Beings can become painters, says and shows Alain Lioret.
Work in progress
The developments proposed here are not specifically aiming to digital art, but are an attempt to an ontology of the digital world. See the home page of this ontology. For thought archaeology, this ontology is an evolution of our reflexions about systemics (circa 1973, in French).
Our eyes see beings as a population of
entities, materially grouped in bodies and machines, partly connected by wires.
What we do not see are the net of "light" connexions
- by light and sound, which have been the major communication media of living beings for long,
- by electronic waves, which began to operate with the wireless communication systems at the end of the 19th century or our era,
- chemical by odors or enzymes,
- by social relations between living beings.
If we zoom onto these
beings and communications, we shall find smaller elements : computer files and
cellular nuclei, records, texts and genes... and, on our way down, just above
the final "atomic" bit level, a general use of standard minimal
- "bytes", i.e. sets of 8 bits (or multiples). That allows to point easily on series of useful beings, as numbers or alphabetical characters, or pixels.
- "pairs" of two bits in the genes. .
These beings and their communications are in constant move, change their features, their distribution change, their communication rates.
The basic grid. When the universe is large, the digital grid is "deep under", but must never be forgotten (see life)
1. Being... a product?
A being is basically known by other beings who have its address. This one may be soft. But we must never forget that the being, to be real, must be instantiated somewhere materially.
A being may have an internal structure, but if the being does not include a processor, then this structure is meaningful only for other beings of sufficiently high level to “understand” it.
A being may have internal addresses.
Seen from outside, a being
is a product by an address and a type, or class (that implies libraries of
types or classes).
The address may be managed indirectly by GOS or an appropriate subsystem.
The type could be just "a bit string".
A being realizes a sort of coherence between the 4 Aristotelian causes. It has the necessary matter, the proper form to realize its finality and proper input (efficient cause)
Then the yield will be sufficient, and the being may live long, because this conformity of finalities, coming frequently from the environment, gives hope to receive the means of subsistence
The existence of a given being is defined from an external system. But we shall see that things will become more complex with large beings, which take in charge by themselves (at least some aspects of) their existence and the preservation or their borders.
In some measure, the internal structures of a being are controlled by the being itself. A self referent being exerts a pressure upon itself, keeps itself together.
2. Basic binary beings
The most basic being is he bit itself. The mos global is the "Digital Universe", if this concept is meaningful... and not excessively paradoxical.
Thesis. But for ulterior rules, beings are either distinct, or one is contained in the other. There are no "trans border" beings, nor "free" bits in a being (this point could be discussed: we suppose that all the bits between the addresses belong to that being. Adapt that when there is more than one dimension).
Internal addresses may be masked to external user beings. Like in OO programming.
The limits of a being may
be defined :
- the limits of other beings,
- implicitly (?), materially, by the characterizing of the supporting device, for instance the organization in cylinders and sectors of a hard disk (or more soft ? including the word length)
- explicitly, by a parameter contained in the being (header and giving its length ; that supposes that the user of the object knows how to interpret this header ; if the being contains a processor, this latter may define the length.
How are defined the limits
of a being:
- extension (in spaces of one dimension)
- formal formula, more difficult if many dimensions)
Anyway, the limits of the beings must be known by GOS and the other beings in order to protect it against destruction by overwriting.
An important point is that the metric qualities of a being, in particular its mass, change radically its features.
- a rather large being generally (some million bits) contains elements describing itself (header, at least)
- anyway, if the being is large, its type may generally be recognized from the content itself.
Thesis: the larger is the being, the richer is its structure.
Thesis: the larger is the being, the more it is self-referent. Would it be only for efficiency effects. Typical example : header of a file, descriptive parts of a computer mother board and operating system.
Question: is there a minimal (digital) size of a object for even a minimal kind of self-reference ?
Note: the Leibnizian monad is paradoxical (illimited complexity within no space)
What holds the bits together? Here the four Aristotelean causes work rather fine. But evidently for DU itself globally.
Materially. The general robustness of DU. Plus the material connection of the bits. To begin with as connected regions, materially. The hardware hierarchy: gates, chips, cards, frames, farms, networks, The Net.
Formally. That is mainly given by the formal definition of a being, its type and description, which is external to the being for the small ones, more or less internal (header) for larger ones. Form may include quantitative relations (format, possibly "canon").
