The people who are developing AI systems have recognised that any AI system needs to be supported by a structure of knowledge which describes the things in the world - and the relationships between them - that are to be talked about. That is what I mean by an ONTOLOGY in this context.
Ontology is not a new word. It has been used for many centuries in philosophy, to describe arguments about what things can be said to EXIST, and in what sense.
Anselm, a mediaeval philosopher, used an Ontological argument to prove that God exists. He said because we can conceive of something that is greater than everything else, this is what we mean by God and therefore He must exist. The question is then "in what sense?". Do only physical things exist? What about legal entities, or nations? What about "memes"? Do even our private memes "exist"?
If an AI system uses language to communicate with humans, then anything that can be of interest to humans in the situation has to be expressed in words and sentences. So my principle is that if we tan talk about it, then we can say it exists.We then have to consider the ''in what sense?'' issue. A thing may exist as a physical item, but it may just be a concept in one or more people's mind. The next question is, is there consensus between what people understand when the word is used? I mentioned 4 problem words in the PLOVER slide show - Freedom, Culture, Soul and Truth. There's a fair chance that people talking about these things may each have a different understanding of what they mean by these words.
Historically, there have been several advances in bringing some structure unto the collection of things we can talk about. Dictionaries are an early example, although a dictionary is only concerned with words in a language, and the only structure is the alphabetic order of the words. Encyclopedias take this same structure further, providing background information about the contexts in which words are used.
A alternative approach is to consider contexts, rather than alphabetic order, as the primary structure. As a child I had a little book which introduced things within contexts; more recenttly several books of lists have been written by Barbara Ann Kipfer. I used to have two books for learning vocabulary in another language which used pictures within contexts; they were for German and Arabic. The German one was Usborne's First Thousand Words in German. Each large page had a set of pictures describing a thing in one context, with the word under the picture and a web link to hear how to pronounce it underneath each picture. The Arabic one was a set of extra pages in my dictionary. This really only works for physical things though.. A similar approach has been followed for mechanical parts catalogues; Shell had one for North Sea oil operations, where additional information gave what each item was used for, which other parts it linked too, and even what other parts might be substituted if the one needed was not available.
An early attempt to provide a structure based on meaning came with Roget's Thesaurus. You can view The structure here. But the items within the structure are still words in English; so it is most useful for finding synonyms and words describing closely related things.
There have been a number of ontology systems put forward by university researchers. One early and notable example is the Bunge-Wand-Weber (BWW) ontology - see my separate notes about this. See also the KNO.E.SIS project at Wright University, Dayton, Ohio, USA.
Ontologies have also been built by software companies and major computer users. A notable example is the US Department of Defense. See this Wikipedia page for some more examples.
For a good introduction to the use of ontologies in AI, see this article by Seth Earley.
If we are going to be relying more heavily on "off-the-shelf" AI apps, we should be aware that we are putting our future in the hands of "off-the-shelf" ontologies, created by someone else. Their view of the world may be similar to ours, but never quite the same. So there is a danger that relying too heavily on AI may lead us to only follow "accepted wisdom", leaving no room for individual inspiration or "thinking outside the box".
In my final few years as a university lecturer, I got involved in developing ideas for an ontology that would support 'information workers', particularly those suffering from 'information overload'. This led to a project called FROLIO, which I continued with for a few years after my retirement. I have retained my old pages about FROLIO and they can be visited via the first link below.
| FROLIO home page | PLOVER index page | Back to Roger's home page | Back to www.tagg.org website |
The links below lead to the individual components of PLOVER
| Philosophy | Language | Ontology | Value | Evolution | Religion |
This
version updated on 6