Category Archives: Wittgenstein – Ludwig

Collective Human Intelligence (CHI)

This text is part of the text “Rebooting Humanity”

(The German Version can be found HERE)

Author No. 1 (Gerd Doeben-Henisch)

Contact: info@uffmm.org

(Start: June 22, 2024, Last change: June 22, 2024)

Starting Point

The main theme of this section is ‘collective human intelligence (CHI)’. However, it does not occur in isolation, detached from everything else. Rather, life on Planet Earth creates a complex network of processes which — upon closer examination — reveal structures that are consistent across all forms of life in their basic parameters, yet differ in parts. It must also be considered that these basic structures, in their process form, always intertwine with other processes, interacting and influencing each other. Therefore, a description of these basic structures will initially be rather sketchy here, as the enormous variety of details can otherwise lead one quickly into the ‘thicket of particulars’.

Important Factors

Basic Life-Pattern

Starting from the individual cell up to the most powerful cell galaxies [1] that life has produced so far, every identifiable form of life exhibits a ‘basic pattern’ of the life process that threads like a ‘red thread’ through all forms of life: as long as there is more than a single life system (a ‘population’), there exists throughout the entire lifespan the basic cycle (i) reproduction of Generation 1 – birth of Generation 2 – growth of Generation 2 – onset of behavior of Generation 2 accompanied by learning processes – reproduction of Generation 2 – ….

Genetic Determinism

This basic pattern, in the phases of ‘reproduction’ and ‘birth’, is largely ‘genetically determined’. [2] The growth process itself—the development of the cell galaxy—is also fundamentally strongly genetically determined, but increasingly factors from the environment of the growth process come into play, which can modify the growth process to varying degrees. Thus, the outcomes of reproduction and growth can vary more or less significantly.

Learning

As growth transforms the cell galaxy into a ‘mature’ state, the entire system can enable different ‘functions’ that increasingly facilitate ‘learning’.

A minimal concept of learning related to acting systems identifies the ‘learning of a system’ by the fact that the ‘behavior’ of a system changes in relation to a specific environmental stimulus over a longer period of time, and this change is ‘more than random’. [3]

‘Learning’ ranges on a scale from ‘relatively determined’ to ‘largely open’. ‘Open’ here means that the control by genetic presets decreases and the system can have individually different experiences within a spectrum of possible actions.

Experiences

Experiences gain their ‘format’ within the following coordinates:

(i) (sensory) perception,

(ii) abstractions of sensory perceptions which are generated internally and which

(iii) are retrievable inside the system. The totality of such conditionally retrievable (recallable) abstractions is also called memory content. [4]

(iv) the possibility of arbitrary abstractions from abstractions, the

(v) storage of sequential events as well as

(vi) abstractions of sequentially occurring events, and the

(vii) free combination of abstractions into new integrated units.

Additionally,

(viii) the ‘perception of internal bodily events’ (also called ‘proprioceptive’ perception) [5], which can ‘link (associate)’ with all other perceptions and abstractions. [6] It is also important to note that it is a characteristic of perception that

(ix) events usually never occur in isolation but appear as ‘part of a spatial structure’. Although ‘subspaces can be distinguished (visually, acoustically, tactilely, etc.), these subspaces can be integrated into a ‘total space’ that has the format of a ‘three-dimensional body space’, with one’s own body as part of it. ‘In thought’, we can consider individual objects ‘by themselves’, detached from a body space, but as soon as we turn to the ‘sensual space of experience’, the three-dimensional spatial structure becomes active.

Individual – Collective

In the individual experience of everyday situations, this ‘inner world of experience’ largely forms in a multitude of ways, largely unconsciously.

However, as soon as ‘human systems’—in short, people—are not alone in a situation but together with others, these people can perceive and remember the same environment ‘similarly’ or ‘differently’. In their individual experiences, different elements can combine into specific patterns.

Coordination of Behavior

It becomes interesting when different people try to coordinate their behavior, even if it is just to make ‘contact’ with each other. And, although this can also be achieved in various ways without explicit symbolic language [7], sophisticated, complex collective action involving many participants over long periods with demanding tasks, according to current knowledge, is only possible with the inclusion of symbolic language.

