- CG - lists.cs.uni-kassel.de

Protege OWL vs. Frames [Was: Technical question on Protege]

by John F Sowa

Chris and Simon, Your comments are consistent with widely used development tools that have been implemented. I agree that frame notation can be quite user-friendly. But it's important to have a formal specification of exactly how it is mapped to and from FOL or some subset or superset. Many people throw up their hands when they see predicate calculus notation, but it can be translated to English-like notations with just nine words: AND, OR, NOT, IF, THEN, EITHER, OR, SOME, EVERY. (Actually, you only need three words -- AND, NOT, SOME -- but the other words make the sentences shorter and easier to read.) However, there are many factors that can make implemented systems more user-friendly. For an overview of the technical issues and examples of various implementations, see https://jfsowa.com/talks/cnl4ss.pdf Interesting point: TQA was a very usable English Query language developed by IBM research. They found that it was much easier to translate English to predicate calculus notation and then to SQL than to translate directly to SQL. Formal notations that are good for computers can be good intermediate notations to and from natural languages In fact, that is an excellent application for LLMs (or more liikely SLMs -- Small Language Models): translate notations and diagrams with good human factors to and from computer systems.. As the cnl4ss slides show, the users loved TQA, but IBM canceled the project because the task of customizing TQA for each application was too difficult for most users and too expensive for IBM. But the current LLM or SLM technology could learn to do the translations very quickly and inexpensively. I wrote cnl4ss long before LLM/SLM were available. But today that technology would be excellent for tailoring any of the systems discussed in that pdf. John ---------------------------------------- From: "Chris Mungall" <cjmungall(a)lbl.gov> Frames actually turn out to be quite useful for meta-modeling of OWL ontologies. I think we took the wrong fork in the path in 2006, as you suggest. A lot of large OWL ontologies turn out to be quite unwieldy and difficult to maintain, leading to a lot of different approaches like templated ontology generation from spreadsheets or design pattern systems. See my keynote from the Ontology Pattern Workshop from 2020: doi.org/10.5281/zenodo.7655184 LinkML provides a language with frame-like semantics for a modern user base (YAML rather than S-expressions; compiling to Pydantic and JSON-Schema), and a more user-friendly way to incorporate IRIs for all elements. We have a framework linkml-owl (https://linkml.io/linkml-owl/) that allows the (historically implicit) metaclasses in an OWL ontology to be modeled in LinkML/frames, with the OWL TBox being "compiled" from LinkML/Frame "instances". This kind of metamodeling is at best very awkward in OWL itself. See the tutorial. On Thu, Aug 15, 2024 at 3:23 AM 'Polovina, Simon (BTE)' via ontolog-forum <ontolog-forum(a)googlegroups.com> wrote: Hi John and all. Protégé should have maintained its Frames version. At https://protege.stanford.edu/conference/2006/submissions/slides/7.2wang_pro…, there is an insightful presentation that compares Frames and OWL side by side. Notably, the leading industry-strength Enterprise Architecture (EA) tool The Essential Project | Enterprise Architecture Tool (enterprise-architecture.org) uses Protégé Frames under the hood, evidenced by its open-source version. OWL did not fit the bill, as Meta-modelling is important (highlighted in the above presentation link). John, you identified these benefits in your sowazach.pdf (jfsowa.com) 1982 paper with John Zachman, the ‘father’ of EA. Hence, your remark about OWL’s limitations in commercial products is well-taken. Simon

7 months, 3 weeks

1
0
0 0

Technical question on Protege

by John F Sowa

Protege is limited to OWL, which is more complex and more limited than first-order logic. But I realize that many uses of a type hierarchy do not require the full power of FOL. My recommendation would be Aristotle's syllogisms for a type hierarchy, supplemented with FOL for a constraint language. This was the original intention for description logic before the decidability gang restricted its expressive power. Unfortunately, the constraint of decidability had three results: (1) it made the language more complex; (2) it seriously limited its expressive power; (3) it made it unusable for a wide range of tools in AI, computer science, and commercial products. There are many reasoning tools that are more expressive, more powerful, and easier to use than Protege. For a brief overview of Aristotle's syllogisms, see slides 25 to 30 of https://jfsowa.com/talks/patolog1.pdf For more detail about Aristotle and modern logics, see all slides of patolog1.pdf and any references cited on any of those slides. John --------------------------------------------------------- From: "Mara Abel" <marabel(a)inf.ufrgs.br> Colleagues We are wondering here if we can use the reasoning of Protege to automatically produce labels for the entities and instances of a domain ontology. Any idea about it? Thank folks!

