Original scientific paper Izvirni znanstveni članek
DOI: 10.32022/PHI33.2024.130-131.15 UDC: 17
Images, Artificial Intelligence, and Informational Nihilism
Veronica Neri
University of Pisa, Lungarno Pacinotti 43, 56126 Pisa, Italy veronica.neri@unipi.it
Abstract
Artificially generated images open up new ethical issues. Since it is no longer easy to understand and discern the true from the false, we can adopt a consciously critical view or a nihilistic view. In the field of information, the criterion of visual truth has become abolished. Artificial images are "nontransparent," subject to a potentially endless process of transformation; they are, moreover, replete with biases that make it difficult to understand the meaning of the image. The crisis of truth, the personalization
Phainomena 33 | 130-131 | 2024
of visual data, the enclosure into knowledge bubbles, machine learning systems, their conveyance and reception, question the very existence of visual information. The risk is the process of "defacticization" of reality, a loss of trust in the facts being told. This can lead to a kind of informational nihilism and the extinction of trust in "the other." Therefore, it is necessary to reflect on an ethics that overcomes the devaluation of the meaning of today's visual signs and allows them to be rehabilitated as bearers of informational, but also social, cultural, and anthropological meaning.
Keywords: artificial images, defacticization, nihilism, regulation, visual information.
Podobe, umetna inteligenca in informacijski nihilizem
Povzetek
Umetno ustvarjene slike odpirajo nova etična vprašanja. Ker ni več lahko razumeti in razločiti resničnega od lažnega, lahko zavestno privzamemo kritičen ali nihilističen pogled. Na informacijskem področju je merilo vizualne resnice razveljavljeno. Umetne 326 podobe so »netransparentne«, podvržene so potencialno neskončnemu procesu preoblikovanja; obenem so polne predsodkov, ki otežujejo razumevanje smisla podobe. Kriza resnice, personalizacija vizualnih podatkov, zapiranje v mehurčke znanja, sistemi strojnega učenja, njihovo posredovanje in sprejemanje postavljajo pod vprašaj sam obstoj vizualnih informacij. Tako se pojavlja tveganje procesa »defaktizacije« resničnosti, izgube zaupanja v predstavljena dejstva. To lahko vodi v nekakšen informacijski nihilizem in izumrtje zaupanja v »drugega«. Zato je treba razmisliti o etiki, ki bi premagala razvrednotenje smisla današnjih vizualnih znakov in omogočila njihovo rehabilitacijo kot nosilcev informacijskega, a tudi družbenega, kulturnega in antropološkega smisla.
Ključne besede: umetne podobe, defaktizacija, nihilizem, regulacija, vizualne informacije.
Veronica Neri
"Information, therefore non-things, is placed in front of things, and makes them fade."
Byung-Chul Han
1. The dawn of artificial intelligence and "other" images. A premise
In the time of artificial intelligence,1 the image takes on a much more complex meaning than in the past. The world seems to manifest itself solely in the form of an image. The traditional dual meaning of an image as a mirror of objective reality—from the Lascaux cave paintings to sculpture, from drawings to engravings, and up to photography and video—or as rêverie no longer suffices.2 We find ourselves in the era of images; that is, in an era where, as Ruggeri claims, two previously separate logics overlap: "the world becomes an image" and "the image becomes a world." Through algorithms, images wield more influence over the information economy and its business models than ever before (Ruggeri 2023, 8); algorithms in the form of images constitute a 327 new type of "experience" that often prompts more unconscious and irrational choices than before.
Conversely, today the images of reality and of tangible "things" seem in fact to have relatively lesser significance compared to the fluid images that populate the internet. Digital images, ab origine analog and later digitized, along with artificially generated images, form a set of powerful "other" images. They exist solely on screens, they are never tangible, they are always within reach.
The latter are computer-generated images, created by artificial intelligence tools (AI). Those are images easily altered or inherently false. In today's visual
1	"AI system" is to be intended, as approved by the EU Parliament in Article 3, as: "[An] 'artificial intelligence system' (AI system) means software that is developed with one or more ofthe techniques and approaches listed in Annex I and can, for a given set ofhuman-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with." (https://eur-lex.europa. eu/legal-content/EN/TXT/HTML/?uri=CELEX:52021PC0206.)
2	Phantasmagoria figuratively conceived of as a figment of the imagination, not as a succession of optical illusions produced by the magic lantern. For the classical term phantasmagoria, see: Mannoni 2000; Casetti 2023a.
Phainomena 33 | 130-131 | 2024
realm (Pinotti 2021; Machin and Polzer 2015), where images increasingly constitute the news, the truthfulness of the image becomes crucial. The so-called artificial image inevitably raises new ethical dilemmas, including the loss of authenticity, authorship, the propagation of social stereotypes, bias, truth in visual information. If information, as maintained by Byung-Chul Han (2022a) and Vilem Flusser, can be understood as a "non-thing" (Unding), we find ourselves immersed in a universe where "non-things are currently entering our environment from all directions, and they are pushing away the things. These non-things are called information" and are gaining irresistible influence.3.
