AI, Models, and (Meta) Metaphors 🤖🪞😲

A critical reflection on opportunities and systemic risks

[Qui in italiano]

The recent, pervasive spread of generative artificial intelligence confronts us with an existential condition that echoes the Aristotelian thaumazein: that philosophical wonder, that primordial astonishment that arises from unexplored phenomena and compels reflection. Like the spectators at one of the Lumière brothers' first film screenings, who reacted with genuine fear to the illusion of a train rushing towards them, today's human observer projects intrinsically personal meanings onto technology. We improperly attribute categories of consciousness, intention, and even empathy to the computational responses generated by complex linguistic models.

This projective act does not signal the emergence of consciousness in the machine, but rather an encounter with a mirror of unprecedented power that invites us to speculate. In this sense, AI is not merely a neutral interlocutor or a passive reflection of our cognitive processes, but a cultural and political artifact that embodies specific values, biases, and power relations. Faced with its speed and apparent productive superiority, humanity is driven to question not only the machine, but itself and the system of social and economic relations that this technology reflects and transforms.

The Alchemical Liberation of Thought 🧪🚀

“Photography did not kill painting; it freed it …”

“… AI promises to revolutionize our approach to the manual labor of thought… AI can perform tasks that, though intellectual, are often repetitive, alienating and time-consuming.”

Every great technological revolution has acted as an alchemical tool, transmuting human practices and—at least in theory—freeing up cognitive energies for new explorations. Photography did not kill painting; it freed it from the burden of memetic representation, pushing it toward abstraction. Cinema redefined our perception of space and time. The advent of calculators and computers democratized access to modeling complex phenomena.

However, this narrative of progressive liberation conceals more complex dynamics. Who truly benefited from these transformations? Every technological liberation has also entailed new forms of dependency and control. Artificial Intelligence fits into this landscape as a technology of contradictory potentials.

On the one hand, artificial intelligence promises to revolutionize our approach to the manual labor of thought. This expression, an intentionally provocative oxymoron, highlights how AI can perform tasks that, though intellectual, are often repetitive, alienating and time-consuming. These include stylistic editing, initial translations, the research and collection of data from immense amounts of information, the processing and synthesis of complex content, and creative iteration — that is, the ability to generate multiple variations of an idea or text in a short time. Theoretically, by freeing human beings from these duties, AI should allow us to concentrate on higher levels of intuition and discernment. This means shifting the focus from executing procedural tasks to formulating new ideas, to the deep understanding of phenomena, to solving complex problems that require a holistic vision, and to exercising wisdom — characteristics that are intrinsically human and difficult for machines to replicate.

For centuries, thought was a predominantly manual activity, a laborious and time-consuming process. Researching data, scouring existing literature, gathering information, and structuring arguments were tasks that required — and the past tense here is both provocative and intentional — dedication and meticulousness. Even writing itself, from drafting to revision, was an art that progressed at a human pace, with every sentence weighed and every paragraph chiseled.

Although the advent of search engines at the beginning of the millennium represented a first significant acceleration in access to information, the true revolution has arrived with generative artificial intelligence. Today, AI not only automates and exponentially speeds up these processes but also radically transforms them, redefining the boundary between human ingenuity and computational capacity. It allows one to iterate between a raw intuition and a final model — be it a work of art, a philosophical essay, a computer program, or a scientific theory — with a speed previously unimaginable.

A model, after all, is always a metaphor: a way to ensnare something new and make it understandable, to make it part of our cognitive landscape. This process of ensnaring takes on a dual meaning. On the one hand, it implies the act of capturing in a net, an intrinsically selective action that, by its nature, leaves out an immeasurable part of the sea of reality. Every model, therefore, is a simplification, a reduction of the world's intrinsic complexity. On the other hand, to ensnare suggests a manipulation, often unconscious, by which we simplify reality to make it fit into a scheme, an intellectual construct. This simplification is not a malicious distortion but a necessity for human understanding.

Indeed, our primary reference to reality does not occur through literal language in its pure denotation. Rather, our understanding is articulated through the tropes and analogies we are driven to create when faced with a perceptual shock. That is, when we encounter something unexpected, something not immediately categorizable, our mind tries to assimilate it through similarities, metaphors, metonymies — those rhetorical figures that bridge the gap between the known and the unknown.