7 months, 3 weeks

2
1
0 0

LLMs, SLMs, and the 70 years of AI and computer science

by John F Sowa

I received an offline note that said I was being too negative about the power of the LLM-based technology. I never wanted to give that impression. I am enthusiastic about that technology and its potential. But I am also enthusiastic about the achievements of the 70 years of AI and computer science. The most important new developments take advantage of both. I often criticize attempts to use LLMs for applications that they cannot support -- deductive reasoning is a prime example. I am most enthusiastic about the hybrids, but I am critical of attempts to make LLMs do things that they cannot achieve by themselves. Therefore, I frequently comment on the failures of LLM applications that do not make a proper balance. There are many, many examples. Many of the ones I discuss were sent to me by other people -- often in offline notes by people who would prefer not to be mentioned. If anybody thinks that I have not achieved a proper balance in one or more notes, please cite the examples. Some people do so on the lists and others send me offline notes. John

7 months, 3 weeks

1
0
0 0

AI-Generated Code Has A Staggeringly Stupid Flaw

by John F Sowa

Ravi, There is a huge difference between the theoretical issues about what LLMs (or the smaller SLMs) can do as the underlying technology and what any particular software system can do. The limitations of LLM technology (summarized in the note that started this thread) cannot be extended by systems that just add various interfaces to the LLMs. But applications that combine LLMs with other technology (from AI, computer science, or many kinds of applications) may support a wide variety of technology. Examples that we have discussed before include Wolfram's use of LLMs to support an English-like front end to their powerful Mathematica system. That does everything that anyone has done with Mathematica and adds an English-like front end to give users a simpler and more friendly interface. Many other companies are supporting such technology with varying degrees of success. They implement a friendly interface to their previous systems. Our VivoMind system in 2010 included technology that was very powerful and ACCURATE for applications that cannot be done with LLMs even today. See https://jfsowa.com/talks/cogmem.pdf Our new Permion.ai system combines a newer version of what VivoMind could do with LLMs to support the interface. Arun Majumdar and I have discussed these issues in talks that we gave in the past year or so. I believe that is the wave of the future: use LLMs as one component of an Ai system that uses other kinds of technology to implement functionality that LLMs, by themselves, cannot support. Answer to your question: The features you're asking for in the note below would be very easy to implement -- just add an on/off button for features you don't want. That does not require new technology. It just requires somebody to add that button. John ------------------------------------------------------------------- From: "Ravi Sharma" <drravisharma(a)gmail.com> John Are we at a point where 1. We can turn AI off say on phone apps and desktops 2. We can limit the content access to AI for inputs to hopefully focus the results better Regards Thanks.Ravi(Dr. Ravi Sharma, Ph.D. USA)

