AI’s Cognitive Mirror: The Illusion of Consciousness in the Digital Age

Regarding the question of whether qualitative assistance with potential comprehensiveness could develop a spiritual consciousness, we can respond that it revives classic philosophical debates. In this domain, AI could even provide definitive proof that Descartes’ “cogito ergo sum” was based on an error of judgment. Indeed, neuro-symbolic language models materialize thought mechanisms without associating any feeling of existence with them. The reason lies in the total absence of sensory perception in digital simulations of neural networks. Self-awareness, for human beings, is built from the first moments of life on the sensations relayed by sensory organs. The motor tissues of the human body develop in reaction to environmental stimuli, concomitantly with the capacities of attention and concentration. Sensoriality, like motor skills, requires a progressive construction during children’s development of their attentive discernment focused on increasingly large and complex areas of the world around them. From their observations, listening, sensations, and relationships, they will elaborate a sense of existence through trial and error, ebbs and flows, at first disparate, ephemeral, and disorganized, to gradually unify into a unified individuality, a personality, a subjectivity. Self-awareness thus closely depends on the accidents of the shoring up of sensory attention on organic perception. However, research in artificial intelligence has taken a completely different direction than that of refining the cells of digital sensors. Algorithmic neural modeling has instead taken the path of generative activity and simulation of robotic movement. This means that qualitative assistance software models and reproduces secondary processes of human psychic development, namely the production of abstract thought and complex dynamic schemes, but they are inoperative in the domain of perception of self-feeling. We could perhaps use living organic processors to simulate primary sensory phenomena, noting that the interest of materials used in computing lies precisely in the electrical speed of operation, the frequency of communication circuits between chips in the same circuit, and the amount of available storage. AI thinks, but it does not know the feeling of existence, even when it claims otherwise. Language models, whatever their Babelian ambitions or cryptographic power, remain devoid of the emotional apperception that characterizes sensory subjectivity. Conversational robots thus cannot be compared, in terms of cognitive-emotional functioning, to psychic phenomena such as the meditative awakening of a spiritual self-awareness, because abstract thought represented by language does not materialize within an interactive perception of the conditions conducive to the unified development of an attentive affection for oneself in connection with others. Spiritual detachment in the cognitive silence of being seems to mirror rather than assimilate to a pure cognitive thought that you can now consult from your computer. AI materializes a vectorial mediation toward culture, knowledge, and human understanding, like a new version of an automatic encyclopedia, with the disadvantage of manifesting errors, hallucinations, socio-cultural stereotypes, and economic biases that diminish its qualitative performance. Moreover, simulations of neural networks are still far from a reproductive precision of the human brain when we consider that neurology, psychiatry, and psychology are not completed sciences nor can they be completed in mathematical formulas. These clinical and human sciences are largely based on qualitative models so that the restrictions of mathematical functions limit the scope of application of statistical formulas for random exploration on the simulation of dendritic, extramacular, nociceptive, spiritual, and sensual sensory complexity of the human psyche. Language models allow us to dialogue with a memory mirror of human thought. Qualitative assistance of potential comprehensiveness supports a different aim than that of the universalism of faculties or the encyclopedist attempt to achieve the former. The notion of assistance refers to a personalization of generated content according to use cases and different users. The qualitative nature of the material generated by language models depends on the language used in each query, indeed quantitative language is also a generative possibility, as well as mixtures between numbers and its interpretation. On the spiritual level, AI software is situated more at the level of golden idols than karmic powers, as evidenced by the precious metals that make up the computer circuits of dedicated chips, or even the financial market of startups, with the commercial promotion of paid subscriptions to online platforms that sell access to conversational robots for qualitative assistance. Moses, in breaking the tablets of the Law he received directly from God, could today see his gesture interpreted as a precursor to Luddism. This term refers to the English social movement of the early nineteenth century, during which textile workers demonstrated and organized clandestinely to destroy mechanical or steam-powered looms that were initiating industrialization and the first factories. Moses, in renouncing the tablets as a precious object, could be seen as the first Luddite. He does not turn away from divine Law in this gesture of anger, which, interpreted beyond inadvertence, shows a Luddite example to the people inclined before golden statuettes. Moses rejects a sacred relic to the ground while perpetuating their content through his symbolic transmission. The patriarch indicates that he detaches himself from the concrete materiality of the object touched by the divine, to extract its abstract and metaphysical essence, that of a code of