- From: Alex Shkotin <alex.shkotin@gmail.com>
- Date: Wed, 1 Oct 2025 20:01:08 +0300
- To: Milton Ponson <rwiciamsd@gmail.com>
- Cc: ontolog-forum@googlegroups.com, public-lod <public-lod@w3.org>, semantic-web@w3.org
- Message-ID: <CAFxxROSURmh4_qTrWMwfFx9kLGB-ctQ0uU7z6gyxoBMoSWJJQA@mail.gmail.com>
Hi Milton, What do you think about representation of our theoretical knowledge as axiomatic theories? Alex ср, 1 окт. 2025 г. в 18:10, Milton Ponson <rwiciamsd@gmail.com>: > As a mathematician I cannot suppress a chuckle here. The problem here is > the implicit discussion about knowledge, knowledge representation and > formal knowledge representation > These are three distinct layers and because we still not have a firm grip > on the first, which is inextricably linked to consciousness, knowledge > representation remains a difficult task to accomplish, and consequently > formal knowledge representation, which we are seeking will remain elusive. > Large language models ignore the first layer and assume we can use token > based systems to create knowledge representation emulation systems that can > capture all formal knowledge representation systems. > If one looks at the groundbreaking paper MIP*=RE, > https://arxiv.org/abs/2001.04383, and what it states about the Connes > embedding conjecture being false, this should ring a bell. > Because we cannot in all cases assume that a finite matrix in a very high > dimensional space can approximate a simulation of an infinite dimensional > space. > Which means that no matter how high we make the dimension and consequently > the number of parameters used, in some cases the simulations will never > even get close to approximate a finite accurate model of infinite space. > Which means generative LLMs are are a mathematical dead end, and will be > the reason why the AI bubble riding on generative LLMs will burst. > > Milton Ponson > Rainbow Warriors Core Foundation > CIAMSD Institute-ICT4D Program > +2977459312 > PO Box 1154, Oranjestad > Aruba, Dutch Caribbean > > On Wed, Oct 1, 2025, 06:53 Alex Shkotin <alex.shkotin@gmail.com> wrote: > >> John, >> >> I agree. Formalization is absolutely crucial, as we're moving toward >> mathematical methods of knowledge processing, where the differences aren't >> very large and mostly lie outside the realm of finite models and algorithms. >> >> But constructing the most accurate formalization is a rather delicate >> matter. And here, the formal language used, while important, is only an >> auxiliary tool. The knowledge being formalized itself must be a >> well-structured theory. And that's quite challenging. >> >> Therefore, it's proposed to store theoretical knowledge, along with its >> various formalizations, in frameworks specifically designed for knowledge >> concentration [1]. Such theoretical repositories with an emphasis on >> formalization exist spontaneously in Isabelle, Coq, and other provers. >> >> Despite the enormous accumulation of theoretical knowledge in science and >> technology, I believe its volume, in a systematic and refined form, would >> be several terabytes. >> >> The key is to create concentrators of such verified and formalized >> theoretical knowledge. >> >> Alex >> >> >> [1] (PDF) Theory framework - knowledge hub message #1 >> <https://www.researchgate.net/publication/374265191_Theory_framework_-_knowledge_hub_message_1> >> рус >> <https://www.researchgate.net/publication/374233866_Karkas_teorii_-_koncentrator_znanij_soobsenie_No1> >> >> "Storing the theory of a particular subject area in one place and >> maintaining it (including formalization) through collective efforts is >> easily possible with the modern development of technology. The >> concentration and verification of knowledge achieved in this case should >> give a powerful ordering of theoretical knowledge, which will facilitate >> their formalization, i.e. mathematical notation, and therefore algorithmic >> processing in many cases, up to the semi-automatic proof of various kinds >> of consequences, for example, theorems. This message describes what the >> framework of the theory is, intended for unified storage and collective >> accumulation of its results." >> >> >> вт, 30 сент. 2025 г. в 23:02, John F Sowa <sowa@bestweb.net>: >> >>> Alex, >>> >>> Wolfram and others make an important check to avoid those errors. >>> >>> Wolfram translates questions or commands in ordinary English to their >>> precise formal notation. Then before they execute the formal version, >>> they translate it back to a precise statement in Controlled English. >>> >>> The CE text looks like English, and it can be read as English. But it >>> has a precise, formally defined translation to and from Wolfram's formal >>> notation. >>> >>> Many systems, including our Permion Inc. systems do that. They either >>> provide an exactly correct answer, or they carry on a dialog to help the >>> human user specify a request that can be processed by exact formal methods. >>> >>> The final answer is exactly correct reply to the formally defined >>> Controlled English. >>> >>> Errors are still possible, but they are the fault of the human user who >>> may not understand the CE reply. That can be corrected by giving the users >>> more options for asking further questions before making a commitment to one >>> particular answer. >>> >>> John >>> >>> >>> ------------------------------ >>> *From*: "Alex Shkotin" <alex.shkotin@gmail.com> >>> >>> Hi Kingsley, >>> >>> A good article about using RDF and user interface functionality. But I >>> believe that any information generated by LLM should be marked "May contain >>> errors." >>> >>> So all those beautiful tables, diagrams, and documents should display >>> this sign prominently. >>> >>> For me, user interface functionality that reflects the power of RDF is >>> more important. >>> >>> Best regards, >>> >>> Alex >>> >>> пн, 29 сент. 2025 г. в 19:48, 'Kingsley Idehen' via ontolog-forum < >>> ontolog-forum@googlegroups.com>: >>> >>> Hi Everyone, >>> >>> It’s been a while! >>> >>> Something important is happening right now, thanks to the emergence of >>> LLMs as the long-awaited generic RDF client (the so-called “killer app”). >>> We all know how Mosaic → Mozilla/Netscape made HTML and HTTP globally >>> usable by end-users and developers alike. Well, the very same thing is >>> finally happening with RDF—albeit some 20+ years later than expected. >>> >>> Here’s a post I recently published on LinkedIn about this critical >>> development: >>> >>> >>> https://www.linkedin.com/pulse/large-language-models-llms-powerful-generic-rdf-clients-idehen-xwhfe >>> >>> -- >>> Regards, >>> >>> Kingsley Idehen >>> Founder & CEO >>> OpenLink Software >>> Home Page: http://www.openlinksw.com >>> Community Support: https://community.openlinksw.com >>> >>> Social Media: >>> LinkedIn: http://www.linkedin.com/in/kidehen >>> Twitter : https://twitter.com/kidehen >>> >>> -- >>> All contributions to this forum are covered by an open-source license. >>> For information about the wiki, the license, and how to subscribe or >>> unsubscribe to the forum, see http://ontologforum.org/info >>> --- >>> You received this message because you are subscribed to the Google >>> Groups "ontolog-forum" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to ontolog-forum+unsubscribe@googlegroups.com. >>> To view this discussion visit >>> https://groups.google.com/d/msgid/ontolog-forum/c6fff330733e409ab403d68c30a52e46%409d7e08195564407192034ca99241e3fa >>> <https://groups.google.com/d/msgid/ontolog-forum/c6fff330733e409ab403d68c30a52e46%409d7e08195564407192034ca99241e3fa?utm_medium=email&utm_source=footer> >>> . >>> >>
Received on Wednesday, 1 October 2025 17:01:26 UTC