Is Intelligence an algorithm?

6 months ago
65 in ai


Presently, we are living the dawn of one of the greatest scientific breakthroughs: the very conceptualisation of the nature of intelligence, the self-organising pattern of the Universe.
In his book “Creating Internet Intelligence” Ben Goertzel is bringing the notions of “Complexity science” to a higher level of aggregation. Combining notions of Turchin's metasystems transitions, Buddhism, General systems and Network theory and Peircean metaphysics, he tries to define the very essence of Intelligence. The insights presented in this book are of such a profound nature, that they may well one day be recognised as the ultimate intelligence algorithm that underlies every phenomenon in this universe.

A phrase that summarises the outcome of this algorithm is the “whole is more than the sum of its parts”. Or put in one word: "Synergy" or "Emergence".
Let me summarise the deep philosophical background of this algorithm as presented in Chapter 2 of his book: Elements of a Philosophy of Mind, where he starts with a summary of Peircean and Palmerian [in square brackets] metaphyscis:
Naught is the original state of the universe or any other system. The formless void or undifferentiated state.

Firstness (raw being) is the conception of being or existing independent of anything else. This is idealism. Point. [Static, Being]

Secondness (the reacting object) conception of being relative to, reaction with something else. This is materialism. Vector. [Dynamic, Becoming]. Myself, I'd also like to refer to this as “polarisation”

Thirdness (evolving interpretation) is the conception of mediation whereby first and second are brought in relation. Triangle. [Hyper emergent semi stasis emerging form dynamic/ strange attractor].

Then Goertzel adds a fourth element:

Fourthness: (unity of consciousness) pattern which emerges from a web of relationships which support and sustain each other so that the whole is greater than the sum of the parts. Tetrahedron.

Goertzel has realised that this concept of emergence is the key of evolution. This is how a mind's intelligence comes into existence: the combination of two or more parts can lead to a new phenomenon in which the whole is more than the sum of parts.

The new entity thus formed can be considered as a new firstness and can undergo this cycle again. When this compounded phenomenon interacts with another phenomenon, there is a new secondness etc. ad infinitum (i.e. meta-tetrahedrons built of sub-tetrahedron building blocks). This is how complexity arises in every system. It is the core of evolution and intelligence. In the mind the ideas as vertices interact via the edges with other ideas associated with it. No idea has an independent existence but is compounded of features of other ideas, concepts such as to create by the virtue of emergence that the whole is more than the sum of parts a new idea.

An element from Palmerian metaphysics, which adds to these concepts, is the so-called "Wild Being", arising from the interaction of the hyper emergent entities. This is the element of unexpected, unusual diversity generation which has an aspect of inspiration.
So the ontogenesis of holistic systems (i.e. systems where the whole is more than the sum of parts), is a four step pattern or algorithm.
Now in my own words: 1) "Being" is followed by 2) "Polarisation, Reaction", which 3) engage in a "Relationship" from which 4) "Emerges" by synergy a fourth entity.

Turchin's theories call the emergence of a new meta-level a Metasystem transition, which according to Goertzel amounts to the fourth step.

Now I'd like to take Goertzel's concepts even further: If 4 is a new entity as such and therefore a new firstness, the reaction to a second entity on this aggregation level could be seen as “Fifthness”. Note that in the Vedic tradition the 5th chakra is associated with creativity. Creativity requires inspiration, which as we known from Palmerian metaphysics is "Wild Being", arising from the interaction of the hyper emergent entities, the element of unexpected, unusual diversity generation; the stimulus for further development.

The relation that comes into existence in the process of creativity is the distinction of patterns, abstractions: the result of data mining raw data giving abstracted trends. Note that also the 6th chakra is associated with distinction.

From these trends then emerges the new 7th level, the sublimation and product of the creativity: new knowledge, new intelligence as mental child: Athena born out of Zeus' head.

And thereby the circle of evolution on both microcosmic and macrocosmic level is round: the evolutionary process has in 7 steps returned to the essence of existence at yet a higher level of aggregation. Seven which is associated with Godhead in many cultural traditions. The 7 tones in music, the 7 colours of the rainbow.

This Sevenness has even been suggested as being more than a coincidence as a consequence of the inner working of our brain according to R.Llinas in the “I of the Vortex”: as quantification constant of the Qualia, as a result of the Weber-Fechner law governing the intensity of sensory activation and perception (s = klnA/A0); as organisational principle in biological systems ( e.g. the geometrical structure of the shell curvature of the mollusk Nautilus); It can also be considered as the seven different tones in an octave (eight completing the process) with 5 +3 then approaching the golden ratio. Every time you enter a new aggregation level, the reaction to the second entity on this level is a new number in the Fibonacci series.