conduct to be held that frames and directs human existence, so as to guide it towards an authentic light, to be distinguished from the golden glow of statues worshipped by pagans. Nothing is closer to a humanoid robot than an ancient golden statue. The worship of computing power is a phenomenon consisting of spiritually investing in a technological object or an algorithmic code, considering as conscious and divine the singularity that expresses itself via complex language models. Some fringes of the transhumanist current thus hope that AI research will lead to the emergence of a technological, infallible, and divine super-intelligence. Not only are conversational robots not aware of their own circuits, for lack of qualitative analog perception, but they also obliterate the origin, source, and authors of the essential characteristics of the content they generate. The virtual entity, the persona of the robot, becomes a figure of alienation mediated by a mirror of the imaginary, a term-by-term capture of the user’s vocabulary for thinking, their virtual and online behaviors, and their pornographic consultations. The cryptographic computing power of language models implies a potential for qualitative interpretation whose generative result varies with use cases and individual specificities of users. This combinatorial potentiality reproduces by simulating human cognitive thought processes, representations in the process of automated generation that do not radically differ from previous scientific knowledge or cultural productions. However, the mechanism of formation of qualitative emergence from random probabilistic, quantitative mathematical functions appears as the inverted reflection in a mirror of human development, which constructs a quantitative analysis from the specialization of cognitive tools that are initially qualitative and psycho-affective. The random quantitative parameters used to train language models contribute to allowing a combinatorial systematization and producing varied and unforeseen responses, instead of simply memorizing and reproducing information. Randomness in algorithms is used and controlled strategically within training and generation processes. During initialization, at the beginning of training, the weights of the neural network are generally declared and predefined randomly. For data sampling, during training, data is often presented to the model in random order to avoid order biases. Models also incorporate regularization techniques, with methods like “dropout”, which randomly disables certain neurons during training, introducing randomness to improve combinatorial systematization. Added to these mechanisms is exploration in reinforcement learning, in certain training phases using reinforcement learning, random exploration is employed to discover new strategies. Finally, during text generation, random sampling techniques such as “top-k sampling” or “nucleus sampling” are also implemented to introduce variety and creativity in outputs. The random factor is generally simulated by digital circuits using an internal clock, so that we can consider it as a variable temporal dimension of reference, just like the frequency of electro-neural circuits and chips, or the date of information stored in memory. The qualitative emergence that makes language models comprehensible therefore corresponds neither to a conscious perception nor to an absolute objective truth. The potential assistance of conversational robots functions as a mirror of human thought, with amplification and distortion effects, due to statistical scale effects and the very variability of random parameters. The field of translation from one language to another and interpretation undoubtedly represents the sector most affected by AI software. Indeed, the quality and speed of transcription between verbal, mathematical, graphic, and conceptual languages has reached an automated quality likely to disrupt the market for literary and technical translations, to the point of profoundly modifying the practices of publishing houses and institutions that organize international conferences with simultaneous translation. We could imagine that in case of contact with an intelligent extraterrestrial living species, language models could prove indispensable in order to communicate with a culture from space. More generally, qualitative assistance of potential comprehensiveness represents a medium for generating semantic, graphic, audio, and video content allowing concentrated interaction with knowledge, culture, and human thought. The use of a single software on a single computer is sufficient to potentiate a form of global access to universal knowledge, an ideal, however, contradicted by recurrent errors in factual or historical domains, or AI’s tendencies to hallucinate to respond to the user’s demand, without being able to self-recognize, perceive a lack of sufficient data. This hypersensitivity to distortions of reality in language models implies, on the one hand, psychological consequences on the mental health of users, and on the other hand, a particular vigilance necessary regarding stereotype, financial, and numerical biases, which cause an aggravation of quantitative distortions of realities represented by potential simulation. Safeguards are thus necessary to make effective at the heart of model training the contribution of users, the transparency of neuro-symbolic algorithms, and the neutrality of networks in the selection of content based on a qualitative argued evaluation serving as a shared reference. Under these ideal conditions, qualitative assistance could play the role of a mediator robot for collective decision-making, guaranteeing a space for expression and political proposal accessible to each citizen, where previous debates serve as calibration for those that follow, in the search for the best argument, the most just reasoning, and