This raises two potential issues: firstly, understanding when an image is Al-generated and to what degree; secondly, questioning whether to believe the conveyed message and its associated representation. In the words of McLuhan, "[t]his fact, characteristic of all media, means that the 'content' of any medium is always another medium. The content of writing is speech, just as the written word is the content of print, and print is the content of the telegraph."
328 (McLuhan 1964, 23-24.) This means that we have the legitimate option to no longer believe what is presented to us, or, in adherence to the notion that "the medium is the message," endeavor to discern the "truth" within the "visual artificial medium" by recalibrating the logic of "remediation."4
However, we exist not only in a visual regime, but primarily in a realm of visual information. An info-iconosphere,5 as stated by Byung-Chul
3	Flusser adds that these non-things are immaterial, "in the true sense of the expression, 'impossible to get hold of.' They are only open to decoding. Of course, as with old-style information, they also seem to be inscribed within things—in cathode-ray tubes, celluloid, micro-chips, laser beams. But although this sounds 'ontological,' it is an 'existential' illusion. The material basis of new-style information is negligible from the existential point of view." (Flusser 1999, 86 f.; Flusser 1993, 81).
4	According to Bolter and Gruisin, "remediation" is a process based on the use of certain characteristics that are typical of one older medium within another. Media continually interact with each other. With the diffusion of IA, the logic of remediation is even more articulated and richer in possibilities (cf. Bolter and Grusin 2002).
5	This term is the fusion between infosphere and iconosphere. The terms "iconosphere" and "visual culture" mark a turn towards the visual as opposed to the linguistic turn theorized by Richard Rorty in the late 1960s (Rorty 1967). "The concept of iconosphere," as noted in the International Lexicon of Aesthetics, "was first introduced by the French
Veronica Neri
Han regarding contemporary information regime (in contrast to Michel Foucault's disciplinary regime), has generated "a form of domination in which information and its processing by algorithms and artificial intelligence have a decisive influence on social, economic and political processes" (Han 2022a, 3). Consequently, ambiguous, hyperreal, unreliable, or unrecognized false visual information can instigate arbitrary choices (Peña et al. 2010, 291-301). Thus, understanding the nature of artificial images and the potential issues, arising from the widespread use of these visual forms in contexts other than news, becomes extremely significant today.6
2. Images that inform. From digital to "artificial"
Information increasingly lies within images, either static or in motion, in the middle of an incessant flow of icons where even the so-called visual journalism
philosopher Jean Wahl (1955) at a conference on the cultural problems raised by the spreading of mass media. According to Wahl, in the age of mass communication the human being is always surrounded by a double envelope, i.e. the 'iconosphere' and the 'phonosphere,' which are like 'a surrounding forest of noises and images' (Wahl 1955: 335)." On the contrary, the term "infosphere" is formed by the words "information" and "sphere." It means a realm of information, data, communication, and knowledge. The first documented use of the term "infosphere" was by Kenneth E. Boulding in 1970. He wrote that "the infosphere [...] consists of inputs and outputs of conversations, books, television, radio, speeches, church services, classes, and lectures as well as information received from the physical world by personal observation. [...] It is clearly a segment of the sociosphere in its own right, and indeed it has considerable claim to dominate the other segments. It can be argued that development of any kind is essentially a learning process and that it is primarily dependent on a network of information flows." An important study on the infosphere was then recently carried out by Floridi (cf. Floridi 2001, 18-19; Floridi 2020; Floridi 2014; Boulding 1970, Wahl 1955, 333-336). Cf. International Lexicon of Aesthetics, s.v. "iconosphere."
6 It seems important, at this point in the article, to pay attention to the difference between information and communication. The difference between information, understood as the one-way flow of a message from a sender to a receiver, and communication, as the two-way flow of a message from a sender to a receiver, certainly appears to be more blurred today than in the analog time, if it has not already totally vanished. In this context, however, I do intend information as a sphere of description of facts and events, keeping apart as much as possible the (personal) point of view of the journalist or those generating the information, including visual information (cf. Manetti and Fabris 2011, 15-16; Fabris 2019; Fabris 2014).
329
Phainomena 33 | 130-131 | 2024
strengthens its role.7 With the spreading of the network, we can now view in real time, regardless of our location, what would have remained unknown before. The ease of multiplying and transmitting images, of remodeling them (within the scope of what is permitted) to make them more effective and expressive, as well as their immateriality and immediacy, are constitutive features of a new way of communicating—more direct, immediate, intuitive, and empathetic. Similarly, the quick spreading of images has altered the concept of time, turning it into a constant present. Information embodies the power of contingency, and the informative image aims precisely to arrest this contingency.8
After all, we are immersed in information, which now surpasses the quantity of things, to the extent that Byung-Chul Han argues for the emergence of a new human practice termed "infomania," a truly fetishistic approach to information. People feel the need to have real-time information on any topic of interest, through data, sound, images, or a combination of them.9
A single informative image can (re)construct an event, recount a story, lending itself to interpretation. Even objects in the world can transform themselves into 330 sources of information, the so-called "infomes," namely "information-processing agents." And we would like them to think. These objects communicate with us inadvertently. These objects listen, act, and respond without explicit requests, guiding us (or tracking us) in our choices from purchases to routes, or suggesting to us what to read or, for example, to do things, such as shutting a window,
7	In journalism, image-based news has a rather recent origin, which we can trace back to the birth and spread of photography by Nicephore Niepce in 1826, then perfected by Daguerre ten years later. We finally have an industrial prototype that can produce serial images, with a level of detail closer to reality than lithographic printing, which, despite a high print run of color copies, is still a manual process. The so-called visual journalism is "based on the combination of text and images and [...] focuses on the expressiveness of the latter" and was born in recent years (Neri 2021, 16).
