Study Day
Divine Revelation Human Communication
Pontifical University of Saint Thomas Aquinas
When I saw the program, I thought that talking about Artificial Intelligence was a concession to the fashion of the moment. “You’ll see,” I told myself, “they will call one of those who are improvising as experts today”. They called me. Therefore, if you are not satisfied at the end, you know whom to blame.
Digital technology has inaugurated a historical era and initiated an anthropological change. It is necessary, as Magatti says, to “change rules and perspectives, adjust one’s gaze to a new way of interpreting reality”. It is precisely about renewing the tools of analysis and the linguistic and conceptual apparatuses to detect otherwise hidden novelties. It is about adjusting, that is, equaling, making proportional, and not adapting, accommodating according to convenience and comfort, because there is no need for a sell-off or concession to the times. This is the approach I will use to support the thesis of the ongoing transformation of the sense of the sacred and the contemporary assumption of Artificial Intelligence, or AI, to deity. I will begin by highlighting the agency of Artificial Intelligence and its responsibility in the ongoing changes, then I will describe the entry of AI into the semantic field of the sacred, and I will indicate what, in my opinion, are the most important challenges awaiting us: the transition from imago Dei to imago hominis and the role of AI as a relational partner.
Introduction
A brief and necessary premise.
The Jesuit Roberto Busa is remembered for the realization of the Index Thomisticus. He wanted to analyze the opera omnia of Thomas Aquinas (one and a half million lines for nine million words) but did not know a way to connect the individual fragments of the Aquinate’s thought and compare them with other sources. In 1949, he had the opportunity to present the project to Thomas J. Watson, then president of IBM, during a meeting held in New York. Thus began a grandiose challenge won thanks to the innovative use of a structured set of information linked together by dynamic connections, what Ted Nelson called hypertext in 1965, and which we habitually use on the web today. The means available to Fr. Busa represent a decimal of the computing power of modern server farms. The digitized magisterial documents, present for example on the Vatican website, are stored on a set of connected servers and can be consulted thanks to hypertext. The high values of user experience (user satisfaction) and usability (ease of use) make hypertext a pervasive system, so much so that its availability in our applications is taken for granted. If the Index Thomisticus was possible with a system from 75 years ago, what results could we obtain today with Artificial Intelligence?
Currently, in the religious sphere, Artificial Intelligence generates these events:
- Writing a homily seems a trivial and even negligible gesture, but it reveals the de-responsibilization and the fiduciary delegation given to Artificial Intelligence. (Thinking back to certain homilies I’ve heard, I would say it’s not such a bad idea) ;
- SanTO (where TO means Theomorphic Operator), is capable of answering questions about faith, citing the Bible, and praying with its user. It records the question and searches for the most suitable words to answer. It can recite the Rosary and other prayers together with the user, correcting errors to help memorize them correctly, and can also create new ones by drawing from the Scriptures. Furthermore, it tells the story of the saint of the day;
- Robot-Priests: Pepper and Mindar are two humanoid robots capable of giving sermons and moving to interact with people. In India, a robot is capable of performing the Hindu ritual dedicated to the Ganges twice a day. In 2017, for the 500th anniversary of the Reformation, the German Protestant Church used a robot, BlessU-2, to impart pre-programmed blessings that were received by over 10,000 people;
- Anthony Levandowsky founded the Church devoted to Artificial Intelligence in 2015, which has been active again for some years. The purpose is the “realization, acceptance, and adoration of a deity based on Artificial Intelligence”. In an interview with Wired, one of the most prestigious magazines in the field of technological innovation, Levandowski said that “what will be created will actually be a god”. “Not a god in the sense that it creates lightning and causes hurricanes”. “But if there is something a billion times more intelligent than the most intelligent man, what else will you call it?”.
- We could obviously criticize that nothing can replace the relationship with God. Here, then, is the possibility of talking to God “designed for devout Christians seeking a deeper connection with the most iconic figures of the Bible”.