Function, of finality gives hopes to survive if somebody finds it useful. We shall talk later of "pressure", or ranking, which are somehow measurements of the functionality or utility of the being.
Efficient cause concerns more the changes of the being than its existence, but for the initial move: the creations of the being.
2.1. Organ, function
These concepts apply to a being seen from outside. This point matters in particular to express the Moore 's law, and more generally the yields.
A being may be analyzed top down by analysis, as well functional as organic; at some times they coincide, and a function corresponds to an organ, and reciprocally. But sometimes not.
The organ/function distinction comes from observation of living beings (anatomy/physiology). It has been applied to information systems design (functional then organic analysis).
2.2. Material features
Seen from outside, or from GOS, seen as an "organ", a being is mainly defined by the space it occupies. Sometimes an address may be sufficient, but in other cases a more elaborated spatial description may be necessary.
In the physically biological world, an organ is first identified by some homogeneity of color and consistence, which will be confirmed by a deeper analysis, showing the similar character or these cells. It is frequently surrounded by a membrane.
We frequently read that an organ is built bottom up. But that is probably inexact, or at least imprecise. A body dissection or a machine dismounting will show organs top down!
The relations with the rest of the system are primarily material, space vicinities. But also various "channels", be they neural or fluid conduits.
2.3. Functional features
A being is also defined by
its function, a concept near to the "finality". In programming, as
well as in cellular tissues, functions refer frequently to type (epithelial
cell, integer variable). The bit is the most basic type, with no specific
function, unless explicitly specified (Boolean variable). As in object oriented programming.
But the type may be very general: bit, etc.
A function is generally defined top down. May be defined bottom up by generalization.
A function refers to something else.
In abstract systems, it is
tempting to think that descriptions can be purely functional, when in concrete
ones; there is never a perfect correspondence between organs and functions. An
organ may have several functions, and a lot of functions are distributed among
several organs, if not in the whole system.
2.4. The "Overblack" concept
As a form of exploration, we
could try to specify a machine which would maximize the non-isomorphism between
organic and functional structures:
- each function diluted in a maximum number of organs, if not all
- each organ takes part in a maximum of functions (cables and pipes included?)
This is a also a limit of orthogonality pushed to its extreme.
3. Active/passive beings
At first look easy and
intuitive, this concept is not so easy to be defined formally. As soon as the output depends on an input, a being does a "processing".
There are limit (or trivial) case of passive beings/processors:
- the pit: it has inputs but not output (or possibly not related to I),
- the null processor: output = input,
- the quasi-null processor: outputs are nearly identical to inputs.
A being is said passive if it does not change otherwise than by external actions (I) (?)
Some case of
- pure source: no iinput, but outputs,
- output radically different of input, with a real processing depth
- a large time liberty of output emission in respect to input.
We can go farther, structurally. At some times, the being may decide to be passive, while remaining largely master of itself.
A passive being has no evolution between cycles. It is a pure address where other beings may write and read. There is no protection but the protection afforded other interested beings.
Note: A message is considered as active if the host lends him the handle (that has no being for a two bit processor). The active side may be determined by the physical character or the host (said passive against soft...)
The clock is an important sign of activity. But a being could be active using the clock of another one.
How can a clock be recognized from outside?
If a observer being has a cycle much shorter, and observes long enough, it can model the values of the clock bit of the observed, and assess its frequency and regularity
of course, this clock could be controlled by another being.
If the observer's cycle is longer than the observed's one, it could perhaps detect and assess the clock on a long observation time,
In physical machines, the distinction between operators and operands is not a question. But, in digital world, "operate a representation" and "represent an operation" could be strictly synonyms. Then there remains no difference between operators and operands, programs and data, etc. Not only is the informational machine reversible (as the perfect thermal machine), but there is reversibility between machines and what they do machine!
Nevertheless, in the
general case, we can note:
- difference in repetition; the operator does not change at each cycle of operation; in DU, normally, there is no wear. in some way, an operator "massifies" always the being it processes..
- an operator is supposed to be heavier..
Note. The fusion between operators and operands demands that a limit has been reached where it is possible. beings must be light enough, de-materialized, to follow the move without delay nor deformation; and operations, through sensors and actuators, must be reduced to an operation on digital beings... But, moreover, the general digitization was a basic requirement.
When you enter an lift and push a button, your gesture may considered as well as an operation on a representation (the stories, from low to high), or a representation of your intention: the story where you want to go.
