Chapter 2. Is the machine or the person creative?¶

The question of the creativity of artificial intelligences is at the center of a heated debate involving not only the technological world but also the philosophical, artistic, and legal worlds. To address this complex question, it is necessary to carefully examine what we mean by creativity and how AI relates to this concept.

Traditionally, creativity has been considered an intrinsically human characteristic, defined as the ability to produce original and valuable ideas or artifacts. In the context of AI, however, this concept requires reformulation. AI creativity could be defined as the ability to generate new and unexpected outputs that go beyond the simple reproduction or random mechanical combination of existing inputs.

The strongest objection raised against the idea of AI as a creative entity is that it does not create but combines existing elements, not being able to "invent." I do not share this statement, but even if it were correct, it is important to recognize that combining existing elements is an integral part of human creativity, and perhaps even its main aspect. Putting together known concepts and objects but in different ways often leads to new and innovative results; in fact, many of the greatest human inventions and works of art are the result of new combinations of preexisting ideas and concepts.

Looking at art history, perhaps the most immediate example is that of Marcel Duchamp and his readymades. Think of his famous Fountain, which Duchamp did not physically create but "created" as a work of art by radically changing the context and perspective from which it is observed.

Or, going further back in time, Giuseppe Arcimboldo's portraits made with fruit and vegetables. An example is The Vegetable Gardener in which he also played with the point of view: the painting seen right-side up is a basket with vegetables, seen upside down it becomes one of his portraits.

AI, thanks to its computing power and machine learning techniques, can combine and rework information and creative elements in ways that go beyond human capabilities in terms of speed and scale. This allows them to explore a much larger space of possibilities than a human could do, potentially leading to combinations and results that a human would never have considered.

Returning to the statement that AI would not be creative as only capable of combining existing elements, the case of AlphaGo (discussed in a few pages) precisely demonstrated that modern artificial intelligences, particularly those based on deep learning models, do not limit themselves to a simple recombination of existing elements, but through processes of abstraction and generalization can create internal representations of concepts that go beyond the specific data on which they were trained. This allows them to generate outputs that, while based on existing information, can be considered new and unexpected.

It is important to note that the "creativity" of AI is different from human creativity, not possessing intentionality, emotions, or understanding of cultural context in the way that human beings experience them. Their "creativity" emerges rather from complex statistical and computational processes, which can produce surprising and innovative results, but without the awareness or intention that we commonly associate with human creativity.

In summary, while AI demonstrates abilities that can be defined as "creative" in certain aspects, the nature of this creativity is fundamentally different from human creativity. AI can generate, combine, and transform in surprising and innovative ways, but they lack the intentionality and contextual understanding that characterize human creativity. This does not diminish the value or potential of AI as creative tools, but underlines the importance of considering AI creativity as a distinct phenomenon, complementary rather than a substitute for human creativity.

A phrase attributed to Pablo Picasso says "computers are useless; they can only give you answers" [See William Fifield, "Pablo Picasso: A Composite Interview," in The Paris Review, 32 (Summer-Fall 1964)] and ironically draws attention to the questions and therefore to the human part of what I call the "hybrid author," because since forever it is the person who must ask the questions, even more so in the era of creative AI, along with directing the machine toward a meaningful result. The responsibility of interpreting and curating the work remains in the hands of the author.

The computational power of AI reaches its full creative potential only within the hybrid author, where it integrates with human awareness and intentionality. It is this synergy that transforms the machine's process into a conscious creative act: the design contribution and artistic vision of the human author guide and contextualize the AI's capabilities, generating creativity that becomes an authentic expression of the hybrid author's personality.

Obviously this makes the person even more responsible; it is in fact their contribution that defines the creativity of the pair.

2.1. AlphaGo: a case study on AI creativity¶

A very interesting example of the creative capacity of AI is represented by AlphaGo, the artificial intelligence developed by Google DeepMind to play the board game Go.

In March 2016, AlphaGo accomplished an extraordinary feat by defeating Lee Sedol, eighteen-time world Go champion.