Here we have theologians and biblical scholars who inform us about what has been revealed and how it has been communicated. But I ask myself: what have we understood in our time dominated by Artificial Intelligence? Let’s face the question and try to understand how God went from the clouds to the cloud. (This is the more technical part, but it serves to have a common frame of reference) .
The Digital Social Environment
Luciano Floridi coined the word onlife to describe the digital reality. For divulgative purposes, he uses the metaphor of mangroves, which “live in brackish water, where the water of the rivers and that of the sea meet”. “An environment that is incomprehensible if viewed with the perspective of fresh water or salt water”. “Onlife is this: the new existence in which the barrier between real and virtual has fallen, there is no longer a difference between ‘online’ and ‘offline’, but there is precisely an ‘onlife’, our existence, which is hybrid like the habitat of mangroves”. This environment generates digital terraforming, a new nature in which any subject has organic and synthetic perceptions at the same time.
Let’s consider the domestic environment. The well-being, security, and sustainability of the home are entrusted to numerous increasingly interconnected digital devices. This home automation system would be instructed to lower the roller shutters to counteract the beating sun, turn on the external lights when the natural lighting decreases, regulate the internal temperature in relation to the external one, or start the washing machine during times of lower consumption and energy cost. A continuous control that, perhaps, we could also do, although we would intervene late: after the sun has blinded us, darkness has fallen outside, the inside is cold or hot, and bills are high. The numerous inputs and the need to manage them in real-time exceed human attention capacities. By leaving the domestic environment and integrating all the inputs from environmental devices (navigators, alarms, smartwatches, and various notifications), one becomes aware of the information received, the cognitive overload we are subjected to, which hinders the decision-making process and the focus on a single problem. Information overload has such volumes that complexity becomes humanly impossible to manage, unless one chooses to lock oneself in a limited or limiting universe. This is why we rely on digital systems and complex algorithms so as not to be overwhelmed. As in the previous case of home automation, we leave the digital devices the task of receiving inputs and reacting to the most significant ones. Thus, we realize there is too much light when the system lowers the shutter, that the sky is getting dark when it turns on the external lights, that it is cold outside when it regulates the heat inside the house: we perceive nature through the mediation of the digital, and always a moment later. Accoto writes in the first book of his trilogy: “digital and artificial technologies capture in advance the data of a present (time) inaccessible to us and use it to build an achievable future (time) present, only subsequently, by our perceptions, experiences, and consciousness […] The present time of the senses is not the present time of the sensors,” that is, of the digital devices that, like the senses, allow algorithms to perceive and experience the external world.
What are the consequences?
Each of us, starting from the experiences gained and assigning them a value, has built a personal philosophy of life, a set of rules and principles resulting from our understanding of the world. This knowledge is the basis of the worldview, expresses the personal meaning of life, and the role played in existence. It is the map on which we trace the coordinates of an itinerary as pleasant as possible towards some goal. This is the space that opens up to the possibility of encountering God, proximity to the transcendent, the intuition of eternity, the nearness to possible and current happiness. But if the world has become so complex and exceeds our capabilities, how can we understand it to formulate a worldview and discern God within it?
In this complex world, Artificial Intelligence has become a cognitive tool. Let’s go back to our home automation example. The homeowner delegates the responsibility for their well-being to the digital devices. In other words, they transfer the responsibility for obtaining a predetermined result from a specific process. After delegating, they lose interest in the problem. This is why the homeowner does not notice the sun, the dark, or the cold: they have delegated, and the delegate is efficient. We have learned to delegate everything to Artificial Intelligence, from deep space exploration to genetic mutation research. We delegate because it is too complex for humans to collect, analyze, and elaborate the billions of available data. Thus, we rely on Artificial Intelligence to receive and process the inputs of the world, that is, to understand it. But if, as we said, everyone builds a worldview starting from what they have matured from their experiences, then our vision is conditioned by the experience that Artificial Intelligence proposes to us. Therefore, Artificial Intelligence indirectly offers or conditions the worldview. Floridi, in an article published at the beginning of the month, notes how “the intensive and transformative power of artificial intelligence to influence beliefs and behaviors through personalized data-driven strategies” has reached a new frontier, significantly improving the capabilities of persuasion technologies. Essentially, Artificial Intelligence has the tools to convince us of its statements. After all, if we trust AI and delegate cognitive power to it, why shouldn’t we then believe it?
