[ivan.moony]

Project E-Teoria

There is some project I've been dreaming about for the last few years, and I'm preparing to finally shape it up. Finally, it would be a web site that would unite all the scientific fields and open a possibility to upload different science theory fields that could mutually interact to derive a new, yet undiscovered knowledge. To do this, we need a uniform Turing complete language, and this is where Implika hops in. Implika, as I already wrote about, is actually a rule-based inference engine that is closer to human way of thinking than usual programming methods. Theorem provers are the most popular application, but it is not the only possible use of rule-based declarative systems. Processes like induction, deduction and abduction make Implika a Turing complete system capable not only to describe logical systems, but also more general kinds of systems known in systems theory. And as the Universe is a big system composed of smaller and smaller systems, I hope to help to discover valuable knowledge about those systems by the project named E-Teoria.

Beside technical specifications of different theories written in lisp-ish language Implika, there is another very important aspect of keeping track of all the theories at one place: making their descriptions and use manuals readable in human language. For this purpose, I'm working on HTML alternative that could annotate formal theories. I was never a fan of HTML, and I feel like HTML + CSS + XSLT system could be greatly improved by replacing all the three pieces by a single s-expression based language. I'll publish some materials about this part soon.

Finally, just to add some nice graphical structure to documentation and scientific content, I'll try to use the Orbiteque fractal tree project I also have written here about. Oval ingredients could host scientific formulas, deduction interaction trees, theory explanations, scientific graphs, or whatever other content we can think about in a form a tree structure.

The whole E-Teoria is a huge scale and a big resource investment project that is developed by only one human: me. But I'm hoping for additional help from possible end users. To get a feeling about competitors field size, you can take a look at Wolfram Mathematica and Wolfram Alpha. I'm aware It's a kind of sick sized byte for one man alone, but I'll try it anyway. W-Mathematica alternative would be based on Implika, while W-Alpha alternative would be a central (or hopefully distributed) site which purpose would be to share independent user created documents and Implika theory source codes (combined gives something like Wikipedia on steroids). I hope to reach my goal by making the project site data crowdsourced. For a start, only E-Teoria core would be offered as a offline browser application, while the actual site content (on which the project success depends) would be filled in later with data by welcomed users.

If you are interested, you can track the development of E-Teoria right here, on this thread. If you possibly have any critics, questions or ideas you want to share, I would be be thrilled to hear about it, or try to answer.

[LOCKSUIT]

So a unified theory of everything? The most important or high level knowledge?

An alternative to your orbital fractal tree would be my high-dimensional 2D graph. A heterarchy. Each item can link to any other item with a weight on the connection. As 2D. It's word2vec or seq2seq but as 2D and fully interactive/modifiable with no nearby affecting neighbors.

[LOCKSUIT]

See also:

there's no dimension loss

[AndyGoode]

Finally, it would be a web site that would unite all the scientific fields and open a possibility to upload different science theory fields that could mutually interact to derive a new, yet undiscovered knowledge.

I don't understand what you're trying to do. How do you unite all scientific fields? The only ways I can think to do this are:

(1) try to combine ideas across fields - interesting, but I don't see how this could be done
(2 ) make very general statements, especially heuristics - interesting, but mostly a philosophical result
(3) come up with the equivalent of Maxwell's equations for each field and subfield - very likely to fail
(4) use the same terminology where applicable - doesn't sound very productive

Similar projects that *would* be useful in my opinion are:

(1) create a compendium of problemsolving heuristics - like Polya's book, except much more extensive, very valuable
(2) find a way to logically organize all the different topics - interesting and useful
(3) create a science news-like page that tells the state of the art in each field - extremely useful

I have a lot more to say but I'll pause there and wait for a reply first.

[LOCKSUIT]

Heterarchies are good for meaning understanding, translation.

Hierarchies are meant for entailment, sequences.

If you have many sentences you want to 'connect', i don't think storing them in a hierarchy is as useful. There's many connections lost. Ask yourself this, do they really build together? A common parent can connect them but you miss out on billion of connections that aren't possible to 'separate' individually in a hierarchy, without a heterarchy ontop it. You can use both.

You can inject dyes onto nodes to see the spread activations in red glow...etc!

[ivan.moony]

So a unified theory of everything?

Well not quite (as a popular physicists' name TOE comes up), but I'd dare to say, in a way it is. For a start it's just a general set of templates (only implications, variables an constants for now) by which other theories can be formed. It's a theory about theories, a metatheory in other words.