7 months, 3 weeks

2
1
0 0

Re: [ontolog-forum] Re: AI-Generated Code Has A Staggeringly Stupid Flaw

by John F Sowa

Michael, The examples you cite illustrate the strengths and weaknesses of LLMs. They show why multiple methods of evaluation are necessary. 1. The failures mentioned in paragraph 1 show that writing a program requires somebody or something that can understand a problem statement and generate a sequence of commands (in some detailed notation) to specify a method for solving that problem. LLMs can't do that 2. The second paragraph shows that ChatGPT had a better selection of answers available in May or perhaps an improvement in its ability to find answers. It's possible that very few dental clinicians had ever used ChatGPT for that purpose. You experiment and the work by the dental clinicians in India may have added enough new patterns that dental clinicians worldwide would have benefitted. 3. The third paragraph shows how ChatGPT learns how to do what it does best: translate from one notation to another. Since you did all the problem analysis to generate Python with miscellaneous errors, it learned how to translate your personal dialect of Python to the official Python syntax. That is an excellent example of LLMs at their best. It was learning how to translate, not learning how to understand. 4. I would say that there is a major difference. Wikipedia is not improved by any method of learning (by humans or machines). Instead, some articles are excellent products of collaboration by experts on the subject matter. But other articles were written hastily by people who don't have the expertise or the patience to do a thorough research of the topic. The Wikipedia editors usually mark those articles that require further attention. But there are many articles that fall between the cracks -- nobody knows whether they are accurate or not. John ---------------------------------------- From: "Michael DeBellis" <mdebellissf(a)gmail.com> [Paragraph 1] I agree. I've asked ChatGPT and Copilot for SPARQL queries, not extremely complicated, either things I thought I would attempt rather than going back to the documentation or in some cases to get DBpedia or Wikidata info because I find the way they structure data to be not very intuitive and it takes me forever to figure out how to find things like all the major cities in India (if anyone knows some good documentation on the DBpedia or Wikidata models please drop a note). I think part of the problem is that people see what looks like well formatted code and assume it actually works. None of the SPARQL queries I've ever gotten worked. [2] We did an experiment in February this year with dental clinicians in India where we gave them a bunch of questions and had them use ChatGPT to get answers and they rated the answers very highly even though almost all of them were incomplete, out of date, had minor or major errors. On the other hand, when I ran the same questions through ChatGPT (and in both cases I used 3.5) in May it was radically different. Almost all the answers were spot on. [3] And for coding, I have to say I find the AI support in PyCharm (my Python IDE) to be a great time saver. Most of the time now I never finish typing. The AI can figure out what I'm doing by figuring out patterns and puts the suggested completion in grey and all I do is hit tab. It's also interesting how it learned. My code is fairly atypical Python, because it involves manipulating knowledge graphs and at first I was getting mostly worthless suggestions. But after a few days it figured out the patterns to read and write to the graph and it has been an incredible benefit. I like it for the same reason I always copy and paste names whenever I can rather than typing, it drastically cuts down on typing errors. [4] All this reminds me of the debates people had about Wikipedia. Some people thought it was worthless because you can always find some example of vandalism where there is garbage in an article. And other people think it is the greatest thing on the Internet. The answer is somewhere in the middle. Wikipedia is incredibly useful and also an amazing example of how people can collaborate just to contribute their knowledge, the way people collaborate on that site is so different than most of the Internet, but you should never use it as a primary source. Always check the references. That's the way I feel about Generative AI. Like Wikipedia I think it is a great resource in spite of the fact that some people claim it can do much more than it really can and that it can still be wrong. It's just another tool and if used properly, an incredibly useful one. Michael https://www.michaeldebellis.com/blog On Saturday, July 27, 2024 at 4:41:24 PM UTC-7 John F Sowa wrote: Another of the many reasons why Generative AI requires other methods -- such as the 70 yeas of AI and computer science -- to test, evaluate, and correct anything and everything that it "generates", As the explanation below says, it does not "UNDERSTAND" what it is doing It just finds and reproduces patterns that it finds in its huge volume of data. Giving it more data gives it more patterns to choose from. But it does nothing to help it understand any of them. This method enables it to surpass human abilities on IQ tests, law exams, medical exams, etc. -- for the simple reason that the answers to those exams can be found somewhere on the WWW. In other words, Generative AI does a superb job of CHEATING on exams. But it is hopelessly clueless in solving problems whose solution depends on understanding the structure and the goal of the problem. For similar reasons, the article mentions that self-driving cars fail in complex environments, such as busy streets in city traffic. The number and kinds of situations are far more varied and complex than anything they have been trained on. Carnegie Mellon University is involved in more testing of self-diving cars because Pittsburgh has the most complex and varied patterns. It has more bridges than any other city in the world. It also has three major rivers, many hills and valleys, steep winding roads, complex intersections, tunnels, foot traffic, and combinations of any or all of the above. Drivers who test self-driving cars in Pittsburgh say that they can't go for twenty minutes without having to grab the steering wheel to prevent an accident. (By rhe way, I learned to drive in P:irravurgh. Then I went to MIT and Harvard,, where the Boston patterns are based on 300-year-old cow paths.) John