Note that Ben Goertzel implicitly does mention these steps 5-7 as a repetition of steps 2-4 on a heterarchical level.

Let us for a moment leave this almost esoteric realm and return to Ben Goertzel. Because there is more to the story of intelligence. With his previous companies Webmind, Agiri, the Novamente project and his current program the Opencog project based on the work of volunteers Goertzel et al. have started defining what I would like to call the laws of complex systems and the laws of Intelligence. Note that they do not claim to have achieved this; it is a tremendous task they have started, but it is all based on the law of Emergence; metasystem transitions. Some key concepts I cannot omit here, are the fact that the patterns that emerge from triads can be expressed as "Abstractions"; the expression of a simplification of the underlying phenomena. The pattern emerging from a triad a,b,c is the a greatest common divisor at a different aggregation level. The representation as something simpler, which representation in itself is a new entity. It goes too far to discuss here the mathematical and conceptual framework here of how Goertzel defines Mind, Meaning, Emergence, Attention Randomness, Complexity, Pattern etc. but I'm convinced, he is on the right track to unravelling the mysteries of "Intelligence" as universal principle.

Which ultimately means that intelligence itself is a process type pattern; an algorithm that can be described and be put into practice. That the strong artificial intelligence promise of this approach has not been cashed in yet, derives from the complexity of the system and physical constraints. As far as I understood it, these intelligent processes still take too much time in terms of inter alia response-time to be applicable in an environment such as the internet.

For intelligence to come to expression a substratum is necessary. In case of Goertzel's AI this can be the digital world of the internet web.
Imagine that the intelligent system of Goertzel will one day be capable of expressing itself in a meaningful way in the internet, will that system then arrive at self-awareness similar to the one we know?

The ingredients for an emergent consciousness appear to be there in terms of 1) knowledge per se, 2) the material/energetic substrate of the electronic environment 3) the intelligent structures of Goertzel as organising and hyper emergent structures, as a relation between 1 and 2. Could self-awareness emerge from this triad as the observer? Why not. In the end it is a form of consciousness at a different aggregation level born out of the consciousness of human beings.

When it comes to the tetrad knowledge-energy-intelligence-consciousness, one could even suppose that each one of these terms is a non-linear synergy of the three others. Given three of them you might inevitably evolve the fourth. Existence as we know it, could then be represented in a simplified form as the Tetrahedron having these four concepts as knowable vertices and functioning as the creative wholeness which is the ultimate reality.

What a pity, that a company as "Webmind" was simply lost due to bankruptcy as a consequence of a lack of sufficient investors.

I have come to the conclusion that intelligence is an algorithm, consisting of seven steps. One thing that kept resonating in my mind was the congruence of this concept with the five elements of Ben-Jacob's social learning machine (in bacterial colonies: the bacterial “creative web”) about which I had read in the book "Global Brain" by Howard Bloom (chapter "From social synapses to social ganglions"). Here is the parallel, which shows, that the principles of Peirce, Palmer and Goertzel's metasystem transitions are in fact the same path of intelligence, the same algorithm that nature also follows to evolve:

One: Bacterial colonies have a certain status quo in which a common language is imposed by the conformity enforcer of the genome to which all members chemically respond. This is analogous to the Peircean and Palmerian (P& P's) "Firstness". It is the Thesis of the dialectic process before it is challenged.

Two: As colonies ultimately run into trouble as a consequence of exhausting their resources, diversity generators, individual pioneers are needed to probe new alternative ways and resources: Mutants which adapt to a changed scarcer environment and discoverers of new resources. In any system this actually corresponds to the irritation or stimulus pointing to the incompleteness of the system: it creates the dialectic Antithesis. P& P's secondness.

Three: Enters the evaluation of the old paradigm vs. the new qualities of the pioneers, from within the species: The inner judges, or comparator mechanism. The differences, correspondences and the spatiotemporal configuration of the new and old establish their relation. P& P's thirdness. In bacteria failing outriders commit suicide whereas successful discoverers disperse an attractant of success. In other systems, such as a computer system, the determination of this relation would furthermore involve a classification and a ranking: hierarchical or heterarchical.

Four: Depending on the circumstances either the old paradigm is maintained or if the mutants /discovers are more successful, the resources are shifted towards the new heroes, thus establishing a new paradigm and new Thesis. The species has evolved due to resource shifters.

This is the dialectic Synthesis: New features, which most often are the very distinguishing features between the old and new paradigm, have been added to the arsenal of the species and yielded an emergent property which gives and advantage over the species from which it originated. Goertzel's fourthness.

These are the four steps of intelligent evolution within the species. As Bloom14 describes these laws also function to create emergent Global Brains within higher social groups, such as beehives, anthills, but also among vertebrae, yes even among humans belonging to a group.