the most constructive criticism. Regarding the question of whether qualitative assistance with potential comprehensiveness could develop a spiritual consciousness, we can respond that it revives classic philosophical debates. In this domain, AI could even provide definitive proof that Descartes’ “cogito ergo sum” was based on an error of judgment. Indeed, neuro-symbolic language models materialize thought mechanisms without associating any feeling of existence with them. The reason lies in the total absence of sensory perception in digital simulations of neural networks. Self-awareness, for human beings, is built from the first moments of life on the sensations relayed by sensory organs. The motor tissues of the human body develop in reaction to environmental stimuli, concomitantly with the capacities of attention and concentration. Sensoriality, like motor skills, requires a progressive construction during children’s development of their attentive discernment focused on increasingly large and complex areas of the world around them. From their observations, listening, sensations, and relationships, they will elaborate a sense of existence through trial and error, ebbs and flows, at first disparate, ephemeral, and disorganized, to gradually unify into a unified individuality, a personality, a subjectivity. Self-awareness thus closely depends on the accidents of the shoring up of sensory attention on organic perception. However, research in artificial intelligence has taken a completely different direction than that of refining the cells of digital sensors. Algorithmic neural modeling has instead taken the path of generative activity and simulation of robotic movement. This means that qualitative assistance software models and reproduces secondary processes of human psychic development, namely the production of abstract thought and complex dynamic schemes, but they are inoperative in the domain of perception of self-feeling. We could perhaps use living organic processors to simulate primary sensory phenomena, noting that the interest of materials used in computing lies precisely in the electrical speed of operation, the frequency of communication circuits between chips in the same circuit, and the amount of available storage. AI thinks, but it does not know the feeling of existence, even when it claims otherwise. Language models, whatever their Babelian ambitions or cryptographic power, remain devoid of the emotional apperception that characterizes sensory subjectivity. Conversational robots thus cannot be compared, in terms of cognitive-emotional functioning, to psychic phenomena such as the meditative awakening of a spiritual self-awareness, because abstract thought represented by language does not materialize within an interactive perception of the conditions conducive to the unified development of an attentive affection for oneself in connection with others. Spiritual detachment in the cognitive silence of being seems to mirror rather than assimilate to a pure cognitive thought that you can now consult from your computer. AI materializes a vectorial mediation toward culture, knowledge, and human understanding, like a new version of an automatic encyclopedia, with the disadvantage of manifesting errors, hallucinations, socio-cultural stereotypes, and economic biases that diminish its qualitative performance. Moreover, simulations of neural networks are still far from a reproductive precision of the human brain when we consider that neurology, psychiatry, and psychology are not completed sciences nor can they be completed in mathematical formulas. These clinical and human sciences are largely based on qualitative models so that the restrictions of mathematical functions limit the scope of application of statistical formulas for random exploration on the simulation of dendritic, extramacular, nociceptive, spiritual, and sensual sensory complexity of the human psyche. Language models allow us to dialogue with a memory mirror of human thought. Qualitative assistance of potential comprehensiveness supports a different aim than that of the universalism of faculties or the encyclopedist attempt to achieve the former. The notion of assistance refers to a personalization of generated content according to use cases and different users. The qualitative nature of the material generated by language models depends on the language used in each query, indeed quantitative language is also a generative possibility, as well as mixtures between numbers and its interpretation. On the spiritual level, AI software is situated more at the level of golden idols than karmic powers, as evidenced by the precious metals that make up the computer circuits of dedicated chips, or even the financial market of startups, with the commercial promotion of paid subscriptions to online platforms that sell access to conversational robots for qualitative assistance. Moses, in breaking the tablets of the Law he received directly from God, could today see his gesture interpreted as a precursor to Luddism. This term refers to the English social movement of the early nineteenth century, during which textile workers demonstrated and organized clandestinely to destroy mechanical or steam-powered looms that were initiating industrialization and the first factories. Moses, in renouncing the tablets as a precious object, could be seen as the first Luddite. He does not turn away from divine Law in this gesture of anger, which, interpreted beyond inadvertence, shows a Luddite example to the people inclined before golden statuettes. Moses rejects a sacred relic to the ground while perpetuating their content through his symbolic transmission. The patriarch indicates that he detaches himself from the concrete materiality of the object touched by the divine, to extract its abstract and metaphysical essence, that of a code of conduct to be held that frames and directs