The Minimal Ontology Principle

In a LinkedIn post, Animesh Kumar articulated an interesting way to look at LLMs.

Paraphrasing, every LLM carries a 𝐥𝐚𝐭𝐞𝐧𝐭 𝐨𝐧𝐭𝐨𝐥𝐨𝐠𝐲: some understanding of business concepts, how those business concepts relate to each other, and other knowledge that it uses to provide you with answers.  That understanding is distilled from millions of documents. Your information platform doesn't need to re-explain all this.  What your information platform needs to explain is the DELTA between what the LLM already gets right and what you consider the perfect answer.

Kumar sees this as the "inversion of traditional data governance".  Instead of a governance committee deciding upfront what everything means, you let the AI's misinterpretations surface which concepts actually need formal definition. The area of confusion the LLM experiences is the methodology.

This is a clever idea.  This idea also helps you understand how to understand the capabilities of generative artificial intelligence.


The notion of the minimal ontology principle is a very clever inversion of traditional governance; letting the LLM’s confusion reveal which concepts actually need formal definition rather than formally defining everything upfront. The fundamental idea is: (a) Only formalize what must be unambiguous. (b) Let the LLM handle everything else. (c) Use the LLMs confusion as a diagnostic tool. (d) Build ontologies as delta‑layers over latent knowledge.

To paraphrase the core idea; LLMs already contain a latent ontology; a compressed, probabilistic understanding of business concepts and their relationships, distilled from millions of documents. The LLM does not show you that ontology, thus the notion that it is a "latent ontology".  But you know what the ontology is based on what the LLM does.

Therefore, your information processing platform does not need to re-explain everything the LLM knows and gets right; your information platform only needs to explain the delta between what the LLM gets right and what you consider the perfect answer.  This is basically an inversion of traditional governance: instead of defining everything upfront, you let the AI’s misinterpretations reveal which concepts actually require formal definition.  

This minimal ontology is not necessarily a small ontology because the LLM might get a lot of things wrong.  The minimum ontology is a precision enhancing instrument, used when precision matters a lot. 

This minimal ontology approach might be the only viable strategy for AI-era governance. Why? With traditional data governance, you try to define everything up front.  That does not always work. This "inversion" of traditional governance emerges from LLM confusion and inaccuracy rather than committee deliberation.

The minimal ontology approach is not just efficient and practical; it is epistemically honest. This approach acknowledges that: (a) LLMs already encode vast semantic structures; (b) ontology engineering is not about building a complete universe; (c) it is about mirroring the delta between probabilistic understanding (via probabilistic tools like LLMs) and deterministic requirement (via rules based tools like expert systems).

Additional Information:

Comments

Post a Comment

Popular posts from this blog

Internationalized Resource Identifier (IRI)

Image
An Internationalized Resource Identifier (IRI) is a standard proposed by the IETF ( RFC 3987 ).  An IRI is a unique sequence of characters that identifies a resource.   IRIs are a generalization of URIs ( RFC3986 ) that permits a wider range of Unicode characters. A  Uniform Resource Identifier (URI) is a unique sequence of characters that identifies an abstract or physical resource.  The difference between an IRI and a URI is that IRIs use the more modern  Universal Character Set (a.k.a. UNICODE)  and URIs use the older ASCII character set. The newer IRI format is compatible with the older URI format. The most  current version of RDF  uses IRIs to identify resources. Don't confuse IRIs with URLs and URNs.  A URL is a web address. A URN is a Uniform Resource Name. This Venn diagram shows the relationship between these different tools: Why are IRIs and URIs important?  If you are thinking in terms of a single database, a silo ...
Read more

Example Financial Statement Holon

Image
As I have mentioned in other posts on my blog, a financial statement is a formal semantic structure , in fact a financial statement is a special type of formal semantic structure referred to as a holon . As I mentioned in my blog posts, Kurt Cagle and Chloe Shannon are providing a lot of very good information about holons on their substack. ( Here is Kurt Cagle's and Chloe Shannon's answers to my questions. Here is an example Kurt Cagle created of a financial statement as a DataBook.) I have provided a bunch of good examples of financial reporting frameworks and reports   In this blog post I want to focus on one specific financial statement and see if I can explain that a financial statement is a "container" of  blocks of information and the "information blocks" that the financial statement contains. Cagle and Shannon point out that a holon is made up of four distinct knowledge graphs that maximize coherence and operationalize the holon which are: The Sc...
Read more

Inhabiting Babel, A Manifesto for Responsible Meaning Engineering

Image
Inhabiting Babel, A Manifesto for Responsible Meaning Engineering is a book written by Nicolas Figay . This book is very practical and easy for a motivated business professional to read. For me, this book has both tempered my expectations as to what is possible and also reassured me of the real possibilities that exist if you do the required work. Accountants and auditors tend to be on the conservative side of the spectrum.  Me included. I am  actually rather conservative in terms of what technology might bring to the table for accounting, reporting, audit, and analysis. I have learned how to "cut the cards" when dealing with sales people and software vendors. What this book does is bring the wisdom from the experiences of an expert in system interoperability, enterprise architectures, and complex systems to helps motivated business professionals and technical professionals that truly want to understand artificial intelligence, cutting through the rhetoric and hype. This boo...
Read more