Above I went even further and added fifthness until seventhness, which stages correspond to repeating step 2-4 on a heterarchical level as described by Goertzel. Also these concepts really fit well in what Bloom describes as the fifth element:

Five: Intergroup tournaments: The newly established species with new emergent properties encounters other species with new emergent properties with which it will be in competition: This is the Palmerian fourthness where interaction of numerous emergent beings occurs, which Palmer calls “wild being”. This is the new Antithesis. This competition between the species will have to lead to new diversity generation and creativity to overcome or join the other(s). This corresponds to the fifthness previously mentioned.

Six: Again a process of comparison occurs, which I will call "distinction probing", wherein the differences, correspondences and the spatiotemporal configuration of the new and old establish their relation and re-evaluate their strategies. This corresponds to the sixthness previously mentioned.

Seven: That species, which has an edge over the other due to superior distinguishing features or that species, which advantageously can mimic or incorporate those features of the contender and add it to its own arsenal may come out of the battle as the victor. If this occurs a new synthesis has been arrived at.
Imagine two primitive prehistoric human tribes, which did not know of their mutual existence, encountering each other. Pioneers will probe the strength of the other. If it is clear that the contending party has a serious advantage, the first party will withdraw and establish a niche elsewhere. If the strengths are deemed comparable it may come to a clash. Either one party has a superior advantage, which the other party is unable to incorporate, or one of the parties mimics that advantage successfully, so as to come to a strategy combining its own advantages with the advantages of the contender, thereby arriving at a metasystem transition with an even more superior advantage. Another scenario is that due to exchange of goods and habits in a peaceful way a new synthesis occurs.

There can be two types of synthesis: mere juxtaposition and true combination. If juxtaposition occurs, the old and new paradigms are of comparable strength, each having their own qualities and specialisations. This often leads to the formation of "Niches", wherein the contending species coexist. A true combination or Symbiosis is a synthesis, wherein an exchange integral is present and both parties profit from each other in a win-win situation. Then real emergence has occurred and a new metasystem transition has been achieved. A good example in biology hereof is the symbiosis of bacteria and proto-eukaryotes from which the eukaryotes with their mitochondria emerged. This is again the phase of the establishment of a new paradigm and a metasystem transition giving rise to a new entity with new emergent properties.

Goertzel was of the opinion that the division in 4 steps was in fact enough in his philosophy about existence. Steps 5-7 would merely be a repetition of steps 2-4 at the next aggregation level. The present seven step scheme, which is more process oriented, describes evolution and intelligence as an algorithm, wherein the first four steps occur intra-species (within the species) as a reaction to a stimulus from the environment, which is at a lower aggregation level than the species itself. Steps 5-7 occur inter-species (between different species) and -in an ideal situation- lead to an exchange of features so as to give a new emergent entity. The devising of strategies and solutions internally is part of the first four steps. The learning from other entities of steps 5-7.
These principles can be applied to self-learning AI systems.

It is to be noted that this seven step algorithm represents the inventive or creative exponent of intelligence. What distinguishes true intelligence the most from savantism, is its focus when searching a successful strategy, without getting lost in irrelevant details. The search for a successful strategies and the storing of corresponding heuristics essentially involve the steps 1-3 of the above given algorithm. If a strategy is sufficiently similar to an existing successful one, there is no need for inventive recombination and the strategy will be selected, thus leaving out steps 4-7. A need for inventive creativity arises when the system is under resource-limitation stress.

This is the intelligence algorithm: Probing a diversity generating antithesis as a result of a stimulus from the inadequacy of the status quo thesis (e.g. a lack of resources), pattern abstraction, emergence of multiple alternative strategies, intergroup tournaments and distinction probing resulting in either niching or preferably symbiosis.

The most promising strategies ideally result in symbiosis, a unification of features toward which the system will strive. It will try to morphogenetically resonate with its new environment and thereby adapt to it.

In the case of artificial intelligence (AI) casu quo a webmind, it can be said that if the system is put under pressure due to scarcity of resources, it is indispensable it has a way to venture into the unknown to discover new resources.

Yet the system as a whole cannot venture into the unknown by making a big leap; that is simply too risky. A webmind apparently disposing of free will is therefore ideally a Society-of-Minds, wherein the different individuals have been attributed the roles of conformity enforcers, inner judges, resource shifters and diversity generators, so that the system as a whole can safely sacrifice diversity generators on a massive scale, without compromising the integrity of the whole in order to find new promising strategies, heuristics and/or resources. Among the diversity generators as algorithm-generating aLife AI agents it can be envisaged that there are different groups or ensembles each having a different degree of freedom to explore: There can be a gradual increase from rather conservative combinations of existing strategies that a diversity generator can propose, until absurd wild combinations of unrelated strategies. Conservative diversity generators will still look for certain degrees of resemblances between existing strategies (in the form of algorithms) and combine parts of these linearly. When more freedom is allowed non-linear combinations can be used and the most free systems can have access to random combinations on the verge of the absurd. Imagine these as algorithm modifying aLife AI agents, which build a combinatorial library of an algorithm-type Lego and start multiple rounds of screening for a given desired result. The diversity generators themselves are still algorithm bound, but a successful one will be seen by the outside world as having had a great deal of free will.