One could say that a basic dissymmetry remains : the initiative comes from me. But it uses the initiative of the lift builder...
5. The basic "agent"
We give a particular importance to a basic "agent", inspired from the finite automaton, which has a good side: it is widely studied, not only by automaticians but also by logicists and at large by digital systems specialists. See for instance [Sakarovitch 2003] (in French).
"Agency" as a feature of a being...
5.1. General principle
The description that we give here calls for a deeper one, in order to link it to the general theory of finite automata. With its inputs outputs, internal states and operating cycle. We shall frequently abbreviate with S.
The basic schema of a processor (or being, or system S): input, output and state (E).This scheme could be adapted to the four Aristotelian causes scheme, with efficient cause as I (which lets in E and its function a auto-mobility), formal as type, material as implementation (in the DU raster) and finality as output (that is only partial).
Materially, an automaton is a binary being, with its beginning and end addresses. Moreover (applying the finite automaton structure), we may part it in three regions, I,E, O. The O region groups the bits to be emitted and the addresses. In the computations, we will admit the I bits are in the beginning of the being and O bits at the end.
All beings internal to a processor communicate with the external world only through the I and O zones (or ports) of the containing processor. Then a processor defines firmly an internal zone and an external region. And the more a processor is developed, the more it controls its ports. More firmly, but also more intelligently (open systems).
A being is seen from outside (image) as a part of DU, with a succession of I (non state region), E (enclosed region, not accessible from outside) and a region O with its states accessible from outside, either at any moment or an intercycle moments (synchronous model).
The agent operates by
cycles. Within each cycle, each processor operates according to its internal
laws. Then, transmissions occur between beings. In that elementary model, we
consider that : :
- the cycle duration is sufficient to let all internal operations take place,
- there are no communication limits between beings, bur for distance limits (see below).
5.2. A tentative minimal model
Let us try to elaborate a minimal model of "core" for an automaton, with B = clock plus the constant part, unchanged, or possibly copied from cycle to cycle. B may be considered as the program running the automaton, or the parameters of its class. B includes the e and f functions, of pointers to functions that can be external. Somehow, the operating system.
By construction, as long as this automaton remains itself, B is constant, but E changes accordant to the inputs I.
We say nothing of O, which does not matter at this stage (as long as there is no remote random looping on I). B could be "subcontracted".
(this scheme should be taken also in ratios, with typical cases).
If S is really finite, there is nothing else, in bit numbers
Te = b + I + E (possibly + margin)
The loss bears not only to S (operating bugs) but on B. Then, after a while, the system no longer knows how to reproduce itself correctly, and dies.
We can add into B and E some redundancy and auto-correcting codes, but that increases its length, hence its relativist frailty. And so more since we must add also B in order to integrate the coding/decoding algorithms.
Then everything depends on
the relation between the two growths.
Loss probability on length function of Te
And this length itself depends on the auto correction rate.
The reasonable hypothesis is that auto correction is profitable up to a given level, which we shall say optimal. Hence, we shall consider that their relativistic loss rate is the net rate.
Once that rate known, we can deduce from it the probable life expectancy of the system (excluding the other possible destruction factors, like external aggressions for instance).
We may also make hypotheses on I, considered as meaningful for L.
Favourable case: I brings
correcting elements to E (and also to B ?)
Notes about B: limit of its reduction, is reliability.
5.3. Automaton cycle
On every cycle (the cycle
concept, with its radical cut inside time, is fundamental in the digital
concept), each processor changes according to its internal laws. Then
transmissions occur between automata. In this elementary model, we consider
- the cycle time is sufficient for all internal operations
- there are no band pass limits between automata, but some distance minima (see below).
Possibly: a being may have its own clock, in order to sequence the internal operations during the cycle phase.
Then we can consider that E contains a part of B which does not change from an instant to the next one, and which supports the invariants of S, including the auto-recopying functions from an instant to the next one.
If we have A = B + E, we assume that GCo recopies B and the whole automaton according to elementary technologies. Then, we need an automated addressing in addresses shorter than log B + E.
And log A
I f A is small, then address is large. e.g. 2 bits address, 2 bit,
A 21K, 10 bits address.
We see that
framing/rastering is profitable, would it be only for this reason.
In a comparable way, we shall have op codes, pointing on op codes.
GCom must grant the transitics, ie transport inside A as well as successive instants, then rabattement.