8 months

1
0
0 0

Logic of Relatives

by Jon Awbrey

Logic of Relatives • https://inquiryintoinquiry.com/2024/08/05/logic-of-relatives-a/ Relations Via Relative Terms — The logic of relatives is the study of relations as represented in symbolic forms known as rhemes, rhemata, or relative terms. Introduction — The logic of relatives, more precisely, the logic of relative terms, is the study of relations as represented in symbolic forms called rhemes, rhemata, or relative terms. The treatment of relations by way of their corresponding relative terms affords a distinctive perspective on the subject, even though all angles of approach must ultimately converge on the same formal subject matter. The consideration of relative terms has its roots in antiquity but it entered a radically new phase of development with the work of Charles Sanders Peirce, beginning with his paper “Description of a Notation for the Logic of Relatives, Resulting from an Amplification of the Conceptions of Boole's Calculus of Logic” (1870). References — • Peirce, C.S., “Description of a Notation for the Logic of Relatives, Resulting from an Amplification of the Conceptions of Boole's Calculus of Logic”, Memoirs of the American Academy of Arts and Sciences 9, 317–378, 1870. Reprinted, Collected Papers CP 3.45–149. Reprinted, Chronological Edition CE 2, 359–429. • https://www.jstor.org/stable/25058006 • https://archive.org/details/jstor-25058006 • https://books.google.com/books?id=fFnWmf5oLaoC Readings — • Aristotle, “The Categories”, Harold P. Cooke (trans.), pp. 1–109 in Aristotle, Vol. 1, Loeb Classical Library, William Heinemann, London, UK, 1938. • Aristotle, “On Interpretation”, Harold P. Cooke (trans.), pp. 111–179 in Aristotle, Vol. 1, Loeb Classical Library, William Heinemann, London, UK, 1938. • Aristotle, “Prior Analytics”, Hugh Tredennick (trans.), pp. 181–531 in Aristotle, Vol. 1, Loeb Classical Library, William Heinemann, London, UK, 1938. • Boole, George, An Investigation of the Laws of Thought on Which are Founded the Mathematical Theories of Logic and Probabilities, Macmillan, 1854. Reprinted with corrections, Dover Publications, New York, NY, 1958. • Peirce, C.S., Collected Papers of Charles Sanders Peirce, Vols. 1–6, Charles Hartshorne and Paul Weiss (eds.), Vols. 7–8, Arthur W. Burks (ed.), Harvard University Press, Cambridge, MA, 1931–1935, 1958. Cited as CP volume.paragraph. • Peirce, C.S., Writings of Charles S. Peirce : A Chronological Edition, Volume 2, 1867–1871, Peirce Edition Project (eds.), Indiana University Press, Bloomington, IN, 1984. Cited as CE 2. Resources — Charles Sanders Peirce • https://mywikibiz.com/Charles_Sanders_Peirce Relation Theory • https://oeis.org/wiki/Relation_theory Survey of Relation Theory • https://inquiryintoinquiry.com/2024/03/23/survey-of-relation-theory-8/ Peirce's 1870 Logic of Relatives • https://oeis.org/wiki/Peirce%27s_1870_Logic_Of_Relatives_%E2%80%A2_Overview Regards, Jon cc: https://www.academia.edu/community/5AEQjj