8	Cf. Luhmann 1996.
9	The term "infomania," coined more than four decades ago by Elizabeth M. Ferrrarini, refers to the relationship between information overload and the inability of humans to regulate all the information they receive and somehow try to store. The anxiety to be constantly up to date has made us more vulnerable and dependent on real-time information. Learning to be well informed qualitatively and not just quantitatively appears to be the goal of a society of hyperconnected individuals (cf. Han 2022a, 7; Han 2022b; Ferrarini 1985).
Veronica Neri
opening a door, etc. Debord foresaw the algorithmic visual domain where "the real world is transformed into mere images, [and] mere images become real beings" (Debord 1967, § 18). These evolving entities interact with humans, exemplifying the informative power of images, as well as their deep charm and their influence on public opinion. Visual information is reshaping the traditional dominion of words in the news, owing to our growing inability to focus beyond a few seconds (as Lisa Iotti provocatively writes; cf. Iotti 2020). Particularly, images swiftly cater to the audience's desire for immediate participation in events and for identification. After all, as Flusser maintains, the image arises from man's ability "to abstract from the surrounding environment and to symbolize the fundamental elements of worldly reality, condensing information into a visible, shareable, ever-present form" (Flusser 2009, x). However, within the info-iconosphere, inhabited by "infomes," we have transcended digital photography by employing artificial intelligence.
This confronts us with a new category—artificial images. What are they? Which role are they taking within the information world that is becoming ever more iconocratic?	331
These images are primarily "technical images that unify through a specific ability, evolving as trust in conventional rules wanes" (Flusser 2009, 15). A technical image can be defined as an image, which is "generated by an artificial apparatus, [and that] gradually prevails through pixel-based algorithmic models reconstructing a unit via point-like signs" (ibid., xiii). These technical images represent not just "attempts to depict the world," but rather "consequences of scientific progress—a quest to depict the concepts the photographer holds of an image" (Menon 2022, 42; cf. Flusser 2004, 102). They unify and combine scattered pieces, transitioning from "visions of objects to computations of concepts" that are multisensory and complex (Flusser 2009, 14, 238).
However, artificial images stand as a separate category within the realm of technical images. Generated by specific GAN-based software, these images may appear more real than reality, yet they ultimately constitute an amalgam of numerous images drawn from reference datasets.10
10 Several artists are using AI—for instance, Mario Klingemann in the Memories
Phainomena 33 | 130-131 | 2024
GAN is an architecture for training a generative model of artificial intelligence, which has revolutionized the way artificial content is created; this type of network was first introduced in 2014 by Ian Goodfellow. GAN consists of two neural networks: a generator and a discriminator. The former has the task of creating new data that can fool the discriminator. The latter has, on the other hand, the task of distinguishing between the data created by the generator and the real data; in essence, it is a classifier, that is, a network to be trained to distinguish between the real data and the "fake" data generated by the generator, while extracting its features. During the training process, the two networks challenge each other: the generator tries to improve the quality of the data it produces so as to fool the discriminator, while the latter tries to improve its ability to distinguish between real and fake data. This competition between the two networks leads to the creation of artificial data that may be quite realistic but nonexistent in reality, which are thus completely birthed by the generative network, even if the individual gives directions to artificial intelligence through prompts.11 332 Artificial intelligence generates images using algorithms that aim to mimic, albeit unsuccessfully, the functioning of the human mind, without truly comprehending its actions. The machine is asked to be "creative" by adhering to a set of more or less concise textual instructions provided by humans.12 It is an exploration of the machine's artificial imagination, a yet-unexplored realm. These images lack a single author; they even lack a single human author. In fact, they are co-authored by both man and artificial intelligence (Barale 2020, 6-16; Somaini 2022, 91-111; Restuccia 2023). If we can say that, with the first prompt, the human being seems to be the mind that guides the artificial generation system, and artificial intelligence is the virtual "hand" that carries out the command, the more the interaction goes on and becomes refined, the more the two roles seem to blur into each other. With images produced by artificial intelligence, we can state that we are faced with a mechanical
of Passersby I, a series of portraits of nonexistent characters, or the Obvious Group, Memo Akten, etc.—, forging new identities and new imaginaries (cf. Barale 2020).