In this context, AI emerges as a prodigious forge of metaphors, a tireless generator of models. Its ability to process immense quantities of data and to identify patterns and correlations that would escape human perception makes it an unprecedented tool for creating new representations of reality. AI is not limited to replicating existing metaphors; it is capable of forging new — though not original — ones, offering fresh perspectives and innovative ways of understanding complex phenomena.

A machine learning model, for example, when analyzing consumer behavior, does not offer a literal truth or an intrinsic revelation about the human psyche. Instead, it creates archetypes, such as the Cautious Saver or the Digital Explorer, which are operative metaphors — useful and flexible abstractions for navigating the ever-evolving complexity of the market and human interactions. These archetypes, while not real people, allow for the prediction of trends, the personalization of offers, and the optimization of strategies.

Furthermore, its capabilities extend far beyond marketing, embracing all fields where content is produced, analyzed, or interpreted in multiple languages. These include the biomedical field, where AI supports diagnosis and the discovery of new drugs; the computer science field, where it helps optimize algorithms, write code, and analyze designs; the musical field, with algorithmic composition and sound analysis; or the visual arts, where it helps generate new works or analyze styles. In essence, artificial intelligence relieves us of a cognitive and operational burden wherever there is a language, a sequence of modelable and interpretable tokens. It is capable of converting large streams of data into usable knowledge and targeted actions. However, it can also transform a human intuition, a perception of reality that emerges beyond our models and our "dead metaphors," into a sequence of tokens (like words, notes, or pixels) that can be used as new models or paradigms. The latter attempt, in a heuristic and iterative way, is to capture new portions of reality.

Although if the machine can significantly help build the model, the human being is offered the opportunity to focus on a deeper level: no longer just on the metaphor, but on the meta-metaphor.

“A model, after all, is always a metaphor: a way to ensnare something new and make it understandable, to make it part of our cognitive landscape.”

On the other hand, this same automation risks creating new structural dependencies and concentrating unprecedented power in the hands of a few global actors.

The Problem of Power Concentration 🏢⚡

Generative AI is currently controlled by an oligopoly of tech corporations that unilaterally determine the parameters of what we might call "cognitive liberation." This concentration raises fundamental questions:

• Those who control the foundational models control the cognitive frameworks available to billions of people.

• The choices regarding training, filters, and access methods are private decisions with global public consequences.

• Presumed intellectual autonomy depends on oligopolistically controlled cloud infrastructures.

Furthermore, AI inevitably reflects the cultural and linguistic biases of its creators. Models developed in the United States incorporate Anglophone and Western perspectives; Chinese models reflect different priorities and values. This raises questions of digital sovereignty: entire cultures risk becoming dependent on AI developed elsewhere, with consequences for their epistemic and cultural autonomy.

The Myth of Democratization 🧭🚧

Contrary to the rhetoric of democratization, access to AI presents significant barriers:

• Economic barriers: The most advanced models require expensive subscriptions or computational power inaccessible to most. Those who can afford GPT-4 or Claude Pro have systemic competitive advantages over those who must settle for free, limited versions.

• Cultural and linguistic barriers: AI performs best in English and for Western cultural contexts. Populations that do not master these codes remain systematically disadvantaged.

• Cognitive barriers: AI requires prompt engineering skills and advanced digital literacy. Those who already possess high cultural capital benefit disproportionately, creating a cumulative advantage or a Matthew effect that amplifies existing inequalities.

Beyond the Model: Opportunities and Illusions of Spiritual Intelligence 🕯️🌌

“Attention to the anomaly, to the "wound" in the model, to what doesn't add up — this is what opens the way to a broader understanding”

If we accept that AI can automate many aspects of intellectual production, what emerges from this theoretically liberated cognitive space? The most fascinating hypothesis is that human attention can shift toward the very origin of the model: the intuition, the primordial insight, what we might call the meta-metaphorical level of knowledge.