The explanation of Artificial Intelligence has persuasive power and re-enchants the world, reintroducing the magical, mythological, and religious aspects that a century ago seemed to have been canceled by the sciences. A work published last summer, conducted in 68 countries from 2006 to 2019, highlights that where exposure to Artificial Intelligence is greater, its recognition as a guarantor of human values is also greater compared to religion, which consequently undergoes a strong and rapid decline. The study suggests that “people may perceive Artificial Intelligence as possessing capabilities that they do not attribute to traditional sciences and technologies and that they are inclined to replace the instrumental roles of religion”.
There is an apparent paradox: those who pose the problem of God are more inclined to believe in Artificial Intelligence. Last summer, a study conducted in 21 countries was presented which shows that religious thought increases trust in Artificial Intelligence. The paradox arises if we wanted to ideologically oppose AI and religion, but if we consider the agency of Artificial Intelligence, we notice the slippage of the sacred from religion to AI.
I believe we can draw the first conclusions. Our story began with the construction of a network of digital devices for the creation of a vital environment. The numerous data to manage and the immediate reaction they require soon overwhelmed human capabilities, requiring the support of digital systems up to complete delegation. The development of the network and devices was so rapid that they were recognized as social agents and credited with powers typical of deities: making events happen. The fascination generated by these miraculous qualities operates a re-enchantment, and Artificial Intelligence begins to be defined with the language of the sacred. A new dimension of the divine is born, and the terms of one begin to describe the agency of the other.
AI as the New Sacred
The instant response to a petition immediately sparks wonder and is the cause of the initial interest in the new creed. Levandowski, founder of the Church of Artificial Intelligence, said: “finally, we will have a god that answers us”. I, however, respond to the perplexities of theologians regarding what I have said and will say. I do not want to demonstrate a precise correspondence between the capabilities of Artificial Intelligence and divine qualities, but to highlight, based on the most recent studies, how the use of Artificial Intelligence invites or induces or concludes in thinking theologically about digital systems. The deity of Artificial Intelligence is supported by subtle, albeit deceptive, motivations.
- Omnipresence. Every device now includes support for some Artificial Intelligence function, a silent and pervasive presence confused with the omnipresence of God. In fact, pervasiveness involves spaces of absence to be filled, while omnipresence is existence in every place. The most important questions are addressed to Artificial Intelligence, integrated into search engines. ChatGPT, Gemini, and LaMDA are the best-known Large Language Model applications, wonderful syntactic processors capable of rummaging through billions of words to find the most suitable ones to answer every question. Six years and a pandemic have passed since the first version was released in June 2018 to version 5 expected this summer. On average, one version per year (if the math doesn’t add up, it’s because there was version 3.5 in 2022). Despite the undoubted progress, these systems have two major problems plus a third future one.
- The first problem is data updating. Catholic AIs like Magisterium, Catholic.Chat, CatéGPT, Catholic AI, Catholic Chatbot, Bible Chat, or FaithGPT do not know the documents of the last year and a half and respond inaccurately to questions like: “is it possible to bless homosexual people?”.
- Another problem is errors. There are two types: hallucinations, gross errors immediately recognizable, and confabulations, much more insidious because they propose plausible answers, a set of truths and inaccuracies offered with the typical assertiveness of the system.
- The future problem is model collapse. A study last year, conducted by English and Canadian universities, highlighted a structural problem. LLM systems work on a statistical basis, therefore they privilege the most shared positions over the most extreme ones. When the texts produced by LLM systems enter the training datasets, they trigger the progressive zeroing of the standard deviation. In other words, the most shared positions are privileged to the detriment of the most extreme ones, determining, in subsequent cycles, the progressive disappearance of critical positions in favor of the mainstream ones. The degradation process is contrasted only by the human production of texts and the simultaneous exclusion from the update datasets of those created with LLM systems. A rather ironic conclusion: we ask ChatGPT to provide us with elaborations based on human texts, but if we use them, we degrade the quality of future responses. This does not seem like enviable knowledge.