11	As Castelle (2021, 19 ff.) argues, this is the same mechanism as the one we can trace in the theory of Bourdieu's habitus.
12	On artificial imagination, cf. Restuccia 2023.
Veronica Neri
interpretation of the digital iconographic world, a mode of information that is partly distorted by the reference dataset. These datasets, predominantly reflecting Western imagery and mainstream themes, have been collecting data since the 1990s, with an exponential surge in the last decade. Some of these hyper-real visual images raise interpretative challenges, making it difficult to discern the true from the false, unless expressly declared by the source. Therefore, we can adopt a consciously critical perspective or a disenchanted and nihilistic view, according to which we no longer believe anything we find before our eyes. Images, now, hold such a level of uncertainty that the observer's search for trustworthy points cannot fully counterweigh such ubiquitous lack of trust.
In the realm of information, the criterion of visual truth is crucial, prompting recent regulatory efforts like the European AI Act. This regulation aims to distinguish deep-fake images from real ones and implements safeguards to prevent the creation of false or illegal content. Concurrently, we are witnessing a rise in immersive journalism, emphasizing an increasing need for individuals to emotionally engage with news through experience.	333
Deep fake refers to photos, videos, and audio created by artificial intelligence, altering or recreating specifics of a body or a face based on contents from objective reality. Using mechanisms akin to deep learning, neural networks simulate the human brain's functioning, learning from substantial data to realistically modify faces in video and digital content. Potential consequences include cyberbullying, computer crimes, and pornography, as these manipulated images can unconsciously influence human choices and actions. Moreover, the datafication of individuals through unauthorized deep fake attacks can breach privacy and compromise security.
Immersive journalism, instead, is a new frontier in the realm of journalistic information. It enables individuals to have first-hand experience of events or situations told by the news programs and documentaries through virtual reality or augmented reality technologies. This sensory immersion allows individuals to immerse themselves in the described scenario (Uskali 2021).
In all these circumstances, we encounter "opaque" images, undergoing potentially endless transformations, rife with biases that make understanding the meaning and the truth of the image, if any, extremely challenging.
Phainomena 33 | 130-131 | 2024
The essential question that is still open is the relationship with the truth of reality, understood as "what we cannot change; metaphorically, it is the ground on which we stand and the sky that stretches above us" (Arendt 1968, 264). If truth seems inherently connected to the earthly, the digital era signals an end to this concept. With artificial intelligence, visual information does not represent events; it merely simulates them. This loss of the sense of reality and truth makes us feel uncertain about how to engage with information, according to Flusser (Flusser 2004, 204). He suggests that human beings no longer function as a homo faber, but turn into a homo ludens, all engrossed in entertainment, hands replaced with fingers. This shift creates a new paradigm, where reality is almost entirely influenced by entertainment and infotainment (Han 2022a, 15). Nonetheless, there exists a space for moral entertainment, serving cultural and social functions simultaneously. By aligning with the audience's knowledge and imagination, conveying moral order becomes more effective. After all, Luhmann maintains that "entertainment strengthens existing knowledge, fulfilling a role similar to the 'new mythology' that 334 romantics aspired to. Mass media entertainment reaffirms what individuals already possess within themselves, intertwining mnemonic performances with learning opportunities." (Luhmann 2002, 78.)
The crisis surrounding truth, the rampant infodemic, the personalized nature of visual data, the confinement within knowledge bubbles, and the generation and reception of images by machine learning and artificial intelligence systems fundamentally challenge the authenticity and, subsequently, the reliability of visual information.
3. "Artificial" visual information and European regulations
It is no coincidence that artificial visual information is currently, in some respects, regulated by the European Union. In June 2023, Europe approved amendments to the proposed regulation of the European Parliament and the Council of Europe, establishing harmonized rules on artificial intelligence.13 On January 2024, the final version of the text was finally submitted by the
13 It amends some previous Regulations of the European Union: C0M(2021)0206 -C9-0146/2021 - 2021/0106(C0D).
Veronica Neri
European Commission, and in March of the same year it was finally approved by the European Parliament.14
The European strategy regarding artificial intelligence aims to make the EU a world-class hub for artificial intelligence and ensure that artificial intelligence is human-centered and trustworthy, based on a classification of risk levels, mostly four: unacceptable, high, limited, and minimal or no risk. The higher the potential risk, the tighter the regulation. This marks the first regulation on artificial intelligence on this scale, aiming to provide guidelines and recommendations for a safe, sustainable, and responsible use of artificial intelligence, ultimately protecting the public from potential harm and discrimination.
In the specific area of visual aspects, several key points have emerged. Firstly, the regulation prohibits biometric analysis using sensitive data (such as gender, ethnicity, citizenship, political opinions, or religious beliefs)15 in real time. It also prohibits using copyrighted material to train generative artificial intelligence.16 Additionally, it calls for restrictions on recommender algorithms and raises concerns about the breach of privacy, stemming from 335 the indiscriminate collection of biometric data from social media or videos in public spaces to create facial recognition databases. These concerns also extend to issues surrounding the authorship of visual content, mostly falling within the legal domain.