8 months

1
0
0 0

Relations & Their Relatives

by Jon Awbrey

Relations & Their Relatives • 1 • https://inquiryintoinquiry.com/2024/07/31/relations-their-relatives-1-a/ All, Sign relations are special cases of triadic relations in much the same way binary operations in mathematics are special cases of triadic relations. It amounts to a minor complication that we participate in sign relations whenever we talk or think about anything else but it still makes sense to try and tease the separate issues apart as much as we possibly can. As far as relations in general go, relative terms are often expressed by means of slotted frames like “brother of __”, “divisor of __”, and “sum of __ and __”. Peirce referred to these kinds of incomplete expressions as “rhemes” or “rhemata” and Frege used the adjective “ungesättigt” or “unsaturated” to convey more or less the same idea. Switching the focus to sign relations, it's fair to ask what kinds of objects might be denoted by pieces of code like “brother of __”, “divisor of __”, and “sum of __ and __”. And while we're at it, what is this thing called “denotation”, anyway? Resources — Relation Theory • https://oeis.org/wiki/Relation_theory Triadic Relations • https://oeis.org/wiki/Triadic_relation Sign Relations • https://oeis.org/wiki/Sign_relation Survey of Relation Theory • https://inquiryintoinquiry.com/2024/03/23/survey-of-relation-theory-8/ Peirce's 1870 Logic Of Relatives • https://oeis.org/wiki/Peirce%27s_1870_Logic_Of_Relatives_%E2%80%A2_Overview Regards, Jon cc: https://www.academia.edu/community/Vj80Dj

8 months

1
3
0 0

Relations & Their Relatives • Discussion

by Jon Awbrey

Relations & Their Relatives • Discussion 24 • https://inquiryintoinquiry.com/2024/07/28/relations-their-relatives-discuss… Re: Daniel Everett • June 20, 2024 • https://www.facebook.com/permalink.php?story_fbid=pfbid02oCRz4EYHAtbrJeAzzo… Daniel Everett remarks: ❝Among the several ideas Peirce and Frege came up with was the idea of a predicate before and after it is linked to its arguments. Frege called the unlinked predicate unsaturated. But Peirce built this into a theory of valency. An unsaturated predicate in Frege's system is a generic term, a rheme, in Peirce's system. So in Peirce's theory all languages need generic terms (rhemes) to exist. Additionally, through his reduction thesis (a theorem proved separately by various logicians) Peirce set both the upper and lower bounds on valency which — even to this day — no other theory has done.❞ Dear Daniel, In using words like “predicate” or “relation” some people mean an item of syntax, say, a verbal form with blanks substituted for a number of subject terms, and other people mean a mathematical object, say, a function f from a set X to a set B = {0, 1} or a subset L of a cartesian product X₁ × … × Xₖ. It would be a great service to understanding if we had a way to negotiate the gap between the above two interpretations. To be continued … Resources — Relation Theory • https://oeis.org/wiki/Relation_theory Survey of Relation Theory • https://inquiryintoinquiry.com/2024/03/23/survey-of-relation-theory-8/ Regards, Jon cc: https://www.academia.edu/community/lzAqAe

8 months, 1 week

1
1
0 0

Re: [ontolog-forum] Fwd: AlphaProof, AlphaGeometry, ChatGPT, and why the future of AI is neurosymbolic

by John F Sowa

Alex, The article you quoted by Gary Marcus is consistent with recent publications about Generative AI. See the excerpt below, which I strongly agree with. Many investors realize that Generative AI is useful for supporting NL interfaces to complex computer systems (Wolfram, Kingsley, and the Permion.ai company that I am working with are examples). But by itself, Generative AI is too unreliable for applications that require accuracy and precision. See the longer commentary by Gary M and many of the articles I have cited in recent notes. As for a collaborative project by Ontolog Forum, this is a discussion group, not a development group. If anybody wants to form a separate project to do anything further, I suggest that they start a separate email list for people on that project. I have too many unfinished projects of my own to work on. And I suspect that is also true of many other subscribers. John __________________ An excerpt from the article by Gary Marcus: https://open.substack.com/pub/garymarcus/p/alphaproof-alphageometry-chatgpt… I think that GenAI is vastly overrated and overhyped, but fear that its collapse may well lead to an AI winter of sorts, like what happened in the mid-1980s, when AI “expert systems” rapidly ascended and rapidly fell. That said, I am certain that the impending collapse won’t lead to the absolute death of AI. There is too much at stake. What the collapse of generative AI might lead to, after a quiet period, is a renaissance. Generative AI may well never be as popular as it has been over the last year, but new techniques will come, new techniques that will work better, and that address some of the failings of generative AI. ---------------------------------------- From: "Alex Shkotin" <alex.shkotin(a)gmail.com> "In the final analysis, expecting AI to “solve” AGI without “System II” mechanisms for symbol-manipulation is like expecting bears to solve quantum mechanics." IMHO GM and GDM are on the way to formalize theoretic knowledge of sciences and technologies. Huge work for many mathematicians for many years. Let's do it. Alex

