On the Self-Similarity of Intelligence

Adaption to the environment has been the core of evolution for years uncountable. It is arguably the simple result of the effect of the forces of nature - just as stone is washed out by running water, more complex entities, even life, are influenced in very much the same manner by nature around them.

In these adaptions, most successful are those which are the most "general" for a given set of circumstances while using optimally a certain set of resources. - In short, to re-use the same solution for the same problem. An organism needs nutrients - then it consumes them. It uses the SAME method of consumption, however, and does not try to "re-invent the wheel" on each lunch - it is not today anaerob and tomorrow photosynthesizing and the day after trying something else.

If circumstances require change, an organism can mutate and adapt. All organisms in this regard are like sets, each organism trying a slightly different approach, some succeeding better, some worse, and thus mutating the set over time. If a circumstance change slightly and slowly, intelligence is not required, but with the dying out of less apt organisms and the spreading of those better adjusted, the set of organisms adjusts to the new circumstances.

Albeit such slow evolutionary adjustment can lead to marvellous results (think of nests of birds or state insects), it is very costly. You lose generation after generation for minimal change, until many generations later, you have a better "fit" to the environment.

This is where intelligence creates an additional advantage: it allows to adjust to patterns more quickly than by evolution. If a human in the forest needs shelter, he will construct it out of the branches around him. If the shelter works, he may construct better shelters in the future, change the trees and skins he uses... he will not need to adapt into doing this like a bird. The problem will be different, but the answer will mutate more quickly. Less generations, less investment in resources are needed for a better adjustment.

Now, an intelligent response to a challenge really means the recoginition of an eventual pattern in the challenge and the response to the challenge with a "known working reply" (this is where learning is useful). You are attacked by a wolf - defend with a spear. You are attacked by a wild cat - defend with a spear (the attack is similar, the pattern is recognised). You are attacked by a swarm of bees - a different attack, the spear is pretty useless, "defend" with jumping into a lake. A human could have evolved to match those patterns (just as a snake eagle has specialised), but recognising the patterns intellectually has been more advantageous.

How is a pattern recognised? - When its sub-patterns are recognised into "sufficient" details. - What patterns are recognised? - Those that repeat. - You certainly have some routine in doing a thing, e.g. having lunch. You know how to handle spoon, knife and fork (and possibly sticks). The food will be different, but you will know how to cut it into pieces. The "challenge" can be met with "responses". You know sufficient sub-patterns in order to be able to react to drier meat, or fattier meat, or potatoes etc. It is a challenge-response thing. Such patterns are themselves parts of greater patterns - "how to handle a workday" etc. There are larger patterns that are not so clear - the pattern of how to live your life, for instance. These are not "repeated" really, hence it is harder to learn the correct "response".

So intelligent behaviour really means responding to a pattern with ... itself. Very simply put: a pattern that repeats is a pattern "that works" - or else you should have died of it. You are greeted - you great back. You have lunch - you eat it in civilized manner - pretty much every single time. Patterns within patterns within patterns ... all get their responses. This pattern-matching behaviour is therefore self-similar. How important a detail is (or when it is not important any longer and thus ignored) or how important a greater pattern is (or when it is too great to be noteoworthy and can be ignored) - these things depend on the circumstances.

Now - each intelligent entity can only match a limited set of patterns. The greater the set of patterns is which the entity can match at any point in time, the "smarter" do we consider it. - If one entity is not sufficient to match a set of patterns, a set of such entities can be used. There is no limit upwards, as long as there are patterns to be matched. - For instance, a nerve cell matches patterns - a certain set of activations triggers a certain response. One nerve cell, even though a fascinating thing, does not, however, have unlimited capabilities. It matches only a certain set of patterns. - That can be improved by having several nerve cells (forming a ganglion, perhaps). If that is not sufficient, a set of ganglia is possible (forming a brain). If that is not sufficient, a set of brains can be formed (forming a society - of humans, of bees, of...). If that is not sufficient, a set of societies can be formed (the countries of the world), etc. - Each time one individual matching mechanism is not sufficient for an abstract enough pattern, a set of individuals can be formed that co-operates into matching said abstract pattern. (No single person can learn the chemistry needed for his shirt buttons while constructing a car or a computer and preparing his medicines - but as a society, we can. - Nay, "it" can - "it", the society, as a form of individual.)

Arguably, then, there is only an "individual" for as long as it is necessary to match a set of patterns successfully. If the need for high abstraction falls away, then the higher individual may fall into its subparts (as a society descends into anarchy - still, "local communities" tend to survive nonetheless; the individual becomes then "the village" or thelike). If the need for abstraction arises yet again, sub-parts can again be composed into a higher individual.