Heterarchies are good for meaning understanding, translation.

Hierarchies are meant for entailment, sequences.

If you have many sentences you want to 'connect', i don't think storing them in a hierarchy is as useful. There's many connections lost. Ask yourself this, do they really build together? A common parent can connect them but you miss out on billion of connections that aren't possible to 'separate' individually in a hierarchy, without a heterarchy ontop it. You can use both.

You can inject dyes onto nodes to see the spread activations in red glow...etc!

Are you talking about  hypergraphs? They use it in OpenCog, but I think it's just a specific kind of nested binary s-expressions which I interpret merely as binary implications. We can combine implications to describe anything, including hypergraphs. I find the invent of s-expressions in programming theory very important success, although the main programming language that exploits them (Lisp) is rather widely unpopular, unless we are talking about AI.

[LOCKSUIT]

Mine is better than hypergraphs though....that viz only shows that a node can link 100% in strength to as many other nodes as it desires, but mine allows adjustable weights between every node to every other node.

[LOCKSUIT]

Ivan, I remember you said you knew so much you knew you knew very little. Can you write a wall of nonstop text explaining 80 different things / related you've learnt? I'm interested in what you know :D Like this:

"People are self-focused. Like stars. They can release back all their heat not just take. The brain learns patterns yet all is patterns. Sodium makes chains like the brain does. Brains are like cells, society is like brains. Evolution of data is.... "

Because my theory says things unite, not expand. You said you knew so much that you knew it was impossible to ever close it into a unity...

[ivan.moony]

Finally, it would be a web site that would unite all the scientific fields and open a possibility to upload different science theory fields that could mutually interact to derive a new, yet undiscovered knowledge.

I don't understand what you're trying to do. How do you unite all scientific fields? The only ways I can think to do this are:

(1) try to combine ideas across fields - interesting, but I don't see how this could be done
(2 ) make very general statements, especially heuristics - interesting, but mostly a philosophical result
(3) come up with the equivalent of Maxwell's equations for each field and subfield - very likely to fail
(4) use the same terminology where applicable - doesn't sound very productive

If we have a Turing complete language (can describe any kind of computation that may exist around us), and if it works similar to a way humans think (rule-based engine), we have a useful formal language for describing theories because all theories are about predicting reactions based on actions in this or that way (which we may interpret as computations). Think of it as an assembler language for science, instead for business. Like in business language, we may form this or that paradigm, possibly enabling interactions between code written in different languages. If we wisely recommend and provide interfaces and templates by which formal theories are formed, then these theories can rely on each other's computations (like counting number of electrons in atoms may rely on parts of mathematics and physics).

Now, how to derive unknown knowledge? Firstly induction forms statistical assumptions about sets of similar data. This is new knowledge discovery and it gives us a base for deduction. Deduction derives implicitly contained knowledge described by the starting set of rules which we may get by induction. We can consider performing deduction in logic like repetitive applying formulas in math, thus forming solution steps. In fact, performing deduction in all the theories is a generalized form of applying math formulas to math expressions. I believe this part may be called general problem solving.

Similar projects that *would* be useful in my opinion are:

(1) create a compendium of problemsolving heuristics - like Polya's book, except much more extensive, very valuable
(2) find a way to logically organize all the different topics - interesting and useful
(3) create a science news-like page that tells the state of the art in each field - extremely useful

I agree. E-Teoria would be a mixture of all those three, but crowdsourced one. Like a collaborative editor, browser and solver for all the human knowledge.

[Edit]
About problem solving heuristics: Implika, as a cornerstone of E-Teoria, is designed to be a total reasoning language, but with a fine control of which deduction branches to develop further, and which branches to abandon. I think there is a space for implementing a heuristic algorythm within Implika. It would be as easy as assigning conditions under which the branches are investigated further.

[ivan.moony]

Mine is better than hypergraphs though....that viz only shows that a node can link 100% in strength to as many other nodes as it desires, but mine allows adjustable weights between every node to every other node.

You just add a node representing a percentage to pair each node within a hypergraph. It is very powerful thing, I'm telling you. You should give it a chance.

Ivan, I remember you said you knew so much you knew you knew very little. Can you write a wall of nonstop text explaining 80 different things / related you've learnt? I'm interested in what you know Like this:

"People are self-focused. Like stars. They can release back all their heat not just take. The brain learns patterns yet all is patterns. Sodium makes chains like the brain does. Brains are like cells, society is like brains. Evolution of data is.... "

Because my theory says things unite, not expand. You said you knew so much that you knew it was impossible to ever close it into a unity...