Another ethically important aspect concerns the censorship of certain image-generating artificial intelligence systems due to ambiguous requests. For instance, prompts centered on images of violence, nudity, etc., may lead to censorship. Lastly, legal issues may arise with the transparency of these images. Actually, the European AI Act states that "users of an artificial intelligence system
14	The EU is the first in the world to have rules on artificial intelligence thanks to the AI Act. The Act was approved at the plenary session with 523 votes for, 46 votes against, and 49 abstentions. Cf. Kroet 2024.
15	Cf. European Parliament 2023.
16	In the AI Act, obligations were introduced to document the indication of copyrighted content used to feed the model, as well as the obligation to indicate whether the content was generated by artificial intelligence. In contrast, the AI Act makes no mention of the protection of works created by Artificial Intelligence systems under copyright law, leaving it to national regulations or case law. Cf. EU Artificial Intelligence Act 2024.
Phainomena 33 | 130-131 | 2024
that generates or manipulates image, audio or video content that appreciably resembles existing persons, objects, places or other entities or events and would falsely appear to a person to be authentic or truthful ('deep fake'), shall disclose that the content has been artificially generated or manipulated."17
Image generators, therefore, yield complex images, which pose risks and are influenced by social biases. In a context reminiscent of Bourdieu's habitus, where social models of classification and perception intersect with generative models, a socially-biased artificial intelligence emerges (Bourdieu 2013, 111-113), influenced by prejudices concerning ethnicity, gender, profession, and age. The prevalent datasets mirror the most active web user population. Biases exist not only within these datasets, but also in the norms, practices, and processes guiding the commissioning, development, implementation, and use of artificial intelligence within a broader societal culture. Creators and designers may inadvertently allow their biases to permeate the entire process, from data selection to model development (Schwartz 2022; Luccioni et al. 2023).
336 Training these systems on distorted data can amplify the social perception of certain aspects, further reinforcing biases reflected by the network, acting as a distorting mirror.18
Flusser encourages us to "create something improbable from the possible":
Men manipulate the keys to entice automatic systems into producing unlikely scenarios in their programs. They use the keys to allow something far-fetched to emerge from the calculated punctual universe generated by the devices. This improbable world, with its unifying ability, must envelop the punctual universe like a skin to give it meaning. The power to standardize aims to provide concrete meaning to this universe, starting from the abstract and absurd universe we find ourselves in. (Flusser 2004, 51-52.)
17	https://www.europarl.europa.eu/doceo/document/TA-9-2023-0236_IT.html. Cf. also Hagendorff 2020, 99-120.
18	Otherwise, the so-called "reverse bias" can be generated: in other words, artificial intelligence makes images with a prevalence of features that are generally unpopular on the internet.
Veronica Neri
Artificial intelligence builds a reality by adhering to specific social norms, while remaining hyper-real, also serving informational purposes. This artificial intelligence poses significant risks, considering the potential impact of certain "false" images on beliefs and decision-making processes.
In this "transparency society," in which anyone can obtain information about anything at any time, the transparency of artificial images is necessarily crucial (Han 2015).
The rules introduced by the regulation aim to ensure safety, transparency, and provide a risk-based approach, establishing obligations for artificial intelligence suppliers and users based on the level of risk that the artificial intelligence itself may entail. The levels of risk associated with the impact of different artificial intelligence systems on people's lives and their fundamental rights are defined, from work to health and safety. These regulations classify artificial intelligence applications into four levels of risk: unacceptable risk, high risk, limited risk, and little to no risk. To avoid the high-risk category, generative artificial intelligence must meet transparency requirements, such as disclosing that the content was generated by artificial intelligence, by 337 designing models that prevent the generation of illegal content, and publishing summaries of copyrighted data used for the training. However, to keep to the limited-risk range, systems should fulfil minimum transparency requirements, so that informed decisions can be made. Users must be informed when they are interacting with artificial intelligence, such as systems that generate or manipulate image, audio or video content (deep fake, for example).
Such regulations certainly appear necessary within the framework of the fundamental rights of every individual, but the moral reflection remains one of the imperatives.
4. Conclusions. Nihilism and the "de-facticization" of reality
It seems that our knowledge and understanding of the world now heavily rely on images. As Heidegger writes:
Bild, image [...] indicates [...] a form, as resonates from the German idiomatic expression: über etwas im Bilde sein, "to be informed about something," "to see clearly" [...]. Sich über etwas ins Bild setzen, "inquire
Phainomena 33 | 130-131 | 2024
about something," means: placing oneself in front of the being itself, making it present to see how it is, and having it constantly before oneself in the way in which it is placed. (Heidegger 2022, 107 f.)
However, the overcommunication of reality through digital and artificial images exposes us to a "high" risk, if we want to use the terms of the AI Act. This risk involves what Byung-Chul Han calls the process of "de-facticization" of reality (and of our culture)—a loss of trust in reported facts and in the illustrated facts, too. Guided by algorithms, the individual sees his autonomy being eroded. We are faced with increasing information entropy that blurs the line between truth and falsehood, making the boundary increasingly fragile. Information navigates a hybrid realm, where fake news can often seem more impactful than real events. Communicative effectiveness takes precedence over truth, often substituting it (Han 2022a, 11).