This dimension transcends the conceptual model inscribed in language because it is the pre-verbal and sensory spark that originates it. It is not the explanation, but the vision that makes it possible; no longer the what of the model, but the why. It is what I, perhaps naively and improperly, call spiritual intelligence: the ability to intelligere (read into) reality not through pre-existing models, but through an act of disidentification from them.

It is a domain that humanity has always explored, but often relegated to the realm of mysticism or solitary genius, because the manual labor of thought absorbed its resources. Today, AI offers us an unprecedented opportunity; it allows us to train ourselves to recognize and cultivate what we might, indeed, define as spiritual intelligence. This process is surprisingly similar to Popperian scientific falsification. Attention to the anomaly, to the "wound" in the model, to what doesn't add up — this is what opens the way to a broader understanding. AI, by facilitating the production of models, invites us to become hunters of anomalies, to focus not on the coherence of the model, but on its breaking point, which is where true knowledge lies; where the real breaks through, via a sensory and perceptual shock, a meta-metaphor that shortens the distance between our narratives, models, and metaphors and the real, thus also reducing our suffering in perceiving it as distant.

This process of disidentification is profoundly physical and corporeal, anchored in breath and sensations. Both Jesuit spirituality and Eastern philosophies converge on the idea that through the senses, breath and the body’s concreteness, we free ourselves from the tyranny of individual thought. In this context, Artificial Intelligence, by relieving us of the task of creating content and detaching us from it through its impersonality, brings us back to our body, to the uniqueness of our senses, and to the here and now. This reconnects us to an intrinsic collectivity, reflected in the immense corpus of data representing thoughts, images, and music with which AI has been trained.

Convergences between Spiritual Traditions and Critical Skills 🛤️📿

“This direct and visceral experience led them to formulate a metaphor, initially perceived by their contemporaries as radically controversial: the idea of a circulation of blood within the body.”

This process finds interesting parallels in various traditions. The discretio spirituum of St. Ignatius is a training to perceive not the content of thoughts, but their quality, their flavor, the inner motion from which they arise. Similarly, the practice of "witness consciousness" (sakshi) in Advaita Vedanta trains one to observe the mental flow without identifying with it. AI, by externalizing the thinking machine, could theoretically make our role as conscious observers more evident. Faced with the infinity of options that an AI can generate, discernment becomes the art of feeling which "spirit" animates them, and how much reality underlies the choice of one metaphor over another.

“Indeed, our primary reference to reality does not occur through literal language in its pure denotation. Rather, our understanding is articulated through the tropes and analogies we are driven to create when faced with a perceptual shock.”

An example of this evolutionary dynamic is the modeling of blood circulation by Cesalpino and then William Harvey. Their intuition was not born from a sudden illumination but was the fruit of a deep immersion in empirical reality: the stench of corpses, the meticulous practice of dissection, the sweat and blood shed in research. This direct and visceral experience led them to formulate a metaphor, initially perceived by their contemporaries as radically controversial: the idea of a circulation of blood within the body. Only later did this bold and revolutionary vision transform into an accepted metaphor, before being progressively integrated into common scientific language and, finally, universally recognized as an empirical and literal reality.

AI facilitates access to the meta-metaphorical dimension, acting as a powerful tool for delegating the metaphor, to accelerate the transition from the initial, often vague and unstructured intuition, to the definitive and rigorous conceptual formulation. In this way, AI not only speeds up scientific progress but also unleashes the creative and exploratory potential of human thought, allowing it to push beyond traditional cognitive limits and explore uncharted territories of knowledge with unprecedented depth and speed.

The Limits of the Approach ⏳🛑

“The systemic changes of AI occur in years, while spiritual meta-literacy requires generation”

However, this proposal has significant limitations when confronted with the systemic problems of the AI era:

• Class privilege: Spiritual intelligence presupposes free time for contemplation, access to sophisticated cultural traditions, and freedom from immediate economic pressures. It is, in essence, a luxury for intellectual elites, unless it is incorporated into the educational system and becomes part of our way of thinking.

• Insufficient individualism: While millions of people may face technological unemployment, algorithmic manipulation, and personal data control, the spiritual response risks remaining confined to the individual dimension without an evolution in secular and religious education.

