Korea needs an ontology, and so does its politics

To make Korea's ambitious artificial intelligence (AI) plan more effective, experts say it needs to make a greater effort to build ontologies.

If you’re like me, that’s the kind of word that makes you stop reading. I’ve looked it up a bunch of times, but it’s one of those words that just doesn’t sink in, so I had to do it again this week.

Here’s where I came across it:

“If the government’s goal was to drive the nation’s AI transformation, a key policy should have been to require each ministry to develop its own ontology and submit a detailed implementation plan.”

The speaker was Rep. Lee Jun-seok of the minor opposition Reform Party of Korea, addressing a roundtable on AI organized by this newspaper.

Given that ontology means “the philosophy of being,” and given that Lee, as you may know, studied computer science and economics at Harvard University, it’s possible he was making a perspicacious observation to demonstrate his evident erudition. Or perhaps he was taking the piss.

But in the world of AI, ontology has a specific and practical meaning. It refers to a map of knowledge, rather than simply a collection of information.

The significance is that AI can potentially do more than retrieve information. It can identify inconsistencies and relationships across millions of records that would be extremely difficult for us humans to detect — and, in some cases, might prefer not to detect.

For example, suppose you have a database of cars and their components. You can search how many complete vehicles are available for sale, how many engines there are and how many rear bumpers. An ontology defines what each component actually is and how it relates to the others. That means an AI system can potentially tell you, before you have even thought to ask, that a labor dispute at a bumper manufacturer could mean there will be no completed vehicles available for sale by January.

Lee offered an interesting real-world example. His staff obtained data on 1.05 million products registered with the Public Procurement Service and records of 2.8 million public procurement orders, converted the information into an ontology and developed an AI system to analyze it.

The system found indications of possible misuse of taxpayers’ money amounting to 562.8 billion won ($418.4 million).

Lee said his office is now working on converting Korea’s laws, enforcement decrees and local government ordinances into ontologies. It has apparently already identified some problems, including references to fines under legal provisions that no longer exist.

The objective here, of course, is not to get people or to have AI decide who is guilty or innocent, but to improve administration by identifying duplication, disproportionate use of resources and wastes of taxpayers’ money.

Still, speaking of getting people, the ruling Democratic Party of Korea has managed the passage of a National Assembly special counsel bill to investigate allegations of “fabricated indictments” under previous governments. The apparent objective is to revisit cases against President Lee Jae Myung that were suspended after he took office and turn the tables on those who brought them.

I think what they need to do instead is create an ontology.

To make the case for this, the National Assembly might start by looking at the work of government agencies and special prosecutors since the martial law fiasco of December 2024. Feed in the numbers of investigators and special prosecutors, what they have investigated, the outcomes in terms of fines and sentences and the cost to taxpayers and then build an ontology.

They might also add the plans of the Serious Crimes Investigation Agency and the police to establish a “special investigation headquarters on insurrection” to investigate the martial law case. Which, in case you just arrived in Korea, has already been investigated and the villains tried.

With an ontology, they could ask the AI system where there might be an element of overkill and what a more rational approach might look like.

It is important to remember, though, that despite the use of the word ontology, AI is neither conscious nor even intelligent. It therefore has no need for a philosophy of being — particularly when it starts meddling in the divine business of politics.

Michael Breen (mike.breen@insightcomms.com) is the author of "The New Koreans." The views expressed here are his own.