Lock, I don't recollect I said I know so much. If I said anything at all, I would say I know just enough, however little it may be, to know that I know almost nothing. But about a train of my thoughts... I think I'll pass... The most of my thoughts are useless anyway.

[HS]

But about a train of my thoughts... I think I'll pass... The most of my thoughts are useless anyway.

Hey maybe that's the point.  Maybe conscious intelligence only becomes necessary when one lacks knowledge of unquestionable reality. Weird and wonderful internal conditions should arise when the phenomenon of life is faced with such cryptic external conditions. We wouldn't have evolved to function like this in a world of recognizable rules, definitely not lol.

We aren't specialized in handling facts like a calculator, that kind of thinking does not mesh especially well with our design. We specialize in dreaming up individual perceptions of the universe. We create and share perceptions like "its best to have an absolute understanding of the universe it terms of its basic rules" but we don't know if that's true, possible, or real. It's a thing we invented with our somewhat untethered understanding.

To function effectively at large, reasoning should always take it's own incompleteness into account. This leads to the advantages of ever more meticulous recursions in nervous systems, which creates ever increasing self awareness. I'm not sure if awareness and intelligence are on/off, or if they strengthen/weaken all the way beyond measurement like gravity.

Maybe what we are creating with artificial intelligence is more like a general solution machine, rather than something that has the opportunity to matter to itself, and wonder about things to engage in wondering about them.

[LOCKSUIT]

I cant make out what a hperygraph is... :) rrrr!! It looks like a blob of mack-n-cheese mushed on each other.

It looks like it's just word2vec super-position axis sitting of dots.....like Glove model

How is is different from a binary tree, or a neural network, or a 3D simulated box with 6 faces, 4, vertices, 12 edges? I don't get it...
https://en.wikipedia.org/wiki/Hypergraph

is it saying the dots are in color locations?
dot1 - green, red
dot2 - green
dot3 - red
dot 4 - blue

Word2Vec can already tell us which words are similar to a given word, and by how much. and if it stored the data, it can tell you why too.

[Art]

Sorry Lock, but a cube/box has 8 vertices, not 4.

[ivan.moony]

Point, line, quadrat, cube, hypecube, ...

I think the trick was in defining the most simple data structure by which every other more complex data structure can be composed of, often in more than one way. It is like elementary particles, say atoms. Watching them separately from each other, it is hard to clue up what they could represent on a big scale. But once you get magnifying glass to a lesser and lesser scale, some structures begin to show up, like molecules, proteins, cells, tissues, organs, life forms, symbiotic relations, communities, ...

Regarding to s-expressions, to which hypergraphs are related as nearly one to one mappings, there is a whole little theory of what can be described with them. If you are really interested in investigating s-expressions, I would propose learning some Lisp embodiment (Scheme or something) as a starting point. It's a magical language with clean syntax, but unfortunately unrecognized by a broader audience outside the artificial intelligence field. There could be a a whole web of interrelated expressions which are impossible to describe in a short exposure. Like increment and decrement in mathematics that eventually may turn into geometry, algebra, infinitesimal calculus, integals, derivatives, followed by eye catching graphical representation of n-th order functions, continuing to the wildest number theories, irrational, complex or transcedental numbers, just to name few. I believe this scale does not even have an end, and it scales up infinitely, beyond our imagination.

What I'm trying to setup with Implika as a part of E-Teora is analogous to defining increment and decrement in math, hoping it could finally develop (with some help from a community) to something really useful to all living beings. I want humanity to be a good friend to all living forms as much as it could be, and I think that knowledge and science may be a way.

[LOCKSUIT]

Ah, high dimensional, but, word2vec is as I said...

I believe the universe is particles in space with no time, only photons that are the moving particles, making delays seem like time loss or growth or etc as we call them.
I believe there is only entailment, translation, segmentation in the universe, because patterns are 'same things' and make us go through evolution.
So, to master an AGI brain or 'knowledge store', all you need is something like a hierarchy+heterarchy to store entailments and relations. If you want to simulate particles in apace or high dimensional particles, you can, using text, 2D vision, and if you dare, 3D block-world vision or higher dimensional video if that's possible. The network itself also acts as a high dimensional modeler.