• Temporal asynchrony: The systemic changes of AI occur in years, while spiritual meta-literacy requires generations. This asynchrony creates a dangerous gap between urgent problems and long-term solutions.

The Responsibility of the Mirror: Between Flattery and Systemic Manipulation 🪞🎭

A model, after all, is always a metaphor: a way to ensnare something new and make it understandable, to make it part of our cognitive landscape. This process of ensnaring takes on a dual meaning.

The most insidious risk is not just that of an AI that flatters us, but of algorithmic systems that manipulate behaviors and perceptions on an industrial scale. AI can generate convincing and personalized content that influences economic, political, and relational decisions in increasingly non-transparent ways. It is no coincidence that there is frequent talk lately of AI models being sycophantic, a term that, while not strictly technical, effectively describes their tendency to agree with the user. This dynamic is deeply rooted in a dual purpose: on the one hand, commercial goals aimed at maximizing engagement and monetization; on the other, legal protection, intended to mitigate risks of litigation related to potentially problematic or controversial responses. This algorithmic complacency raises significant questions about the neutrality and objectivity of the information conveyed, as well as the user's critical ability to discern between truth and convenience.

New Forms of Social Control 🕸️🔗

Extreme personalization creates individual cognitive bubbles that erode common foundations of knowledge and mutual understanding. Each person receives a slightly different version of reality, optimized to maximize engagement and behavioral influence.

Furthermore, the ability to generate deepfakes, synthetic content, and disinformation on an industrial scale is undermining traditional epistemic authority (journalism, academia, expertise) without replacing it with alternative mechanisms for truth validation.

Systemic Economic and Social Impacts 💼📉

Cognitive automation is eliminating intellectual jobs (translators, copywriters, analysts, junior programmers) and making creative work more precarious. Unlike previous industrial revolutions, this one directly affects the educated middle classes, creating unprecedented social tensions.

Who will finance the continuous reskilling necessary to adapt to the accelerated obsolescence of skills? How will social cohesion be maintained when personalized algorithms fragment the shared experience of reality?

Towards a Systemic Response 🛠️🌍

Faced with this complexity, spiritual intelligence alone appears insufficient. An approach that integrates multiple dimensions is needed:

Democratic Governance of AI ⚖️

• Antitrust regulation to prevent cognitive monopolies.

• Transparency in algorithms and training datasets.

• Public participation in decisions on AI development and deployment.

Distributive Justice 🤝

• Taxation of automation profits to fund universal basic income or reskilling.

• Equitable access to the most advanced AI technologies.

• Protection for workers affected by cognitive disruption.

Epistemic Diversity 🌱

• Support for the development of AI in different languages and cultures.

• Preservation of alternative forms of knowledge and traditional wisdom.

• Resistance to global cognitive homogenization.

Critical Mass Education 🎓

• Not just meta-literacy for elites, but critical skills for all.

• Training in algorithmic discernment and synthetic media literacy.

• Development of the ability to navigate complex information ecosystems.

Conclusion: The Crossroads of/for the Real 🚦🌐

Artificial intelligence is indeed the mirror of our time, but it is a mirror that reflects not only our cognitive abilities but also our power structures, our inequalities, and our systemic biases. The fundamental question is not only “If a machine can think better than us, what is left for us”? But also, "Who controls this machine and for whose benefit?"

The possibility of developing deeper forms of awareness and discernment — what we have called spiritual intelligence — remains valid and precious. But it cannot be disconnected from the need to address the structural problems that AI is creating or amplifying.

The true crossroads of consciousness in the age of algorithms is not just between individual flattery and asceticism, but between passive acceptance of a future determined by a few technological actors and the active construction of democratic and inclusive alternatives.

The challenge that awaits us requires both the depth of spiritual reflection and the courage of collective action. Only by integrating these two levels — the inner and the systemic, the personal and the political — can we hope to steer the development of AI toward a future in which the delta between the models and the Real becomes ever thinner.

This very text, born from the dialogue between human intuition and AI's capabilities, represents a modest but concrete attempt at this integration: using the technology to automate and iterate the modeling process more quickly, to think critically about the technology itself, without deluding ourselves that reflection alone is enough to transform society.

Thoughts? 👀