John F Sowa wrote: > > Ravi, Michael, Mark, > > Every communication in any notation of any kind is based on some ontology at some level. > > Michael: can you explain

inherent in LLM technology when used as the primary or only version of AI. The most secure AI systems mitigate or avoid these problems by using traditional symbolic AI technology as

with > LLM-powered tooling aimed at improving communication and related workflows. > > Kingsley > On 1/1/26 9:51 PM, John F Sowa wrote: > > Jans and Kingsley, >

originally developed LLM technology as a major improvement for language translation, both natural and artificial. For that purpose, it has been an outstanding success. It's

John F Sowa wrote: se amounts of $$$ have been poured down the AI drain in hopes that something spectacular will happen. This kind of overinvestment has caused periodic Boom & Bust

John F Sowa : >> >>> Alex, >>> >>> Wolfram and others make an important check to avoid those errors. >>> >>> Wolfram translates

John F Sowa (sowa@bestweb.net) kirjoitti: > >> Alex, >> >> What I did was to call the company that used that service. Then I >> explained that the service

of the LLM-based AI technology. I agree with his negative evaluation of AI systems that use only LLMs and large volumes of data from which the LLMs find or generate answers to any requests

John F Sowa : > Alex, > > There are infinitely many finite models. In fact, it's impossible to > observe or compute more than a finite number of instances. That implies

applications of LLM technology, but methods that allow people -- especially students -- to avoid thinking for themselves weaken their ability to learn and think in deeper and more

of an LLM, an approach as used in the latest version of law analysis in the Netherlands. In short: 1. Definitions of the TERMS used in the data definitions, derivation rules and processes

John F Sowa : > Alex, > > I am not saying that the general goal is bad. In fact, it is a very > important and very difficult research issue. > > That is why I keep repeating

John F Sowa Sent: Friday, July 4, 2025 12:05 AM To: ontolog-forum ; CG Subject: [ontolog-forum] News about Apple and Amazon Two items from the newsletter by Anand Sanwal show current

John F Sowa wrote: > Mike, > > You wrote "you tend to have a didactic style that I have noted before > is off-putting." When I'm writing a note to a friendly

John, From "All those rules are versions of deduction." I got that you are simply classifying any real particular rule of knowledge

John F Sowa wrote: > Neven, Alex, Anatoly, > > These systems are very useful and they are getting better. But they are > very, very far from superintelligence. Unfortunately

John F Sowa wrote: > Following is a note I sent to another email list, and I thought it would > also be useful for the Ontolog and CG lists. In fact, the three links > below are to

component is LLM is a fact, but what is there is still unknown. My opinion: they almost certainly use some filters for the results issued by LLM. This began with the fact that LLM began

John F Sowa : >>> >>>> Alex, >>>> >>>> I agree that any ontology implemented on a digital computer cannot be >>>> informal

beyond just LLM processing. Unlike most people who wrote opinions about DeepSeek, Arun had studied the publications by the DeepSeek developers. His talk is one of the very few presentations

John F Sowa wrote: > A hybrid system that combines LLMs with symbolic reasoning provides > the best of both worlds. And it does so with just a tiny fraction of > the amount of Nvidia

approach with LLM-based methods for translating languages of any kind -- natural, logical, diagrammatic, multidimensional, and perceptual (as mapped to and from sensory input

to the LLM-style of processing. They can use the symbolic methods to detect and eliminate LLM errors and hallucinations. John ___________________ Permion Listed in AWS "ICMP

Alex, The data structure of an ADEPT LION instance is a single table in this demo (the fixed schema doesn't require the use of a graph database

every current LLM system that does not use a symbolic component to detect and evaluate the inevitable errors and hallucinations. Our largest clients (some US gov't agencies)

, and LLM-based services. > My impression is that this type of user-friendly environment and glaring > applications are missing in our community. In my opinion, Protégé and

number of LLM/ChatGPT courses that >> are offered. And, I came away with two conclusions from that writer. >> >> First, there are very few good courses, and the best

system combines LLM technology with traditional AI methods to do the evaluation. This is a hybrid technique that our Permion.ai system uses. I don't know anything more that I read

John F Sowa wrote: > Mike and Igor, > > I'm glad that you mentioned design patterns. But instead of an > ontology of design patterns, I would say that design patterns

John and all, The theme of knowledge processing (KP) is so important. We first need to classify a little what kind of KP we use in science, technology