The photograph of an event, either captured in the moment and altered later or entirely constructed artificially—as recently attempted by photographer Zanon to 338 document the war in Ukraine or the Channel 1 project—19 "has something to do with the Resurrection" of a fact, as Barthes observes (Barthes 2000, 63-64). While analog photography freezes and shows something that was present, exposing it to its audience, digital photography presents the appearance of the fact itself.20 By manipulating the image, there is a tendency to erase the referent. With artificial generation systems, "[a] new expanded reality is generated, a hyperreality devoid of any correspondence with the actual referent. Photography created in this digital manner becomes hyperreal." (Han 2023, 43 f.) From an etymological point of view, the term "intelligence" derives from the Latin inter-legere, meaning "to read between" involving a choice between potential options. On the other hand, "artificial" implies the artificium, that is, the use of artifice or expedients to achieve specific ends. Therefore, artificial intelligence represents a working method that selects the most suitable options and creates an illusion, enhances appearances. However, the artifice as such is seductive, but lacks empathy, passion, and truth.
19	Cf. Mantengoli 2023; Caprodossi 2023.
20	As a matter of fact, Benjamin deals with the difference between exhibition value and cult value in The Work of Art in the Age of Its Technological Reproducibility (1935).
Veronica Neri
Choosing between options crafted with artifice implies building a world of very peculiar non-things, narrated without certainty of their actual existence. This eclipse of facts might lead to a form of informational nihilism and erode trust in "the other," understood both as a subject and as a narrated object. Facts cease to exist, and the interpretation of the fact becomes the fact itself. This interpretation is not necessarily a product of human thought, but is often assumed to be the creation of the algorithm's thoughts and imagination. In order to algorithmize imagination, it must be noticed that there is no real form of imagination in artificial image generators.
Some scholars claim that there may be a form of imagination in artificial intelligence, with which human imagination could converse. Perhaps one can speak of algorithmic techno-imagination, which can be understood, picking up on a point made by Finn, as imaginative thinking to try to prefigure possible futures, but it remains algorithmically constructed, the fruit of a computation. The computational space of imagination is the space, in which algorithms "work-work" and we will never know "how algorithms know what they know." Rather than imaginative space, we might call it a space, in which algorithmic 339 laws hold based on human stimuli and the datasets used (Finn 2014, 181 ff.).
The dematerialization of facts and objects brought about by the digital era clashes with the material culture that prevailed until the 1990s, while the derealization of the world appears to have emerged within the info-iconosphere.
The pervasive nihilism we find ourselves surrounded by results from the indiscriminate cacophony of numerous visual non-things. What is lacking is that visual "silence," which often communicates more deeply than many images.
Today, artificial images feel far more vibrant than those described by Barthes in a time, when image-generating artificial intelligence was not so widespread yet:
Observing the patrons of a bar, someone rightly said to me: "Look how dull they are; Nowadays, images are more alive than people." One of the distinctive signs of our time is perhaps this reversal: we live in accordance with a generalized imagination. Take the United States, for example: everything there is transformed into images; only images exist, are produced, and consumed. (Barthes 2000, 118.)
Phainomena 33 | 130-131 | 2024
Today's visual non-things appear even more ambiguous, luminous, and persuasive. We cannot exclude their existence; in numerous cultural and mass media contexts, they constitute useful communicative and expressive tools. Yet, visual non-things, even when conveying true messages, raise numerous doubts about their reliability. If the medium, as Bolter and Gruisin state, "is what remediates. A medium appropriates techniques, forms, and social meanings of other media and tries to compete with them or reshape them in the name of reality," we can attempt to "dismantle" the visual medium, in order to understand its nature, determining what is reliable to believe and what not to believe (Bolter and Gruisin 2000, 65). "Digital remediation," after all, adapts old media content to the new one, without altering the content itself. The medium, as Casetti claims, is something that "is in the middle, between one point and another," but also something that contributes to altering the terms it connects (Casetti 2023a; Casetti 2023b, 5). The remediation applied to intelligence in the visual algorithmic field, understood as "the ability to behave effectively in new situations," requires a further effort; it requires the 340 ability to understand the meaning of what we have before our eyes in situations that were unknown until a few years ago, and the ability to contextualize their meaning (Cristianini 2023).
We must learn to understand image alterations by engaging in a conscious interpretation, and must be helped in this by regulations as well (such as the recent AI Act). Techno-visual information, as Flusser argued, strongly influences society, uniting people under certain values and guiding their choices and actions. The same thing can happen with artificial images, but the opposite could happen, too (Flusser 2011). Their arbitrariness devalues reality, diminishing its significance. Hence, it is crucial to reflect on an ethics that may transcend the devaluation or hyper-valuation of today's visual signs. By careful assembling and reassembling, these signs can be rehabilitated to convey new meanings—not only informative, but also social, cultural, anthropological, and imaginative. The "de-facticization" of reality can be deeply dangerous, as it leads to the devaluation of facts, thus desensitizing us to pain. This phenomenon becomes even more apparent, when we are faced with the distressing news of our times, conveyed through digital and artificial representations that portray wars, conflicts, and violence against populations
Veronica Neri
and women. The representation is so vivid and consistent with what a certain fact or emotion may feel like in our imagination that we more readily believe the artificial rather than the photos actually taken of a certain event.