The ability of humans to use languages seems to be fundamentally genetically conditioned. [8] ‘How’ language is used, ‘when’, ‘with which memory contents language is linked’, is not genetically conditioned. People must learn this ‘from the particular situation’ both individually and collectively, in coordination with others, since language only makes sense as a ‘common means of communication’ that should enable sophisticated coordination.

Linguistic Meaning

The system of sounds (and later symbols) used in a language is ultimately decided by those who initially use the language together. [9] Given the fact that there are many thousands of languages [10], one can conclude that there is considerable freedom in the specific design of a language. [11]

This ‘freedom in the specific design’ is most evident in the association between language elements and potential meanings. The same ‘object’, ‘event’, or ‘fact’ can be expressed completely differently by people using different languages. [12] And this ‘difference’ refers not only to the naming of individual properties or objects but occurs within a ‘larger context’ of ‘perspectives’ on how the everyday world is perceived, classified, and enacted within a specific language community. These differences also occur ‘within a language community’ in the form of ‘milieus’/’social strata’, where the ‘practice of life’ differs.

The most important aspect of the relationship between the language system (sounds, symbols, the arrangement of sounds and symbols) and possible ‘meaning content’ is that any type of ‘content’ is located within the person (inside the ‘system’).[13]

As the brief sketch above on the space of experience suggests, the elements that can constitute an experiential space are so layered and dynamic that an assignment between language elements and elements of experience is fundamentally incomplete. Additionally, two different systems (people) can only agree on elements of experience as ‘linguistic meaning’ if they have ‘common reference points’ in their individual experiential spaces. Due to the structure of the system, there exist only the following ‘types of reference points’:

(i) There are sensory perceptions of circumstances in a commonly shared environment that can be perceived ‘sufficiently’ similarly by all participants (and are automatically (unconsciously!) transformed into more abstract units in the experiential space).

(ii) There are internal body perceptions that normally occur similarly in each person due to a genetically determined structural similarity.[14]

(iii) There are internal perceptions of internal processes that also normally occur similarly in each person due to a genetically determined structural similarity. [15]

The more ‘similar’ such internal structures are and the more ‘constant’ they occur, the greater the likelihood that different language participants can form ‘internal working hypotheses’ about what the other means when using certain linguistic expressions. The more the content of experience deviates from types (i) – (iii), the more difficult or impossible it becomes to reach an agreement.

The question of ‘true statements’ and ‘verifiable predictions’ is a specific subset of the problem of meaning, which is treated separately.

Complex Language Forms

Even the assignment and determination of meaning in relatively ‘simple’ linguistic expressions is not straightforward, and it becomes quickly ‘blurred’ and ‘vague’ in ‘more complex language forms’. The discussions and research on this topic are incredibly extensive.[16]

I would like to briefly remind you of the example of Ludwig Wittgenstein, who first experimentally played through the very simple meaning concept of modern formal logic in his early work, ‘Tractatus Logico-Philosophicus’ (1921), but then, many years later (1936 – 1946), reexamined the problem of meaning using everyday language and many concrete examples. He discovered — not surprisingly — that ‘normal language’ functions differently and is significantly more complex than the simple meaning model of formal logic assumed. [17] What Wittgenstein discovered in his ‘everyday’ analyses was so multi-layered and complex that he found himself unable to transform it into a systematic presentation.[18]

Generally, it can be said that to this day there is not even a rudimentary comprehensive and accurate ‘theory of the meaning of normal language’. [19]

The emergence and proliferation of ‘generative artificial intelligence’ (Generative AI) in recent years [20] may offer an unexpected opportunity to revisit this problem entirely anew. Here, modern engineering demonstrates that simple algorithms, which possess no inherent meaning knowledge of their own, are capable of producing linguistic output by using only language material in linguistic interaction with humans. [21] This output is substantial, structured, and arranged in such a way that humans perceive it as if it were generated by a human, which is ultimately true. [22] What a person admires in this linguistic output is essentially ‘himself’, though not in his individual language performances, which are often worse than those generated by the algorithms. What the individual user encounters in such generated texts is essentially the ‘collective language knowledge’ of millions of people, which would not be accessible to us without these algorithms in this extracted form.