I know you will ask, "where is the individuality?" - the individuality matters only for the mutation of the pattern matching. One single element can deviate sufficiently from the usual pattern matching that it learns to match new patterns (heal a new illness, construct a new machine). If no-one ever deviates, the "swarm" or "state" is dumbed down (a regrettable state that befell most if not all of human societies - the Greeks and Romans somehow never reached the steam engine, perchance even overlooked it [referring to the aeolipyle, known likely even to the Egyptians]). Then meeting a new obstacle means not sufficient patterns are known to tackle it (it is not without irony how China was subdued by the very gunpowder it helped invent, but failed to further refine). If the deviation is too high, old and still needed patterns cannot be matched reliably and the entity dies without being useful (commonly known as "crazy people").

You may ask, "what is with self-consciousness?" - To this I will answer - how self-conscious are you in your sleep? In your awakening? In your dreams? As an infant? Under drugs? - This question is asked mostly, as if it were a black and white thing - as if consciousness were some bright inner light that is kindled. - I truly believe, it is a very gradual thing. It exists in some form in every kind of pattern-matcher. Its core is the "evaluation of evaluation" - to not blurt out a response to a challenge in a reflexoid manner, but to "re-consider", i.e. to ask, "is that response suitable" etc. - That only means to extend the input pattern. This is no change of concept, but only makes the input pattern "bigger" - so there are more sub-patterns to consider when delivering an answer. - The response still is "reflexoid", a response to a challenge, just on a broader set of circumstances and with more sub-patterns.

For instance, when you get a "good evening" from your girlfriend or wife, it is different than a "good evening" from a Gestapo interrogator in the 1940s. Both might be waiting for you in your home. Both might be smiling. Still you will think it is different. - The reason of the difference is memory, retrieval of patterns - what the evening will probably unfold like. The "challenge" in both cases is different, though not expressed solely in the words "good evening". And yet, the challenge-response-mechanism is in place here, too. (Not all responses necessarily go into the outer world. You might not tell the Gestapo guy about your suicide capsule or concealed gun and you might not tell your girl straight how you will let chocolate melt on her skin.) But the "challenge", the situation, will nonetheless elicit a "response", regardless of the details of its expression. - This goes for humans as it goes for bees or for single nerve cells (all those nervecells which have a connection causing an evaluation of themselves or other nerve cells, rather than of sensors or actors.) Even a fruit-fly has "consciosness", though perhaps on a very, very low level.

The reason why we think of "our consciousness" is simply because we know no higher consciousness. You cannot predict your opponent's next 20 moves in chess. We actually know of no lower concsiousness, either. You cannot truly imagine to be a worm. You would be sensitive to light, chemical challenges... but you have no idea what that is like. Yet a worm is such a simple being, is it not?

If we are truly sincere, we cannot even imagine simple things we take for granted. For instance, say, I challenge you, imagine a dove. - Now, you have it in front of your inner eye. That dove. Look at it. Got it? - No, you do not. - And I shall prove it to you: what precisely is the shape of its feathers on the rim of its left wing? Are its feet pinkish or greyish, are the nails long or short, does it stand perfectly still or walk around naturally? Where does it have precisely little speckles? Are they white, or greyish, or black, or a combination? Are the feathers on its head perfectly smooth, or are there a few standing out a bit and shaken by the wind?

A second ago, you were sure you had that dove before your inner eye, and now, it unfolds to you in details that were NOT part of your mental image before. How complete is your mental image anyway? - Originally, you "imagined" an abstract dove, an abstract challenge - a more detailed challenge was not required. Only when I started to discuss the details, then you imagined them - as the challenge refined. So it is with everything: what will you do in the evening? - What will you watch? How about your supper? How EXACTLY will you place it on the plate? - You do not consider these things until detailed challenges require them. And then you do the detailed pattern-matching.

So there goes the individual - you are no more an "individual" (matching patters through nerves) than the human society is itself (matching patterns through people). Everyone who has had a convulsion or involuntary movement will know what "the individuality of nerve cells" really can mean, too. There is "individuality" in all levels; but the question is not about the individuality - it rather concerns the ability to deconstruct patterns into regularities in order to reply to them. The question whether an entity is intelligent really is the question whether it can respond to patterns of challenge to itself - no matter by what sub-parts it is composed, nerves, ganglia, animals.

What is the final goal? - Ideally (and not practicably), to have only one (very general, made out of many subparts) challenge, and to know a (very general made out of many subparts) response to it. (When "every day is the same", but you have sub-patterns of whether you eat, drink, lead war, paint or do push-ups.)

Therefore, the efforts in creating a "true" (strong) artificial intelligence should be directed at the (de-)composition of patterns in a flexible manner, with the purpose of being able to re-combine the knowledge and reply to higher patterns. - The method of exactly how this is done - whether with a neural net, or Bayesian reasoning, or that theory of logical triangulation I proposed here earlier, or any other method, is irrelevant; the point really is, it must "work", i.e. reply properly to the challenges by matching their underlying patterns. It is not the goal to imitate the human, in my opinion; but the goal is to reach intelligent responses.

- Nino Ivanov