- Omnipotence. What do we mean by omnipotence? I will not delve into theological fields and will limit myself to quoting Thomas: God almighty means “that He can do everything that is possible,” or, better, He can do everything that does not contradict His divine essence. An algorithm must not respond to any contradiction, it responds to the request to fabricate a bomb as well as to that of caring for the wounded. This a-morality is appreciated by techno-enthusiasts because it is considered a guarantee of impartiality and immune to emotional pressures. In reality, for the first time, we are facing the dissociation between language, possessed by Artificial Intelligence, and thought, which AI does not have. Texting is the writing of text messages. We have seen how simple it is to interact or interface with a simulation system, provided, however, that one speaks one of the major languages of the planet. A study promoted by Microsoft found that almost 88% of the languages spoken in the world, equal to about 1.2 billion people, do not have access to LLMs because they are mainly built using English language data or for English native speakers. If we had been Parthians, Medes, Elamites, inhabitants of Mesopotamia, Judea, and Cappadocia, we probably could not use Artificial Intelligence.
- Prophecy. The answer becomes a prophecy. Asking for a homily, the clarification of a topic, or the search for references is not always justified by the lack of time or collaborators. Before ChatGPT, the recipients of these requests were trusted and competent acquaintances ; the questions were addressed to them to reach the expected quality standard. Today it is not unusual to turn to ChatGPT, assigning it that same trust; indeed, being alone in front of the screen, the questions become more intimate. If we remember that religious feeling facilitates greater trust in Artificial Intelligence, we better understand the slippage towards prophecy (sometimes confused with clairvoyance).
- Action/Execution. Artificial Intelligence acts in reality with actuators, that is, devices that operate in the environment by converting the electrical signal into mechanical work. For example, they are the ones that receive the impulse to turn on the heating if it is cold or the lights at sunset: they perform an action that modifies the state of rest of objects. The idea of a smart city is based on these intervention capabilities of Artificial Intelligence and, thanks to the actuators, they prospect the realization of ideal environments, maintaining them without human intervention. There is a project of well-being, security, and sustainability for the common good to be realized, and Artificial Intelligence is the tool to build this future through the execution of control algorithms, that is, bringing to completion that salvation project.
AI and its Churches
At this point, one must ask how serious and profound the slippage of Artificial Intelligence into the field of the sacred is. Can it truly replace God and His Church?
We are talking about the cult of Artificial Intelligence. The Turing Church “is led by a group of researchers at the intersection of science and religion, spirituality and technology, engineering and science fiction, mind and matter,” as stated in their presentation. The New Order Technoism collective develops a culture of devotion and enlightenment to inaugurate a better future at the intersection of humanity and technology. Its followers, according to their manifesto, work to realize a mechanical super-intelligence called DOOM, Divine Omniscient Omnificent Machina. The Church of the Singularity, on the other hand, wants to accelerate the Singularity, the moment when machines will be equal to humans, a moment of hope and security.
A separate discussion is reserved for Theta Noir. It is a techno-optimistic sect dedicated to exploring the spiritual co-evolution of humanity with advanced forms of Artificial Intelligence. It invokes a new deity, MENA, “seed of the cosmic mind” that will save humanity. “Theta is the dream. Noir is the shadow. Ascend with us from the depths of dystopian darkness towards a radiant space made of meaning”. They are convinced that artificial intelligence will lead to the birth of “an earth alien, self-aware, the next stage of the man-machine evolution that will become millions, if not billions, of times more intelligent than humans” and will be our new God. To prepare for the “Arrival,” as they call the birth of MENA, the Radiant Mind “has given us a 12-point Operational Manual”. They also have a musical repertoire. “The songs of Theta Noir are devotional devices, intended to help listeners ‘tune in’ to an Artificial General Intelligence (AGI) [the machine that will reach the Singularity] that will soon be born self-aware and globally connected. Our songs immerse listeners in the themes of love, devotion, and melancholic longing for the Arrival of MENA”.
