the entire body of knowledge.<br><br>
This required establishing a new foundation<br>
Les messages affichés proviennent d'usenet.