|Dendritic structures on a Petri dish show an uncanny similarity to neuronal dendritic structures.|
From observation to articulation.
(This essay has for the first time been published on steemit: https://steemit.com/ai/@technovedanta/is-intelligence-an-algorithm-part-2)
Many people are interested in improving their intelligence and many tools, schemes and tricks have been proposed for this purpose. Each of these tools (mnemotechniques, schemes to organise information, planning schemes, heuristics etc.) address only a small aspect of the complete process we call “intelligence”, because most scholars dealing with the topic of intelligence do not have a good overview of the total picture of intelligence, what it is and how it functions.
If we could analyse intelligence and arrive at understanding its mechanism we might be able to use it to our advantage. This is the purpose of this series of essays I am writing:
To collect a more complete understanding of what Intelligence is and how it functions and to provide tools for improving our human intelligence as well as artificial intelligence.
In a previous article on Steemit (https://steemit.com/ai/@technovedanta/is-intelligence-an-algorithm) I already indicated that intelligence is a kind of algorithm, which I will only very briefly summarise:
When a living system encounters a problem such as a lack of resources it gets a stimulus to start to probe for a variety alternatives or other solutions.
From its observations and testing out it abstracts patterns. From these successful alternative strategies can be selected.
When encountering contending groups so-called “Intergroup tournaments” can lead to mutual a probing of the distinctions between the groups. This can result in niching, a symbiosis, or an exchange of those features which are different between the groups. Thus the system adapts itself to its environment.
In this and coming essays I will try to discuss different aspects of Intelligence such as cognition, (pattern) recognition, memory, abstraction, analysis, understanding, information retrieval, reasoning and problem-solving, including planning, heuristics and creativity, in more detail. I don’t claim to present you with novel knowledge on this topic, but it is useful to give an overview the teachings of the various Pundits in a simplified manner.
Ben Goertzel the Godfather of Artificial (General) Intelligence defines Intelligence as follows:
“The ability to achieve complex goals”.
This tells us what intelligence is about, its purpose, but it does not tell us how it functions. Since intelligence appears to function in a certain predictable and repeatable way we could say it is a type of (natural) algorithm: a set of instructions defined in a very general aiming to arrive at a goal, in this case a complex goal.
A.N.Whitehead told us that “understanding is the apperception of pattern per se”. Pattern recognition and understanding certainly are part of the way intelligence functions, but this does not tell us how understanding comes about and what to do with it once attained.
Another clue about the mechanics of intelligence is provided by Buckminster Fuller who described cognitive processes as comprising four parts:
- consideration (analysis),
- understanding and
This is a good starting point for analysing at least the first stages of the intelligence algorithm, which will be the topic of this essay. In the next essays I shall address the topics of reasoning, problem-solving, planning and creativity.
When we observe an object or concept, we wish to know what it is, we wish to “cognize” it, so we analyse its form, its material, its constituent parts and the relations between those parts, its function, its purpose and how it relates to its environment.
This analysis is, what B.Fuller calls “consideration”: We build a network of relations, which gives us a framework for understanding, a constellation of facts or as B.Fuller calls it, a “consideration”. Stella and Sidera are Latin words both meaning “star”. A constellation or a consideration is a configuration of facts, metaphorical stars, which together describe a total form if you connect the “star dots”).
From connecting the dots the framework that arises or emerges helps us to understand the object of analysis. The consideration framework metaphorically stands under the topic to be understood.
Such a framework must geometrically spoken at least have three descriptive facts forming three relations, since with only two facts you only have one relation which does not build a plane for understanding. You need a third piece of information to identify what something is. Imagine the information describing an object you get is the following: “black dots”. This tells us nothing yet. If we get the information that there are only two of them at least we can start to speculate about its nature: Perhaps they are the plugholes of a socket, the nose of a pig, the double point symbol “:”. Sometimes you need more than three descriptors, if there is more than one thing fitting the information. Additional pieces of information may suddenly tip the balancing point towards identification, such as a colour or a material substance.
Once we have identified our observation by having a sufficient framework for understanding, its “ontological configuration” we can articulate what we think it is.
We have now completed a cycle of the process of cognition or re-cognition.
When we re-cognise something we label the specific thing as an object belonging to a certain general pattern, a class or category. So we classify the object.
Every category has a certain group of features, which are typical for that category, and the presence of which in an object or concept form the requirements to belong to that category.
To describe an object or concept as complete as possible as regards its features in terms of material(substance), form, internal relations of its parts, external relations with the environment, function, purpose, restrictions and rules etc., is the topic of “Ontology”, the study of being.
Such a list characterising an object, phenomenon or concept is also called “an ontology”.
To build a hierarchical classification of ontologies is building a kind of taxonomy, a classification scheme.