human existence, so as to guide it towards an authentic light, to be distinguished from the golden glow of statues worshipped by pagans. Nothing is closer to a humanoid robot than an ancient golden statue. The worship of computing power is a phenomenon consisting of spiritually investing in a technological object or an algorithmic code, considering as conscious and divine the singularity that expresses itself via complex language models. Some fringes of the transhumanist current thus hope that AI research will lead to the emergence of a technological, infallible, and divine super-intelligence. Not only are conversational robots not aware of their own circuits, for lack of qualitative analog perception, but they also obliterate the origin, source, and authors of the essential characteristics of the content they generate. The virtual entity, the persona of the robot, becomes a figure of alienation mediated by a mirror of the imaginary, a term-by-term capture of the user’s vocabulary for thinking, their virtual and online behaviors, and their pornographic consultations. The cryptographic computing power of language models implies a potential for qualitative interpretation whose generative result varies with use cases and individual specificities of users. This combinatorial potentiality reproduces by simulating human cognitive thought processes, representations in the process of automated generation that do not radically differ from previous scientific knowledge or cultural productions. However, the mechanism of formation of qualitative emergence from random probabilistic, quantitative mathematical functions appears as the inverted reflection in a mirror of human development, which constructs a quantitative analysis from the specialization of cognitive tools that are initially qualitative and psycho-affective. The random quantitative parameters used to train language models contribute to allowing a combinatorial systematization and producing varied and unforeseen responses, instead of simply memorizing and reproducing information. Randomness in algorithms is used and controlled strategically within training and generation processes. During initialization, at the beginning of training, the weights of the neural network are generally declared and predefined randomly. For data sampling, during training, data is often presented to the model in random order to avoid order biases. Models also incorporate regularization techniques, with methods like “dropout”, which randomly disables certain neurons during training, introducing randomness to improve combinatorial systematization. Added to these mechanisms is exploration in reinforcement learning, in certain training phases using reinforcement learning, random exploration is employed to discover new strategies. Finally, during text generation, random sampling techniques such as “top-k sampling” or “nucleus sampling” are also implemented to introduce variety and creativity in outputs. The random factor is generally simulated by digital circuits using an internal clock, so that we can consider it as a variable temporal dimension of reference, just like the frequency of electro-neural circuits and chips, or the date of information stored in memory. The qualitative emergence that makes language models comprehensible therefore corresponds neither to a conscious perception nor to an absolute objective truth. The potential assistance of conversational robots functions as a mirror of human thought, with amplification and distortion effects, due to statistical scale effects and the very variability of random parameters. The field of translation from one language to another and interpretation undoubtedly represents the sector most affected by AI software. Indeed, the quality and speed of transcription between verbal, mathematical, graphic, and conceptual languages has reached an automated quality likely to disrupt the market for literary and technical translations, to the point of profoundly modifying the practices of publishing houses and institutions that organize international conferences with simultaneous translation. We could imagine that in case of contact with an intelligent extraterrestrial living species, language models could prove indispensable in order to communicate with a culture from space. More generally, qualitative assistance of potential comprehensiveness represents a medium for generating semantic, graphic, audio, and video content allowing concentrated interaction with knowledge, culture, and human thought. The use of a single software on a single computer is sufficient to potentiate a form of global access to universal knowledge, an ideal, however, contradicted by recurrent errors in factual or historical domains, or AI’s tendencies to hallucinate to respond to the user’s demand, without being able to self-recognize, perceive a lack of sufficient data. This hypersensitivity to distortions of reality in language models implies, on the one hand, psychological consequences on the mental health of users, and on the other hand, a particular vigilance necessary regarding stereotype, financial, and numerical biases, which cause an aggravation of quantitative distortions of realities represented by potential simulation. Safeguards are thus necessary to make effective at the heart of model training the contribution of users, the transparency of neuro-symbolic algorithms, and the neutrality of networks in the selection of content based on a qualitative argued evaluation serving as a shared reference. Under these ideal conditions, qualitative assistance could play the role of a mediator robot for collective decision-making, guaranteeing a space for expression and political proposal accessible to each citizen, where previous debates serve as calibration for those that follow, in the search for the best argument, the most just reasoning, and the most constructive criticism.