These generative algorithms [23] can be compared to the invention of the microscope, the telescope, or modern mathematics: all these inventions have enabled humans to recognize facts and structures that would have remained invisible without these tools. The true extent of collective linguistic performances would remain ‘invisible’ without modern technology, simply because humans without these technologies could not comprehend the scope and scale, along with all the details.

Preliminary Interim Result

The considerations so far only give a first outline of what collective intelligence can be or is.

[1] Reminder: If we assume that the number of stars in our home galaxy, the Milky Way, is estimated at 100 – 400 billion stars and we take 200 billion as our estimate, then our body system would correspond to the scope of 700 galaxies the size of the Milky Way, one cell for one star… a single body!

[2] We know today that genes can change in different ways or be altered during various phases.

[3] Simplifying, we can say that ‘randomness’ in a ‘distribution-free form’ means that each of the known possible continuations is ‘equally likely’; none of the possible continuations shows a ‘higher frequency’ over time than any of the others. Randomness with an ‘implicit distribution’ is noticeable because, although the possible continuations are random, certain continuations show a different frequency over time than others. All these particular individual frequencies together reveal a ‘pattern’ by which they can be characterized. An example of ‘randomness with a distribution’ is the ‘natural (or Gaussian) distribution’.

[4] The concept of ‘memory’ is a theoretical notion, the empirical and functional description of which has not yet been fully completed. Literature distinguishes many different ‘forms of memory’ such as ‘short-term memory’, ‘long-term memory’, ‘sensory memory’, ‘episodic memory’, and many more.

[5] For example, ‘joint positioning’, ‘pain’, ‘feeling of fullness’, ‘discomfort’, ‘hunger’, ‘thirst’, various ’emotions’, etc.

[6] One then perceives not just a ‘blue sky’ and feels a ‘light breeze on the skin’ at the same time, but also experiences a sense of ‘malaise’, ‘fever’, perhaps even a ‘feeling of dejection’ and the like.

[7] In biological behavioral research, there are countless examples of life forms showing impressive coordination achievements, such as during collective hunting, organizing the upbringing of offspring, within the ‘family unit’, learning the ‘dialects’ of their respective languages, handling tools, etc.

[8] Every child can learn any known human language anywhere, provided there is an environment in which a language is practiced. Many thousands of different languages are known.

[9] The exact mechanism by which a language first arises among a group of people remains unclear to this day. However, there are increasingly more computer simulations of language learning (language evolution) that attempt to shed light on this. Given the enormous number of factors involved in real language use, these simulations still appear ‘rather simple’.

[10] As we know from many records, there were many other languages in previous times that eventually died out (‘dead languages’), and only written records of them exist. It is also interesting to consider cases in which a language has ‘evolved’, where older forms continue to exist in texts.

[11] Language research (many different disciplines work together here) suggests the working hypothesis that (i) the type and scope of sounds used in a language, due to the human speech apparatus, represent a finite set that is more or less similar across all languages, although there are still different sets of sounds between the major language families. Furthermore, (ii) many analyses of the structure of spoken and then written language suggest that there are ‘basic structures’ (‘syntax’) that can be found—with variations—in all languages.

[12] Anyone who comes into contact with people who speak a different language can experience this up close and concretely.

[13] In light of modern brain research and general physiology, the ‘location’ here would be assumed to be the brain. However, this location is of little use, as the research into the material brain in the body, along with its interactions with the surrounding body, has hardly been able to grasp the exact mechanisms of ‘meaning assignments’ (just as one cannot identify the algorithm being executed based on the physical signals of computer chips alone from those signals).

[14] For example, ‘hunger’, ‘thirst’, ‘toothache’…

[15] The ‘remembering’ of something that has happened before; the ‘recognition’ of something sensually concrete that one can ‘remember’; the ‘combining’ of memorable things into new ‘constellations’, and much more.

[16] Parts of this discussion can be found in the context of ‘text analyses’, ‘text interpretations’, ‘hermeneutics’, ‘Bible interpretation’, etc.

[17] Which is not surprising at all, since modern formal logic could only arise because it had programmatically radically departed from what has to do with everyday linguistic meaning. What was left were only ‘stubs of an abstract truth’ in the form of abstract ‘truth values’ that were devoid of any meaning.