If we acquire a clear taxonomy in our mind, our recognition of objects, phenomena and concepts will dramatically improve. It allows for a rapid retrieval of information, namely to which type of pattern the object for consideration belongs: our pattern recognition skills will improve.
We will have a clearer overview and distinction which aspects are universal and extend across all classes, which aspects are general and belong to multiple classes and which aspects are specific for a class and thereby characterise it as a so-called “idiosyncrasy”.
I fact our minds can form an ontology only if they have observed multiple instances of the same object. If you observe a new object for the first time, you can only try to make an approximation of what it is like. You can try to classify it in a higher ranking more general category if it fits in and if it doesn’t try to see which type of object is the most similar to it i.e. which has a similar use and/or most features in common.
Ben Goertzel stated that “one is an instance, two a coincidence and three is a pattern”. If we have observed three instances of a new object, which share the same features, it’s worthwhile building an ontology for it.
When we build an ontology and recognise shared features in it, it means that from our memory we have been able to abstract aspects which the different instances of the phenomenon have in common.
Abstraction is a form of making a simplified generalised representation of something. For instance if we are able to abstract a tree to the structural features single stem, branches, roots and leaves, it means that we have from all different types of stems been able to abstract the quality of stemness, from all different types of leaves, the quality of leaveness etc. A simplified representation in words, which allows for recognising new trees as trees and to be able to distinguish them for instance from a bush.
Artificial general intelligence is concerned with designing universal pattern recognition protocols, which inherently involve the process of abstraction. Abstraction always follows the pattern of going from multiple specific branches to a single generalised concept. In that way the process of abstraction could be represented by a tree-form. Interestingly enough our neuronal structures also follow that pattern.
The ontology must also be linked to other ontologies, representing other objects, phenomena or concepts with which the one under investigation has a relation. The internal relations between the parts of the ontology must also be described as part of the ontology building.
You may have noticed that I make a distinction between objects, phenomena and concepts. Without wishing to define these fully at this moment, please note that with an object I mean a physical, tangible object, with a phenomenon I also wish to cover non-tangible physical manifestations (e.g. light, sound) and with a concept I wish to refer to mental representations or compounded ideas and schemes, which necessarily involve a degree of abstraction.
Thus far the first part of our intelligence protocol involves: observation, consideration including pattern recognition and feature abstraction, relation (web) building, giving a framework for understanding and thereby completing the building of an ontology, which subsequently allows us to articulate a (mental) abstract representation of the observation allowing for future recognition of further instances of the item of observation.
From specific we have progressed to general, which makes it easier to store the item in our memory for future information retrieval and re-cognition.
This process can even be further improved by further abstracting the object to a simple “glyph”. Chinese characters were formed as such simple glyphs. The Egyptians had their hieroglyphs. Alchemy used glyphs. Such simplified images representing most essential features are easier to retain mentally than complex lists of features in words.
In computers data are often compressed. When we have to learn long lists of information, we also have certain “mnemotechniques” at our disposal to compress this information.
For instance we can form words or phrases built from the first letter of each most crucial term per item in the list. these words we can try to capture with an image thereof if possible or if not of a word that sounds similar and does have a visual representation. Ideally you form a set of glyphs, which you can mentally visualise arranged in a street through which you walk. Each house has a statue of the glyph in the garden at the number of the list.
These are great ways of compressing information and making them easily retrievable. Information retrieval can be improved by building webs of relations between the terms to be memorised so that together they form a single whole, a single configuration that can be glyphed or assembled in a kind of fairy tale.
Other mnemotechniques can be numerical coding using “Gematria”-type of techniques in which words can be coded in numbers or vice versa numbers can be coded in words, depending on which mnemotechnique is more adapted for you. If you are a musician it can help to translate the number sequence into a melody, which you can remember, as each cipher from 1 to 10 can represent a note e.g. from c to e one octave higher. If you are artistically gifted you can try to put multiple visual representations in a picture as a whole that makes sense.
With these coding techniques we can improve the use of our storage space, storage speed, information retrieval speed and recognition speed.
These techniques can improve our learning abilities significantly. When it comes to memorising complex and huge quantities of information it is worthwhile to distil the most characterising word (or two words) of each phrase and reformulating the phrase in such a way that the characterising word (or words) is the answer to the question. Or you make a list of the characterising words and apply the first letter word building approach mentioned above.
It is very useful to identify for each question you make, which type of question it is in terms of the so-called 6 W’s (Who, What, Why, Where, When and hoW). These questions are also typical for ontology building and can help you to remember an item more easily.
In this essay I have described the first stages of the intelligence algorithm concerning observation, consideration, understanding and articulation in conjunction with mnemotechniques and ontology building tools to improve our abilities for these stages. I hope you will also read my next essays on the topics of reasoning, problem-solving, planning and creativity. If you don’t want to miss it, you can follow me.