Moreover, these representations often lack clarity about their authorship, leaving us uncertain whether they originate from a human or an artificial source, blurring the boundaries of truth. The opacity of this boundary weakens trust in what is in front of our eyes and in the image-generating systems themselves. While the reliability of the systems is crucial, and is built into the image generator when it is programmed, what matters is the use that the individual makes of it, depending on the context and the people whom the generated image targets. Thus, responsibility for the risks that may affect certain images lies with all those involved. It is necessary to make an ethical use of technology, without necessarily promoting one's alignment with the imagegenerating systems, but rather with knowing how to distance oneself critically and use them by bending them to one's own ethically-oriented communicative goals. One must be aware that this technology represents both a threat and an opportunity for the AI designers of the future. Otherwise, there is a risk 341 that radical nihilism will become the individual's most natural approach to the environment, artificial and otherwise, in which we find ourselves immersed.21
Bibliography | Bibliografija
Arendt, Hannah. 1968. Between Past and Future: Eight Exercises in Political Thought.
New York: Viking Press. Barale, Alice (ed.). 2020. Arte e intelligenza artificiale. Milano: Jaka Book. Barthes, Roland. 2000. Camera Lucida. Reflections on photography. Transl. by R.
Howards. London: Vintage. Benjamin, Walter. 2008. The Work of Art in the Age of Its Technological
Reproducibility. London: Penguin Books. Bolter, Jay David, and Richard Grusin. 2000. Remediation: Understanding New Media.
Cambridge: MIT Press. Boulding, Kenneth. 1970. Economics as a Science. New York: McGraw-Hill.
21 The question of maintaining control over the use of artificial intelligence and its consequences is still open (cf. Taddeo 2021, 170-171; Floridi 2023; Chiriatti 2021; Cristianini 2024).
Phainomena 33 | 130-131 | 2024
342
Bourdieu, Pierre. 2013. Outline of a Theory of Practice. Transl. by R. Nice. Cambridge:
Cambridge University Press. Caprodossi, Alessio. 2023. "It telegiornale creato con lAI è già tra noi." Panorama. https://www.panorama.it/telegiornale-creato-intelligenza-artificiale. Accessed: April 29, 2024.
Casetti, Francesco. 2023a. Screening Fears: On Protective Media. New York: Zone Books.
---. 2023b. "Il concetto di medium." In Immagini, media, dispositivi. Studi e
prospettive, ed. by A. Fabris. Pisa: Edizioni ETS. Chiriatti, Massimo. 2021. Incoscienza artificiale. Come fanno le macchine a prevedere
per noi. Roma: Luiss University Press. Cristianini, Nello. 2023. The Shortcut: Why Intelligent Machines Do Not Think Like Us.
Boca Raton: CRC Press. ---. 2024. Machina Sapiens. Bologna: Il Mulino.
Debord, Guy. 1967. La Société du Spectacle. Paris: Éditions Buchet/Chastel. EU Artificial Intelligence Act. 2024. https://artificialintelligenceact.eu/the-act/.
Accessed: April 20, 2024. European Parliament. 2023. AI Act: a step closer to the first rules on
Artificial Intelligence. https://www.europarl.europa.eu/news/en/press-room/20230505IPR84904/ai-act-a-step-closer-to-the-first-rules-on-artificial-intelligence. Accessed: April 3, 2024. Fabris, Adriano. 2014. Etica della comunicazione. Roma: Carocci. ---. 2019. Etica del giornalismo. Pisa: Edizioni ETS.
Ferrarini, Elizabeth. 1985. Infomania: The Guide to Essential Electronic Services.
Boston: Houghton Mifflin. Finn, Ed. 2014. What Algorithms Want: Imagination in the Age of Computing.
Cambridge: The MIT Press. Floridi, Luciano. 2001. "Ethics in the Infosphere." The Philosophers' Magazine 16: 18-19.
---. 2014. The Fourth Revolution: How the Infosphere is Reshaping Human Reality.
Oxford: Oxford University Press. ---. 2020. Pensare Infosfera. La filosofia come design concettuale. Milano: Raffaello Cortina Editore.
---. 2023. The Ethics of Artificial Intelligence: Principles, Challenges, and Opportunities.
Oxford: Oxford University Press. Flusser, Vilém. 1993. Dinge und Undinge. Phänomenologische Skizzen. Munich: Hanser.