[18] His posthumously published ‘Philosophical Investigations’ (1953) therefore offer ‘only’ a collection of individual insights, but these were influential enough to impact reflections on linguistic meaning worldwide.

[19] The list of publications titled around the ‘meaning of language’ is exceedingly long. However, this does not change the fact that none of these publications satisfactorily solve the problem comprehensively. It is currently not foreseeable how a solution could emerge, as this would require the cooperation of many disciplines, which in current university operations are well distributed and separated into ‘existences of their own’.

[20] With chatGPT as an example.

[21] Millions of texts produced by humans for the purpose of ‘communicating content’.

[22] Which ultimately is true, because the algorithms themselves do not ‘invent’ text, but use ‘actually used’ linguistic expressions from existing texts to generate ‘highly probable’ combinations of expressions that humans would likely use.

[23] These cannot be seen in isolation: without extremely powerful computing centers along with corresponding global networks and social structures that make widespread use possible, these algorithms would be worthless. Here, indirectly, what has become possible and functions in everyday life due to collective human intelligence also shines through.

THE OKSIMO CASE as SUBJECT FOR PHILOSOPHY OF SCIENCE. Part 1

eJournal: uffmm.org
ISSN 2567-6458, 22.March – 23.March 2021
Email: info@uffmm.org
Author: Gerd Doeben-Henisch
Email: gerd@doeben-henisch.de

CONTEXT

This text is part of a philosophy of science  analysis of the case of the oksimo software (oksimo.com). A specification of the oksimo software from an engineering point of view can be found in four consecutive  posts dedicated to the HMI-Analysis for  this software.

THE OKSIMO EVENT SPACE

The characterization of the oksimo software paradigm starts with an informal characterization  of the oksimo software event space.

EVENT SPACE

An event space is a space which can be filled up by observable events fitting to the species-specific internal processed environment representations [1], [2] here called internal environments [ENVint]. Thus the same external environment [ENV] can be represented in the presence of  10 different species  in 10 different internal formats. Thus the expression ‘environment’ [ENV] is an abstract concept assuming an objective reality which is common to all living species but indeed it is processed by every species in a species-specific way.

In a human culture the usual point of view [ENVhum] is simultaneous with all the other points of views [ENVa] of all the other other species a.

In the ideal case it would be possible to translate all species-specific views ENVa into a symbolic representation which in turn could then be translated into the human point of view ENVhum. Then — in the ideal case — we could define the term environment [ENV] as the sum of all the different species-specific views translated in a human specific language: ∑ENVa = ENV.

But, because such a generalized view of the environment is until today not really possible by  practical reasons we will use here for the beginning only expressions related to the human specific point of view [ENVhum] using as language an ordinary language [L], here  the English language [LEN]. Every scientific language — e.g. the language of physics — is understood here as a sub language of the ordinary language.

EVENTS

An event [EV] within an event space [ENVa] is a change [X] which can be observed at least from the  members of that species [SP] a which is part of that environment ENV which enables  a species-specific event space [ENVa]. Possibly there can be other actors around in the environment ENV from different species with their specific event space [ENVa] where the content of the different event spaces  can possible   overlap with regard to  certain events.

A behavior is some observable movement of the body of some actor.

Changes X can be associated with certain behavior of certain actors or with non-actor conditions.

Thus when there are some human or non-human  actors in an environment which are moving than they show a behavior which can eventually be associated with some observable changes.

CHANGE

Besides being   associated with observable events in the (species specific) environment the expression  change is understood here as a kind of inner state in an actor which can compare past (stored) states Spast with an actual state SnowIf the past and actual state differ in some observable aspect Diff(Spast, Snow) ≠ 0, then there exists some change X, or Diff(Spast, Snow) = X. Usually the actor perceiving a change X will assume that this internal structure represents something external to the brain, but this must not necessarily be the case. It is of help if there are other human actors which confirm such a change perception although even this does not guarantee that there really is a  change occurring. In the real world it is possible that a whole group of human actors can have a wrong interpretation.