Go is an ancient game known for its strategic complexity and for the fact that it requires deep intuition, in addition to simple computing power. Before AlphaGo, it was believed that AI could not compete at a professional level in this game, precisely because of the need for an "intuitive" understanding of positions on the board.

What made AlphaGo's victory particularly significant from the point of view of creativity is the way it played. AlphaGo did not limit itself to replicating human strategies or calculating all possible future moves (which is impossible in Go due to the enormous number of possibilities). Instead, it developed completely new approaches to the game, executing moves that human commentators initially considered mistakes, but which then proved to be deeply strategic.

One move in particular, number 37 in the second game, amazed human observers. It was a move that no human player would have considered, but which proved crucial to AlphaGo's victory. This move demonstrated that the AI was not simply replicating human strategies but was generating new strategic ideas.

Lee Sedol's opinion on the matter is as follows:

I thought that AlphaGo was based on probability calculation and that it was just a machine. But when I saw this move, I changed my mind. Surely, AlphaGo is creative. [Lee Sedol, quoted in DeepMind, AlphaGo]

For those who would like to learn more, an account of the individual games can be found on the dedicated English Wikipedia page [https://en.wikipedia.org/wiki/AlphaGo_versus_Lee_Sedol].

The AlphaGo case demonstrates that AI can not only match human performance in complex tasks but also generate innovative approaches that go beyond conventional human thinking. This raises fascinating questions about the nature of creativity and intelligence, challenging our conceptions about what it means to be "creative."

On closer examination, however, the creativity demonstrated by AlphaGo is limited to a specific albeit highly structured domain like Go. While an interesting type of creativity emerges on one hand, on the other it remains a form of creativity different from human creativity and lacking the versatility and adaptability of human creativity, which instead can range across different domains and unstructured contexts.

2.2. The role of the human author in the era of creative AI¶

In the era of creative AI, the role of the human author has evolved significantly, further strengthening the transition from direct creator to orchestrator and curator of the creative process. This dynamic requires a reconsideration of what it means to be an author and how human creativity manifests itself in collaboration with AI.

Obviously the case of Refik Anadol is an extreme example (see paragraph 1.2), both for the project and for the availability of resources necessary to create a work destined for MoMA.

However, the creativity of the hybrid author does not depend on the complexity of the tools used: even through normally accessible AI services and a simple computer it is possible to fully express one's creative potential and create original works.

Operationally, where can the "normal" author intervene if they cannot concretely participate in the construction of the work? Let's start from an observation: working creatively with a generative AI is an interactive and iterative process.

The interaction between human being and artificial intelligence develops through three fundamental phases: input, generation, and selection. These phases, through iterative cycles, lead to the desired result. Together they represent the moment of actual image generation, but they are preceded by a crucial phase: project definition. It is at this moment that the idea is born and takes shape in the author's thought.

Project definition, together with the input and selection phases, remains firmly in the hands of the human being. These phases share a distinctive element: they are the moments of choices. The centrality of the decision-making process in the creative journey is such that it has become one of the criteria used in the legal field to evaluate the degree of authorship of the person in generating an image, as explored in chapter 4, dedicated to legal and regulatory aspects.

The iterative process is articulated in several well-defined phases. It begins with setup, during which the human author provides the initial inputs of each cycle through the definition of instructions and any parameters. This phase is determining because it outlines the direction and boundaries within which the entire creative process will develop, significantly influencing all subsequent phases.

In the generation phase, the AI processes the inputs received and produces a series of outputs. The author then proceeds with selection, examining the generated outputs and choosing those that best correspond to their creative vision. The interpretation phase follows, during which the author contextualizes the selected outputs and attributes meaning to them, inserting them into a broader conceptual framework.

The process can continue with the next iteration cycle, in which the author can refine the prompts or modify the parameters to guide the AI toward results increasingly adherent to their vision.

The journey concludes with a final phase of selection and organization of the generated material. This moment includes various activities: the identification and preservation of elements immediately useful for the project in progress; the selection and archiving of potentially useful material for future projects. It is common practice that in this phase the majority of images generated by the AI that do not fall into the previous categories are eliminated.