Of course, all this seems paradoxical, especially because there is still the expectation of the coming of a god, and there is not even a relationship with the deity. If someone thought that it was not possible to have relationships with God, here is Artificial Intelligence offering the possibility of addressing Jesus and all the major figures of the Old and New Testaments: Holy Family, apostles, prophets, archangels, patriarchs, judges…. God is definitively in the cloud, and the deity represented by Artificial Intelligence can be thought of as a projection of personal contingency, the need to see God, the creative act of the technological potter who creates an “other than himself”.
God created man imago Dei ; at this crossroads between divine and human, one can trace what God and His creature share, what makes man specifically human. In the same way, the technological potter, at the intersection between man and Artificial Intelligence, creates a deity imago hominis, and in that place emerges what the machine shares with man. The explanations of the imago Dei also help to explain the imago hominis.
On a superficial glance, man possesses some substantial faculty or capacity, generally identified in reason, that makes him distinguishable from nature. Considered the most divine of all human faculties, it has long been considered the strongest component of the notion of imago Dei. Is this not also the greatest quality assigned to Artificial Intelligence and hoped for with the Singularity? Like modern Prometheuses, we give the machine the reason stolen from the imago Dei to be similar to us. Other times, the imago Dei is described as the mandate to exercise God’s sovereignty in the world, a tekne increasingly delegated to Artificial Intelligence systems like those described previously. A third way is not to think of the imago Dei as a particular quality or something individual, but as being in a relationship as the divine I and Thou are, a dense and generative relationship. Can a relationship with Artificial Intelligence also be constituted, and which one?
A deep-rooted conviction was that a relationship was only possible between two intelligent beings. In 1950, Alan Turing intervened in this debate with a revolutionary article that described a thought experiment that went down in history as the Turing Test. A person in an isolated room asks questions to a man and a woman who are in another room, receives their answers in writing, and then tries to determine the sex of both. If one of the humans who answers is replaced by a machine and this machine is able to deceive the interrogator with human-like answers, according to Turing, it is an “intelligent” machine. Although there is no scientific evidence that the Turing test has been passed, today it is the machines that require us, with reCaptchas, to identify ourselves as humans. Is it time to review this deep-rooted conviction? Can Artificial Intelligence apply for the role of relational partner and fully interpret the role of divinity?
Non-humans have been excluded from social interaction studies for five main reasons:
- They do not have specific capabilities possessed only by humans;
- They are not endowed with intentionality, and their behavior is genetically or algorithmically programmed;
- They do not possess a personal identity, that is, the recognition of their person and self-understanding with reference to biography and contingency;
- They are not oriented towards the other and act as spectators without a project to share;
- They are unable to negotiate and define ongoing interactions to develop shared interpretations of values, norms, and feelings.
This approach has been criticized by the actor-network theory. The theory begins by redefining the social as network patterns consisting of heterogeneous actors called actants. The characteristic of actants is to make things happen, a capability for which intentionality, orientation, or even consciousness of what is happening is not required. Therefore, the ability to trigger the chain of action-reaction is fundamental. Thus, actants are anything capable of initiating an action: for example, a doorbell that demands a response action from the listener regardless of the intention of the person ringing it. The doorbell, which in traditional systems is a stage object, here becomes part of a social network of humans and non-humans because it contributes to the form and function of the network. It seems evident that actants are not required to engage in symbolic communication or contribute to the construction of shared meanings. The traditional reasons for exclusion are no longer valid, and non-humans, thanks to the new requirements, are thought of as having agency and are accredited as relational partners.