SYMBOLIC COMMUNICATION AND MEANING

It is a specialty of human actors — to some degree shared by other non-human biological actors — that they not only can built up internal representations ENVint of the reality external to the  brain (the body itself or the world beyond the body) which are mostly unconscious, partially conscious, but also they can built up structures of expressions of an internal language Lint which can be mimicked to a high degree by expressions in the body-external environment ENV called expressions of an ordinary language L.

For this to work one  has  to assume that there exists an internal mapping from internal representations ENVint into the expressions of the internal language   Lint as

meaning : ENVint <—> Lint.

and

speaking: Lint —> L

hearing: Lint <— L

Thus human actors can use their ordinary language L to activate internal encodings/ decodings with regard to the internal representations ENVint  gained so far. This is called here symbolic communication.

NO SPEECH ACTS

To classify the occurrences of symbolic expressions during a symbolic communication  is a nearly infinite undertaking. First impressions of the unsolvability of such a classification task can be gained if one reads the Philosophical Investigations of Ludwig Wittgenstein. [5] Later trials from different philosophers and scientists  — e.g. under the heading of speech acts [4] — can  not fully convince until today.

Instead of assuming here a complete scientific framework to classify  occurrences of symbolic expressions of an ordinary language L we will only look to some examples and discuss these.

KINDS OF EXPRESSIONS

In what follows we will look to some selected examples of symbolic expressions and discuss these.

(Decidable) Concrete Expressions [(D)CE]

It is assumed here that two human actors A and B  speaking the same ordinary language L  are capable in a concrete situation S to describe objects  OBJ and properties PROP of this situation in a way, that the hearer of a concrete expression E can decide whether the encoded meaning of that expression produced by the speaker is part of the observable situation S or not.

Thus, if A and B are together in a room with a wooden  white table and there is a enough light for an observation then   B can understand what A is saying if he states ‘There is a white wooden table.

To understand means here that both human actors are able to perceive the wooden white table as an object with properties, their brains will transform these external signals into internal neural signals forming an inner — not 1-to-1 — representation ENVint which can further be mapped by the learned meaning function into expressions of the inner language Lint and mapped further — by the speaker — into the external expressions of the learned ordinary language L and if the hearer can hear these spoken expressions he can translate the external expressions into the internal expressions which can be mapped onto the learned internal representations ENVint. In everyday situations there exists a high probability that the hearer then can respond with a spoken ‘Yes, that’s true’.

If this happens that some human actor is uttering a symbolic expression with regard to some observable property of the external environment  and the other human actor does respond with a confirmation then such an utterance is called here a decidable symbolic expression of the ordinary language L. In this case one can classify such an expression  as being true. Otherwise the expression  is classified as being not true.

The case of being not true is not a simple case. Being not true can mean: (i) it is actually simply not given; (ii) it is conceivable that the meaning could become true if the external situation would be  different; (iii) it is — in the light of the accessible knowledge — not conceivable that the meaning could become true in any situation; (iv) the meaning is to fuzzy to decided which case (i) – (iii) fits.

Cognitive Abstraction Processes

Before we talk about (Undecidable) Universal Expressions [(U)UE] it has to clarified that the internal mappings in a human actor are not only non-1-to-1 mappings but they are additionally automatic transformation processes of the kind that concrete perceptions of concrete environmental matters are automatically transformed by the brain into different kinds of states which are abstracted states using the concrete incoming signals as a  trigger either to start a new abstracted state or to modify an existing abstracted state. Given such abstracted states there exist a multitude of other neural processes to process these abstracted states further embedded  in numerous  different relationships.

Thus the assumed internal language Lint does not map the neural processes  which are processing the concrete events as such but the processed abstracted states! Language expressions as such can never be related directly to concrete material because this concrete material  has no direct  neural basis.  What works — completely unconsciously — is that the brain can detect that an actual neural pattern nn has some similarity with a  given abstracted structure NN  and that then this concrete pattern nn  is internally classified as an instance of NN. That means we can recognize that a perceived concrete matter nn is in ‘the light of’ our available (unconscious) knowledge an NN, but we cannot argue explicitly why. The decision has been processed automatically (unconsciously), but we can become aware of the result of this unconscious process.

Universal (Undecidable) Expressions [U(U)E]

Let us repeat the expression ‘There is a white wooden table‘ which has been used before as an example of a concrete decidable expression.