Each phase of this process contributes significantly to the final result. It is worth underlining how the setup phase is not simply preparatory but represents a fundamental creative moment in which the author defines the essence and direction of the project, a process that can also develop independently of interaction with AI.

Speaking of the theme of responsibilities, those of the human author cross the entire arc of the creative process, manifesting in various fundamental areas. The conception of the work represents the starting point: the initial idea and the overall vision remain the exclusive prerogative of the human author. Despite the power of AI as a tool, the definition of the purpose and meaning of the work remains firmly in the hands of the author.

The direction of the creative process constitutes another central responsibility. The author guides the path through the accurate formulation of prompts and the calibration of parameters, orienting the AI toward the realization of their own artistic vision. In the field of selection, the author applies their critical judgment and their understanding of the artistic and cultural context to evaluate and organize the outputs generated by the AI.

Contextualization represents a further area of responsibility: it is up to the author to place the work in a broader artistic, social, or cultural framework, attributing meaning and relevance to it. The author must also consider the ethical implications of using AI in artistic creation, addressing fundamental questions such as the authenticity of the work, the correct attribution of authorship, and the management of potential algorithmic biases.

What can be called a reasoned creative process belongs to the person, to counterbalance the "creative exuberance" of the machine. The generative capacity of AI must be considered simultaneously an opportunity and a challenge for the creative process. The ease with which these systems can produce a multitude of content requires the author to take a disciplined and conscious approach. It is fundamental to resist the temptation to rely exclusively on massive generation, which would risk replacing the quality of human creative thought with the mere quantity of automated output.

The reasoned creative process develops through well-defined steps. Critical reflection requires an in-depth evaluation of each AI output, considering not only the aesthetic aspect but also consistency with the overall artistic vision. Intentionality must permeate every decision, from the formulation of prompts to the final selection. Conscious iteration privileges the progressive refinement of results over the indiscriminate generation of variants.

The ability to synthesize allows the author to integrate the various AI outputs into a unitary and significant work. It is also essential to maintain openness to the unexpected, recognizing the creative potential even in apparently misaligned results, which can suggest new artistic directions or inspire innovative developments.

The quantity of outputs generated by AI cannot replace the author's creative thought. Creativity in the AI era manifests itself in the ability to guide, interpret, and synthesize these outputs in a meaningful and original way. This confirms the central and irreplaceable role of the human author in providing vision, context, and meaning, transforming the potential of AI into authentic artistic expression.

The effective use of AI in artistic creation requires a specific set of skills. A solid foundation of traditional artistic knowledge is necessary, including understanding of composition, color, form, and style. Equally important is a good understanding of how AI works, accompanied by specific prompt engineering skills for the effective formulation of instructions and parameter management.

Curatorial skills play a crucial role in selecting, organizing, and contextualizing outputs into a coherent work. Finally, conceptual skills allow for the development of artistic ideas that innovatively exploit the potential offered by AI. The challenge for contemporary authors lies in the ability to master this new medium while maintaining their distinctive artistic voice and a critical approach to the creative process.

This set of skills leads to the emergence of new professional figures; many artists working with AI now present themselves with hybrid titles such as "AI artist," "AI developer," or combinations of these concepts. These names reflect the interdisciplinary nature of their work and the fusion of previously distinct roles.

We can therefore say that the human author is not simply a user of new technologies, as an evolved artisan might be, but a pioneer who explores new forms of artistic expression. Through collaboration with AI, these authors are redefining the boundaries of creativity, challenging traditional conceptions of originality, authorship, and creative process.

On one hand, the author must maintain a clear creative vision and guide the process toward results consistent with this vision. On the other hand, they must remain open to the unexpected possibilities offered by AI, recognizing the creative potential in apparently random or unrelated outputs.

2.3. Brainstorming with AI: exploring new creative frontiers¶

Brainstorming with AI represents a new frontier in the creative process, offering unique possibilities for exploration and idea generation. Unlike traditional brainstorming, which is based solely on human interaction, this hybrid form of ideation exploits the capabilities of AI to generate, combine, and transform concepts in ways that are often surprising and non-intuitive for the human mind.