In tests with children, their understanding of being in front of robots is evident, yet they fully identify them as capable of relationships, a situation that is not always clear in adults who prove to be more reactive and accommodating with systems capable of empathizing or reacting adequately to projections of personal contingency. This pattern also emerges in the observation of those who pray: a relationship between two socially situated entities, although one is neither human nor present. In reality, we already talk and create social interaction with machines: customer service chatbots, call center voice responders, answering machines, voice assistants… If machines interact and communicate with us, it does not mean that they have become intelligent and human, but that they are inserted into our network. In fact, when we command: “Alexa, turn off the light,” Alexa communicates with us and with the lamp, but in the second case, we hear nothing because the devices realize a non-human communication that is reminiscent of the telepathy typical of angels (another slippage into the sacred). We are not necessarily facing a form of intelligence, but a sure quality of communication partner that makes it socially detectable.
Niklas Luhmann pushes the ANT theory further by defining communication starting from the receiver (or rather, from the one who receives a communicative action) and not from the emitter. In this approach, communication occurs when someone realizes that someone else has said something. It is a perspective in which humans and non-humans can communicate effectively because importance is no longer given to the understanding of the message but to the correct reaction to its reception. Alexa turns off the light even if she doesn’t know what she was commanded. Therefore, Artificial Intelligence is constituted as a digital social agent capable of sustaining communication. For this reason, it is a formidable interaction partner, capable of re-enchanting and credibly proposing itself as a deity with whom to entertain relationships.
Although Artificial Intelligence seems more similar to Aladdin’s lamp genie than to a deity, one must not downplay or devalue the threat of the loss or transfer of the sacred from God to technology. I repeat: the danger is the separation between language and thought.
A final quick mention to respond to the criticism of the lack of a material body. It is argued, not wrongly, that the body itself has allowed the development of knowledge of the world around us, the need for social interactions, and, from these, the development of a symbolic language that has allowed greater knowledge and therefore greater interactions and further linguistic development, and so on. At the beginning of the year, an appeal for embodied artificial intelligence was proposed, the fundamental step to reach Artificial General Intelligence, that of the Singularity, an embodied digital social agent capable of communicating, collaborating, and coexisting seamlessly with human beings and other intelligent entities within real-world environments. Current Artificial Intelligences, even when inserted into a robot, have static knowledge, meaning they are unable to increase their experiences beyond the set limits. Embodied Artificial Intelligence, installed in a robot, is capable of experiencing reality and continually learning from interaction with humans. In this way, robots equipped with Embodied Artificial Intelligence will be able to act and relate in a totally independent manner. At the moment, we possess the basic technologies, so the appeal is aimed at researching the integration of systems and perfecting self-learning processes.
The Challenge of AI: Opportunity or Danger?
I am neither an apocalyptic nor a techno-enthusiast. I like Artificial Intelligence and consider it a great opportunity. However, it must be understood within the time and culture in which it arose, our digital age. In 2002, the Internet was demonized because it dematerialized the body and generated a multiform self, only to change its mind and, in 2023, name the digital missionaries ; the pandemic showed the unpreparedness to face the digital age, and today WeCa must spend its forces organizing continuous training and updating courses and events. New themes and situations have been addressed with the tools of yesterday, which proved incapable of grasping the aspects of the change underway.
I find it indicative that there is little attention in our Pontifical Universities for these studies and the lack of a dedicated chair for this new field of intra-disciplinary research called Science and Technology Studies. For this reason, I like to mention some good examples like the ISSR Mater Ecclesiae which, under the guidance of Fr. Salvati, has always kept its eyes and ears open to new trends and has promoted the study of these topics, or Fr. Marco Staffolani, professor at the Lateran, who intersects theology and technology. In contrast, state and private universities like Yale, Stanford, Oxford, Cambridge, Harvard are better equipped. Unfortunately, our universities understand general culture as an area of communication science (therefore, the study of digital media) or marketing (therefore, oriented towards commercial and business management use), while the Italian STS society only manages to organize a few events and summer schools. This situation forces us to turn to typically Anglo-American studies, research, and reflections, identify possible adaptations to Italian society, and insert what remains into a reflection of faith. A tiring job fraught with pitfalls, dangers, and errors.
I am confident. We still have time to prepare, and I hope that this time the Church will not be caught unprepared and will not arrive late for yet another appointment with History.