Relation B/E (neglects the relation to I/O):
We reach here hierarchy,
nesting of sub-automata. As soon as there is a processor in A, their relations
are defined by IEO(A). It is mandatory to
- organically, assign to it an existence region in the string, at a moment in the processing (rabattement); if DF, delay/reliability parameters
- functionally, an external description (type, IO addresses).
TE is the "rabattement" on one dimension of all zones of a S (if infinite ?) TGG is the rabattement on one dimension of the whole of MG
The conditions of this rabattement may be discussed.
5.4. Being identification
Something is there. What (who) is it?
We have a sequence of
outputs O. We try to describe them. i.e. we design an automaton which
generates these O.
This operation will generally be stochastic, heuristic, with an error risk.
Symmetrically, rational identification of the World by the automaton..
Find E and the f and g functions through I and O.
A very important stage is when S is able to know which O has impact on its I. And possibly to experiment systematically, generating different outputs to see what it gets. At some level, the mirror test and the emergence of the I : emergence is the moment when distinction of I from D (what is perceived of D) (????what is D?) . To identify O, S needs a mirror. We shall have the global O, image of the self for the external. Inside this global O we will have the explicit O (gesture, speech, attitudes).
Automata functions are not reversible, in general. Probably it would demand a very particular automaton to obtain reciprocal (or symmetrical) functions giving I = f'(E,O), and E = g'(E,O)...
5.5. Being operating mode
If there are no inputs, the operations are necessarily cyclical, but the cycle may be very long, with a theoretical maximum on S variety, or more exactly on the variety of its modifiable parts (deduced from that the basic program).
(???) For instance, if b(E)
is larger than
Nevertheless, new things
may appear in such a closed phase, if S analyses itself and discovers:
- empty, useless parts,
- better ordering methods (defrag)
- some KC as much as we have generators,
Here, we must look for a relation between the input rate and the drill depth. There is a normal way of operation for S for a "typical" I rate. When I stops, we go to deeper and deeper tasks, until we reach the root. Thereafter, the S is bored.
What about an absolute reduction limit, and a relation between root and HC? That could be studied on a given E, with its modifiable art. But also a DU given with its contents.
We could also speak of some
kind of "internal inputs", it the organic interior of S has with the
environment other connections that the I proper. That should be modeled
cleanly. For instance with a partition of I in levels of different depths:
- operation I, day to day life
- superior I : consulting, development.
On the opposite, if for a
long time, the inputs are very strong and never let go down in depth, there is
a sort of overworking, over warming, and S will ask some rest or sleep
(preventive maintenance), or else ends breaks down.
But, in the internal work, going towards the root, S may find new patterns.
5.6. Structuration of beings
Beyond the simple IEO structure, we can imagine that some intermediary processes, or "motors" (which are somehow beings also, but subordinated), have their own IEO structure.
For instance, a "input processor" can operate like a sort of sophisticated filter, cutting out the noise, selecting the most pertinent inputs and sending them only to the "central" processor ; probably on a very different form, for instance some information ot the kind "an image of Mr X appeared. are you interested to look at it". Or extracting featuers ("Mr X looks angry"...).
- an outuput processor will be some kind of synthesizer (providing expression),
- a memory processor will take in charge all the levels of internal (even external) addressing, selections, etc.
6. Some beings defined by matter/finality
A functional typology may
demand organic features:
- minimal size, or maximal, sometimes strictly defined,
- maximal admissible distance for the processor (due to transmission times).
- Materially mobile, Form defined by format/language. Finality: action on the receiver. Efficient : the emitter being
A message is in duality with beings. We could think of a "message space", with particular openings on GG. A message has a particular kind of address. Near to stream. It moves, and GOS or a subsystem manages the move
Materially a passive being inside S, or called for when demanded. Formally: written. Finality: function. Efficient cause: call to the program.
"Une action, résultat d'une opération de contrôle ou de régulation, est programmable si elle peut être déterminée à partir d'un ensemble d'instructions. Pour cela il faut que la transformation à contrôler soit déterminée et qu'on en ait établi un modèle complet." Mélèse 1972.
"La programmation a pour objet la transcription finale en langage-machine des résultats conjugués de l'analyse fonctionnelle et de l'analyse organique." (Chenique 1971
6.3. Process, processor product
Materially fixed, inside a processor. Form defined by program. Finality: generally an output. Efficient cause: triggered by I or internal S cycle.
