Invoking Questions

Warning: Dangerous bend ahead


English has a well-defined set of question words. In brief, we claim all of them can be expanded into a what phrase. To demonstrate:

• what ::= what
• who ::= what be the actor of/that
  
• when ::= what be the time of
  
• where ::= what be the location of
  
• how ::= what be the way of
  
• which ::= what be the choice of
  
• whose ::= what be the owner of
• why ::= what be the reason of

Lets pursue this further.
In mathematics, when one shows the different levels of infinities via the ordinal numbers (and there's an infinite chain of such levels), one accomplishes the construction of such a hierarchy using a very simple trick - names. The moment we name a mathematical entity, we can use it explicitly to create an even bigger mathematical object, using the previous operations we had at our disposal.

I argue this is also the case in the above example of natural language. All primitive thought can be modeled in a very explicit (and artificial to the reader) syntax, clearly postulating the exact meaning of what is implied using only a very poor vocabulary. Translate for example "who and why ate my food" to "what is the actor and what is their reason of eating my food". This is extremely uncomfortable to use in communication, indeed. However, it is already a big step closer to being machine understandable! The smaller the vocabulary and the more explicit the content, the less ambiguity and more precision a machine could obtain from trying to understand the language.

Now, machines are stupid, as they have no real sentience of any sort. Human beings, on the other hand, can deal with adopting new names and use them in novel contexts, handling abbreviation, unexpected abbreviations, metaphors, ellipsis and so on and so forth. We have a very robust and adaptable language processing mechanism in our heads which allows us to communicate with very high efficiency and to overload our languages with multiple meanings and contexts. Ah, sentience.

Well, lets give this post some structure.

Artificial Intelligence:

the Pillar of the Winding Staircase

Back to names. It is my suspicion that achieving automatic and clever ways of introducing, understanding and forgetting names in Artificial Intelligence is one of the keys to approaching sentience. Names essentially allow abstraction, multiple levels of modularization of thought, distinguish between concepts and reveal relationships. It is just a hunch, but it seems very logical.

Next, we need context. Without understanding the local and mid-range relationships occurring in a discourse (a text, a signal, a formula, virtually any non-trivial piece of information) one can do very little. This has been shown and sadly remarked upon both when one tries to automatically generate and automatically understand information. Sentience requires at the very least the ability to generate and understand non-trivial statements, which is the next step on top of understanding the atomic symbols (i.e. the names) of the system.

Third, there is pragmatics. A young child has a first explanation of what a canine is and that a particular dog is of a certain breed. The next time the child meets this dog, or a different dog, it will inevitably make the comparison and relate with its previous knowledge. Its memory allows for a dynamic creation of a "knowledge base" which later serves as a basis for reasoning. This basis of reasoning is shortly known as pragmatics, when one refers to the black-box setting of previously attained knowledge or experience.

These three components have been known for quite a while to the scientific communities dealing with branches of AI, but have been generally insufficient for automatic reasoning to be attained with any "sentient" dynamics. There is a fourth, magic piece of the puzzle, which I personally think makes the whole machinery tick with transcendental grace, but I will hold it as a surprise for now. After all, I should make a proof of concept before making any claims, shouldn't I?

You can already see the mad direction which this post has taken...

Invoking Questions

Recall my translation of why - " What be the reason of ".

This is the driving force behind the topics I will blog about in this calm and quiet online space of mine. The metaphysical drive to always discover the relational causality, the purposeful dependencies, the holistic or reductionist puzzle solutions,
the search for the reason of things.

It is easy to see an inquisitive mind would never run out of question to ask, as the chain of the meta-levels is clearly infinite (See Hofsdatder's GOD example in GEB). But then why ask at all?

Well, do hang around and see for yourself. If you dare.