Generative AI and the kenosis of humanity

We begin with some diachronic considerations concerning thought and language. For brevity's sake we will only distinguish between external language, linguistic utterances externalized in the context of communication or record-keeping and internal language, the inner spiritual substrate (including habitual memory and inner discourse) from which external language proceeds. And let us consider the following simplistic model. A given time $t$ the is a population $P(t)$ of $n$ speakers, for every $p,q \in P(t)$ there is $L(p,q)(t)$, the set of linguistic utterances uttered by $p$ to $q$ and a set of texts $L(t)$ (each authored by some element in $P(t)$) which are public and accessible.  We note that normally, $L(p,q)(t)$ is tied to context, dialogue, questioning, spiritual-mental states, relationships, goals, to the specific community to which $p$ and $q$ may belong - all of which we can consider as taking place in a microlocal time the moments of which are not distinguished by the global time $t$. 

For a given macro-time t, the total possible linguistic content $TPLC$ of $P(t)$ has to determined in terms of the collection of the externalizations and recordings in all possible hypothetical communicative scenarios between elements of $P(t)$.  For instance there are questions which could have been posed to individual a in time t which would have given rise to utterances of great value, deemed worthy of recording, but this scenario historically never in fact happened. We can very roughly think of $TPLC$ as being parametrized by different graphs over $P(t)$ with marked vertex "$a$" representing the social-context for the recorded utterance of $a$. We can think of elements of $P(t)$ as containers which as a rule only pour parts of their contents when interacting with other containers or group of containers. 

Now we can describe the result of the socio-technological transformation which lead to the world wide web and from the world wide web to generative AI (and from generative AI to future totalitarian agentic monitorization and simulation). Generative AI seeks to reach a state P(s) and which $TPLC$ has already been recorded and in which the possible communicative contexts themselves have been automated and made public. Generative AI seeks to effect a universal kenosis (emptying) of the human inner-spiritual linguistic substrate by rendering such a spiritual substrate and all the human social connections and contexts irrelevant and unnecessary.  Returning to the vessel metaphor, generative AI seeks to empty all vessels as to how they empty in every one of their possible interactions and mash together the content into a public monstrous idolatric simulacrum. The essential function of the world wide web was precisely to get people to "empty" themselves onto the great heap of the internet so that such a heap could be then "mashed" together by generative AI. 

Such a kenosis is however illusory and doomed to failure for the following reasons. Behind this thinking is the fallacy that meaning of a linguistic utterance can be precisely determined and exhausted by the extensional sum of all its possible occurrence windows and communicative contexts. This ignores the inexhaustible nuanced connotational dynamic multi-leveled intensional nature of meaning and its ultimate inseparability from the inner spiritual life, past experience and personal relationships of the subject. The contexts in question cannot be captured or formalized on the superficial level in which generative AI operates (the same goes for the earlier shallow encyclopedias and question-forums of the world wide web). Furthermore, the public availability of the generative AI version of $TPLC$ will itself contaminate its own training process in a way recalling diagonalization and self-reference arguments. We can consider prompts like: what would person $X$ think of subject $Y$ if person $X$ interacted with LLM $Z$ in way $W$ for length of time $T$? If LLMs are treated as oracles then they must answer questions which involve oracles. 

Generative AI will effect a kenosis of all the spiritual value and meaning of human language for human spiritual life and culture, not to mention its communicative value and formal objectivity. Generative AI will succeed in making language itself the enemy of mankind robbed of its spiritual, expressive, liberational and higher epistemic function - and historical-cultural continuity with the past.  Equally drastic counter-measures are called for, the development of radically alternative languages (including formal languages and logics) and symbolic modes (maybe including those based on silence and gestures) of thinking, communication and recording which serve and preserve intact the infinite spiritual content of human meaning, experience and social communion.  We note that there are aspects of human experience which can carry profound spiritual and semantic content and connection related to olfactory, gustatory, tactile, proprio-perceptive senses and inner modes of vision and illumination which will remain forever beyond the realm of generative AI. And the experience of a romantic connection, of falling in love, though it can occur in the context of audio-visual data,  cannot be reduced to it. Make philosophical logic and the philosophy of language together with the theory of literature and literary stylistic analysis the core of human education. Encourage multilingualism and metaphoric poetic writing and the appreciation of literary genius. Increase the subtlety, nuances, subtext, connotations of text and metaphoric symbolic expression. Revive symbolist poetry and ancient languages. Develop irreducibly personal, private or intimate languages as forms of resistance (there are some interesting historical precursors).  Demolish arguments against a private language and demolish behaviorism. Use pure mathematical theories as a form of symbolic metaphoric personal private communication. Develop entirely new mathematical theories, puzzles, formal systems and games which have not yet been "extensionalized" in the web to be used by LLMs.  Develop a new system of encryption in which all communications and records can only be decrypted by small groups of people sharing the "private group key". Develop context-dependent programming languages which are intractable by LLMs.