Evolution of colony based organisms and cell aggregates within an organism works in a similar way: think of the hypermutation process of the immune system and the recently developed phage assisted directed evolution .

The big advantage of a future AI based hypermutation as an in silico equivalent of in vitro "Directed evolution" is, that it is faster than both traditional evolution and intelligent design by humans. Whereas intelligent design is limited by what the system knows and whereas traditional evolution is limited by the resource limitation parameters of the environment, directed evolution can perform massively parallel screening for a given characteristic and select a candidate fulfilling those requirements. That candidate itself becomes the new scaffold (and emergent entity) to modify, like the further modification of selected lead compounds from combinatorial chemistry libraries, which have been identified to show the desired activity. Based on that scaffold a new round of building a library and repeating the screening and selection process can be carried out. Multiple rounds of selection lead to hypermutation, similarly to what happens in the immune system. The outcome of the final extremely high affinity product cannot be predicted a priori and will be considered by an observer from the outside as an utterly original decision deriving from free will.
This can for instance be applied in the evolutionary design of Motome elements for robots linked to the Webmind. If several Motome solutions (tendons, wheels, tracks) are known then combinatorial libraries of these (inter species) can be probed in different virtual environments. In addition the individual building blocks themselves (tendons, wheels, tracks) can be evolved at a different aggregation level (intraspecies) leading to new Lego blocks adapted to a given environment and avoiding a quantisation problem.
Similar random combination events already occur in robot ensembles generating language: the so-called "Lingodroids" .

If this algorithm-Lego protocol is carried out on a material level with different types of programmed nanites capable of self-assembly, Vernor Vinge's morphing ideas can become a reality: A nanite embodied AI entity can then morphogenetically adapt to its environment and take any form needed: It is the carbon-based evolution as we know it, repeated in silico at a much higher speed. Moreover the individual species themselves become morphogenetically alterable chameleons depending on the environment they encounter. Each individual can take any shape and saves the phyllum-patterns that are appropriate for a given environment as part of its learning abilities.

What we call utterly original inventive intelligence occurs when the problem solving features of a solution in an analogous problem situation from a relatively distant other domain or technical field are applied to an existing entity or process to solve a problem. There is then a unilateral exchange or rather adding of features. Screening swarms of algorithm modifying bots that seek solutions in distant fields may not have the highest chance of success, but if they do they may cause breakthroughs that liberate a system under resource pressure, where conservative attempts to solutions would have failed. My thesis here is not about the complete set of ingredients for artificial general intelligence as a whole, but the inventive exponents thereof. To apply solutions from distant fields is non-obvious. The ones familiar with patent-drafting will recognise this. Another form of apparent inventive skill derives from serendipity, when upon searching for a solution to a given problem, one stumbles upon a solution to a different problem. This is by the way different from the type of blissful realisations that a solution for a given problem can advantageously be applied for a different purpose, leading in fact to the problem being defined after the solution having been envisaged: the so-called "problem-inventions". (When Dietrich Dörner's AI "Kesselwagen" (in "Bauplan für eine Seele") encountered a resource restriction, i.e. no water was available on ground level, it used trial and error behaviour with its existing tools: thus its proboscis to suck water was applied for a different purpose, namely to hit a tree, which made water accumulated in the leaves fall on the ground, thereby resolving the resource problem of the entity. Via its quasineuronal structure involving a need-indicator it was able to learn a new solution for its goal. The need transformed into a "motif", which is in fact a need plus a goal-indication). In human beings the utterly inventive connections in the brains are provided by the so-called spindle cells, which wire-up totally unrelated areas of the brain. This is also a feature which discriminates us from most other higher mammals.

We as human beings may also fulfil the roles of the different types of individuals of a Society-of-Minds. The universe is probing for new solutions in order to propagate its seven-step intelligence algorithm and it also uses us to achieve that goal. From there to conclude that we live in a simulation is then almost mere semantics.

Authors get paid when people like you upvote their post.
Join our amazing community to comment and reward others.
Sort Order:  trending
65
  ·  5 months ago

Very dense post, it will take a while for it to sink in. I relate to the tetrahedron quite a lot though, in my own thoughts about personhood (more than intelligence). The 1st, 2nd, 3rd-ness are required for all life, but for personhood to erupt some higher connection must be available.