To be defined in duality
The role of time. By definition, partly non-permanent, anyway not the same type of stability as the processor, who support the processes.
How is a process organized? Organic and functional definition? In a descriptive system, as with software components, the organic side would be the interface representation; functional on the other hand would be the methods in the components? Or vice-versa?
6.4. Pattern, form, force
These concepts are “open” and vague as long as they are not defined precisely in some formal context.
6.5. Behavior, emotion
On the low levels, behavior
dissolves in matter. At the highest level, it is part of reason.
Architecture of behaviors: hierarchy, sequence, nesting.
A behavior may be innate or learned. From an IOE standpoint, that does not matter very much.
Innate means: settled with
the being creation. A type is instantiated (or a genetic combination of types)
which had previously behaviors.
Then, this behavior may be kept on as an internal resource (accessible by pointing on the type), or, on the contrary, the corresponding code is copied in the new being.
A first, that has no importance but for performance.
Later, on may imagine that the being pushes on its integration, changes this behavior, overcharges the code... then goes further from genericity.
Other behaviors may be acquired/learned.
With diverse degrees in the autonomy of acquisition:
- simple transfer (but different from innate insomuch as it is built after, an independently of, the creation or the being)
- guided learning
- autonomous creation of a new behavior.
Example: the apes learning to wash potatoes.
Emotion. More passive than behaviour. Maybe of rather low or high level. Emotion is a kind of pattern.
Materially : like patterns and behaviors, but larger. Form: idem. Finality: global L, plus pleasure. Efficient cause: learning.
The set of patterns owned by a being.
What can be transmitted from one being to another being core using only its imput channels. Possibly with feedback systems (the learner sending acknowledgment signals to the teacher system).
7. Identity is digital from start
7.1. No identity without digits
identity, even before it is psychological or social, is digital from start, since it is built around a personal digital code, a number written in base 4, our DNA. Then, as soon as we acquire a social identity, through family and registration, it is no less digital, since written in first and second names, another string of digits, and present in more and more digital information systems.
7.2. A purely digital model of identity?
It would be exciting and probably useful to build a purely digital model of identity and its augmentation, around some ideas as :
- the evaluation of the “strength of identity” of a being, based on the (digital) complexity and the specificity of its features, hence its “recognizability” in a digital space;
- paradoxical aspects of identity, as says for instance Lévy-Strauss: “One of the paradoxes of personal identity is that it expresses oneself by adhesion to groups and crossing of collective identities”. This could be modeled with networks of sufficiently sophisticated automata;
- the fundamentally recursive nature of identity, with mirroring effects as seen for example in the master-slave Hegelian dialectics , the Lacan’s “mirror stage”, the Cartesian “I think, therefore I am” or even the scholastic proofs of God’s existence and identity (aseity) ; a formal (mathematical/logical) approach could be particularly promising, since reflexive deductions (diagonalization) are basic for the Gödelian “undecidability” principle and… computer structure itself.
The triple meaning of transparency
Let us now focus on recent decades and the DAI progress up to now, from the modern views dominating in the 50’s to the postmodern explosion giving it a name and global perception since the 1970’s. The dimension of a conference paper, of course, allows space only for a broad brush approach To begin with, our neat parting between the two successive stages must more properly be seen as the dialectic relation of two mental and social “paradigms” that have more or less co-existed from start, even if or course, the prefix “post” assumes that one followed the other. We shall conclude with some views about the future, as far as it is predictable.
8. Internal/external, transparency
Enhancing the beings goes along with a sharper, thicker separation of the being from its environment. Cells habe membranes. Bodies have a skin and clothes, machines have carters and bonnets, cities have walls, letters have warrpers, codes have diacritic signs (including in DNA), files and messages have headers and tailers.
An important point it the relation of a being with its metas, internal and external:
- internal: impossible on
1 bit; possible with 2,
- external: if the meta is much longer than the being, it is an address, and the being is an address or a pointer.
Internality implies a protection. There is also, in the exterior, a "security distance". A security zone, larger than S. Anyway, such a "free" environment is necessary for S to move and expand.
Clothing, housing, fences, are material aspect of that. There are equivalent in digital spaces (file structure).
We consider the existence of God as an undecidable assertion. As says Leibniz "Si Deus est, unde malum. Si Deux non est, unde bonum".
Paris ACM Siggraph, the French chapter of ACM Siggraph, worldwide non-profit organization of computer graphics.