If one looks to the different parts of this expression then the partial expressions ‘white’, ‘wooden’, ‘table’ can be mapped by a learned meaning function φ into abstracted structures which are the result of internal processing. This means there can be countable infinite many concrete instances in the external environment ENV which can be understood as being white. The same holds for the expressions ‘wooden’ and ‘table’. Thus the expressions ‘white’, ‘wooden’, ‘table’ are all related to abstracted structures and therefor they have to be classified as universal expressions which as such are — strictly speaking —  not decidable because they can be true in many concrete situations with different concrete matters. Or take it otherwise: an expression with a meaning function φ pointing to an abstracted structure is asymmetric: one expression can be related to many different perceivable concrete matters but certain members of  a set of different perceived concrete matters can be related to one and the same abstracted structure on account of similarities based on properties embedded in the perceived concrete matter and being part of the abstracted structure.

In a cognitive point of view one can describe these matters such that the expression — like ‘table’ — which is pointing to a cognitive  abstracted structure ‘T’ includes a set of properties Π and every concrete perceived structure ‘t’ (caused e.g. by some concrete matter in our environment which we would classify as a ‘table’) must have a ‘certain amount’ of properties Π* that one can say that the properties  Π* are entailed in the set of properties Π of the abstracted structure T, thus Π* ⊆ Π. In what circumstances some speaker-hearer will say that something perceived concrete ‘is’ a table or ‘is not’ a table will depend from the learning history of this speaker-hearer. A child in the beginning of learning a language L can perhaps call something   a ‘chair’ and the parents will correct the child and will perhaps  say ‘no, this is table’.

Thus the expression ‘There is a white wooden table‘ as such is not true or false because it is not clear which set of concrete perceptions shall be derived from the possible internal meaning mappings, but if a concrete situation S is given with a concrete object with concrete properties then a speaker can ‘translate’ his/ her concrete perceptions with his learned meaning function φ into a composed expression using universal expressions.  In such a situation where the speaker is  part of  the real situation S he/ she  can recognize that the given situation is an  instance of the abstracted structures encoded in the used expression. And recognizing this being an instance interprets the universal expression in a way  that makes the universal expression fitting to a real given situation. And thereby the universal expression is transformed by interpretation with φ into a concrete decidable expression.

SUMMING UP

Thus the decisive moment of turning undecidable universal expressions U(U)E into decidable concrete expressions (D)CE is a human actor A behaving as a speaker-hearer of the used  language L. Without a speaker-hearer every universal expressions is undefined and neither true nor false.

makedecidable :  S x Ahum x E —> E x {true, false}

This reads as follows: If you want to know whether an expression E is concrete and as being concrete is  ‘true’ or ‘false’ then ask  a human actor Ahum which is part of a concrete situation S and the human actor shall  answer whether the expression E can be interpreted such that E can be classified being either ‘true’ or ‘false’.

The function ‘makedecidable()’ is therefore  the description (like a ‘recipe’) of a real process in the real world with real actors. The important factors in this description are the meaning functions inside the participating human actors. Although it is not possible to describe these meaning functions directly one can check their behavior and one can define an abstract model which describes the observable behavior of speaker-hearer of the language L. This is an empirical model and represents the typical case of behavioral models used in psychology, biology, sociology etc.

SOURCES

[1] Jakob Johann Freiherr von Uexküll (German: [ˈʏkskʏl])(1864 – 1944) https://en.wikipedia.org/wiki/Jakob_Johann_von_Uexk%C3%BCll

[2] Jakob von Uexküll, 1909, Umwelt und Innenwelt der Tiere. Berlin: J. Springer. (Download: https://ia802708.us.archive.org/13/items/umweltundinnenwe00uexk/umweltundinnenwe00uexk.pdf )

[3] Wikipedia EN, Speech acts: https://en.wikipedia.org/wiki/Speech_act

[4] Ludwig Josef Johann Wittgenstein ( 1889 – 1951): https://en.wikipedia.org/wiki/Ludwig_Wittgenstein

[5] Ludwig Wittgenstein, 1953: Philosophische Untersuchungen [PU], 1953: Philosophical Investigations [PI], translated by G. E. M. Anscombe /* For more details see: https://en.wikipedia.org/wiki/Philosophical_Investigations */