8 months, 1 week

1
0
0 0

AI-Generated Code Has A Staggeringly Stupid Flaw

by John F Sowa

Another of the many reasons why Generative AI requires other methods -- such as the 70 yeas of AI and computer science -- to test, evaluate, and correct anything and everything that it "generates", As the explanation below says, it does not "UNDERSTAND" what it is doing It just finds and reproduces patterns that it finds in its huge volume of data. Giving it more data gives it more patterns to choose from. But it does nothing to help it understand any of them. This method enables it to surpass human abilities on IQ tests, law exams, medical exams, etc. -- for the simple reason that the answers to those exams can be found somewhere on the WWW. In other words, Generative AI does a superb job of CHEATING on exams. But it is hopelessly clueless in solving problems whose solution depends on understanding the structure and the goal of the problem. For similar reasons, the article mentions that self-driving cars fail in complex environments, such as busy streets in city traffic. The number and kinds of situations are far more varied and complex than anything they have been trained on. Carnegie Mellon University is involved in more testing of self-diving cars because Pittsburgh has the most complex and varied patterns. It has more bridges than any other city in the world. It also has three major rivers, many hills and valleys, steep winding roads, complex intersections, tunnels, foot traffic, and combinations of any or all of the above. Drivers who test self-driving cars in Pittsburgh say that they can't go for twenty minutes without having to grab the steering wheel to prevent an accident. (By rhe way, I learned to drive in P:irravurgh. Then I went to MIT and Harvard,, where the Boston patterns are based on 300-year-old cow paths.) John ________________________________________________ AI-Generated Code Has A Staggeringly Stupid Flaw It simply doesn’t work. https://medium.com/predict/ai-generated-code-has-a-staggeringly-stupid-flaw…. . . So, what is the problem with AI-generated code? Well, one of the internet’s favourite developers, Jason Thor Hall of Pirates Software fame, described it best in a recent short. He said, “We have talked to people who’re using AI-generated code, and they are like, hey, it would take me about an hour to produce this code and like 15 minutes to debug. And then they are like, oh, the AI could produce it in like 1 minute, and then it would take me like 3 hours to debug it. And they are like, yeah, but it produced it really fast.” In other words, even though AI can write code way faster than a human programmer, it does such a poor job that making the code useful actually makes it far less efficient than getting a qualified human to just do the job in the first place. . . . Well, AI doesn’t actually understand what it is doing. These generative AI models are basically overly developed predictive text programs. They use statistics based on a stupidly large pool of data to figure out what the next character or word is. As such, No AI actually ‘knows’ how to code. It isn’t cognitively trying to solve the problem, but instead finds an output that matches the statistics of the data it has been trained on. As such, it gets it massively wrong constantly, as the AI isn’t actually trying to solve the problem you think it is. As such, even when the coding problem you are asking the AI to solve is well-represented in its training data, it can still fail to generate a usable solution simply because it doesn’t actually understand the laws and rules of the coding language. This issue gets even worse when you ask it to solve an AI problem it has never seen before, as the statistical models it uses simply can’t be extrapolated out, causing the AI to produce absolute nonsense. This isn’t just a problem with AI-generated code but every AI product, such as self-driving cars. Moreover, this isn’t a problem that can be easily solved. You can’t just shove more training data into these AIs, and we are starting to hit a point of diminishing returns when it comes to AI training (read more here). So, what is the solution? Well, when we treat AI as it actually is, a statistical model, we can have tremendous success. For example, AI structural designs, such as those in the Czinger hypercar, are incredibly efficient and effective. But it falls apart when we treat AI as a replacement for human workers. Despite its name, AI isn’t intelligent, and we shouldn’t treat it as such. [End]

8 months, 1 week

1
0
0 0

2025

2024

2023

2022

2021

CG