---. 1999. "The Non-Thing 1." In V. Flusser, The Shape of Things: A Philosophy of
Design, transl. by A. Mathews. London: Reaktion Books. ---. 2004. La cultura dei media. Transl. by T. Cavallo. Milano: Bruno Mondadori. ---. 2009. Immagini. Come la tecnologia ha cambiato la nostra percezione del mondo. Transl. by S. Patriarca. Roma: Fazi.
Veronica Neri
---. 2011. Into the Universe of Technical images. Minneapolis: University of Minnesota Press.
Hagendorff, Thilo. 2020. "The Ethics of AI Ethics: An Evaluation of Guidelines." Minds and Machines 30: 99-120.
Han, Byung-Chul. 2015. The Transparency Society. Stanford: Stanford University Press.
---. 2019. Good Entertainment: A Deconstruction of the Western Passion Narrative.
---. 2022a. Le non cose. Come abbiamo smesso di vivere il reale. Torino: Einaudi.
---. 2022b. Infocracy: Digitization and the Crisis of Democracy. Cambridge: Polity Press.
Heidegger, Martin. 2022. "Lfepoca dell'immagine del mondo." In M. Heidegger, Holzwege. Sentieri erranti nella selva, ed. by V. Cicero. Milano: Bompiani.
International Lexicon of Aesthetics. https://lexicon.mimesisjournals.com/index.php. Accessed April 19, 2024.
Iotti, Lisa. 2020. Otto secondi. Viaggio nellera della distrazione. Milano: Il Saggiatore.
Kroet, Cynthia. 2024. "Intelligenza artificiale: il Parlamento Ue approva il testo finale dell'AI Act." Euronews. https://it.euronews.com/my-europe/2024/03/13/ intelligenza-artificiale-il-parlamento-ue-approva-il-testo-finale-dellai-act. Accessed: April 1, 2024.
Luccioni, Alexandra Sasha, et al. 2023. "Stable Bias: Analyzing Societal Representations in Diffusion Models." arXiv e-prints 2303.11408.
Luhmann, Niklas. 1996. Entscheidungen in der "Informationsgesellschaft." https:// www.fen.ch/texte/gast_luhmann_informationsgesellschaft.htm. Accessed: April 20, 2024.
---. 2002. La realta dei mass media. Transl. by E. Esposito. Milano: Franco Angeli.
Machin, David, and Lydia Polzer. 2015. Visual Journalism. London: Macmillan Education, Palgrave.
Manetti, Giovanni, and Adriano Fabris. 2011. Comunicazione. Brescia: La Scuola.
Mannoni, Laurent. 2000. La grande arte della luce e dellombra. Torino: Lindau.
Mantengoli, Vera. 2023. "Intelligenza artificiale, la fotografa Zanon: 'Il mio bellissimo (ma falso) reportage dall'Ucraina, tutto fatto al computer." La Reppublica. https:// www.repubblica.it/cronaca/2023/04/04/news/intelligenza_artificiale_falso_ reportage_foto_ucraina_barbara_zanon-394896322/. Accessed: April 30, 2024.
McLuhan, Marshall. 1964. Understanding Media. New York: McGraw-Hill.
Menon, Marco. 2022. Vilém Flusser e la «rivoluzione dell'informazione». Comunicazione, etica, politica. Pisa: Edizioni ETS.
Neri, Veronica. 2021. Notizie visive. Pisa: Pacini Edizioni.
Peña, Nonny de la, et al. 2010. "Immersive Journalism: Immersive Virtual Reality for the First-Person Experience of News." Presence: Teleoperators and Virtual Environments 19 (4): 291-301.
Pinotti, Antonio. 2021. Alla soglia dell'immagine. Da Narciso alla realta virtuale. Torino: Einaudi.
343
Phainomena 33 | 130-131 | 2024
Restuccia, Francesco. 2023. "Un'immaginazione artificiale? Flusser e l'intelligenza artificiale generative." In Immagini, media, dispositivi. Studi e prospettive, ed. by A. Fabris, 68-93. Pisa: Edizioni ETS.
Rorty, Richard. 1967. The Linguistic Turn. Chicago: University of Chicago Press.
Ruggeri, Eugenio. 2023. Capitale algoritmico. Brescia: Scholé, Morcelliana.
Schwartz, Reva, et al. 2022. Towards a Standard for Identifying and Managing Bias in Artificial Intelligence. Gaithersburg: National Institute of Standards and Technology.
Somaini, Antonio. 2022. "On the altered states of machine vision. Trevor Paglen, Hito Steyerl, Grégory Chatonsky." AN-ICON. Studies in Environmental Images 1 (1): 91-111.
Taddeo, Maria Teresa. 2021. "Costruire l'etica dell'intelligenza artificiale." In Il potere del piferraio magico, ed. by G. Fregonara, 165-178. Pisa: Pisa University Press.
Uskali, Turo, et al. 2021. Immersive Journalism as Storytelling: Ethics, Production, and Design. London: Routledge.
Wahl, Jean. 1955. La culture est-elle en péril? Débat sur ses moyens de diffusion:
presse, cinéma, radio, télévision. Rencontres Internationales de Genève. Neuchâtel: Editions de la Baconnière.
344