One of the most powerful applications of brainstorming with AI is the development of vague or incompletely formed initial ideas. In this context, AI can act as a catalyst for the expansion and concretization of abstract concepts.

For example, an artist might have a vague idea of a work that explores the concept of "loneliness in the digital age." Using open and iterative prompts with a generative AI for images, the artist can explore different visual representations of this concept. The AI might generate images of people surrounded by technological devices but with distant expressions, or densely populated urban landscapes but with isolated and disconnected figures.

The key in this phase is the formulation of prompts that are open enough to allow the AI to explore freely, but sufficiently focused to maintain a coherent direction. An effective approach consists of starting from a base prompt and generating a limited series of variations, noting the most interesting elements of each result. These elements can then be recombined into new, more specific prompts, creating an iterative process that progressively refines and develops the initial idea. This method allows exploring creative possibilities in a structured way, avoiding dispersion in too many simultaneous directions.

Another approach to brainstorming with AI is to give the machine "carte blanche," allowing it to generate ideas and concepts without specific constraints. This method can lead to surprising and unexpected responses, opening new creative directions that the artist might not have initially considered.

In this scenario, the artist might provide the AI with a very generic or even random prompt, such as "blue Tuesday dance." The output might contain a series of surreal images that combine elements of color, emotion, and movement in unexpected ways. The artist's role here would be that of explorer and interpreter, grasping interesting ideas from the AI's proposals and developing them further.

The challenge in this approach is to maintain a balance between openness to the unexpected and the ability to recognize and develop potentially valid ideas. Not all AI proposals will be useful or relevant, but the human author's ability to grasp unexpected ideas can play a crucial role in the creative process.

The effective integration of AI brainstorming into the broader creative process requires a conscious strategy. It is important to balance AI output with human artistic vision, using technology as a tool for inspiration and exploration, rather than as a substitute for human creative thought.

An effective strategy develops through several interconnected phases. The process is based on alternating between traditional brainstorming sessions and AI-assisted sessions, creating a continuous dialogue between human intuition and machine suggestions.

Structured documentation represents a fundamental element of this strategy. It is essential to develop an organized system for recording and classifying the ideas generated during sessions with AI, so as to facilitate their recovery and development in the subsequent phases of the creative process.

Each brainstorming session must be followed by a moment dedicated to critical analysis of the AI outputs. This phase allows identifying recurring themes, evaluating the most promising ideas, and outlining possible directions for future explorations.

The ideas generated during this process do not represent an end point but constitute the basis for further developments. They can be used as a starting point for new iterations, both through interaction with AI and through traditional methods of creative elaboration.

AI brainstorming also raises important ethical and practical questions. One of the main concerns regards the intellectual property of ideas generated during these sessions. While the initial input comes from the person, the AI's role in generating specific ideas can make the line of authorship blurred.

Furthermore, there is the question of how to maintain artistic authenticity in an AI-assisted ideation process. Artists must be aware of the risk of becoming excessively dependent on AI suggestions, potentially at the expense of their own unique creative voice.

AI brainstorming should be seen as a phase of "pre-creation," where the ideas generated are considered raw material to be reworked and transformed through the filter of human sensitivity and artistic intention.

AI brainstorming is redefining the creative process, offering artists new tools to expand the boundaries of their imagination. This practice represents a bridge between traditional human creativity and the possibilities offered by artificial intelligence.

Looking to the future, we can expect a continuous evolution of these techniques, with increasingly sophisticated AI capable of participating in even deeper and more nuanced ways in the ideation and creation process. However, the ultimate value of these technologies will remain in their ability to amplify and—in a certain sense—inspire human creativity, rather than replace it, acting as catalysts for creativity.

2.4. The hybrid author: a new creative figure¶

We thus come to reflect more specifically on what I call the "hybrid author." This figure embodies the synergy between the technological capabilities of AI and human artistic vision, creating a bridge between the world of traditional art and the frontiers of technological innovation.

The hybrid author is a new paradigm in collaboration between human being and machine in the field of creativity. This collaborative model opens the way to expanding the boundaries of human creativity, offering new creative horizons. However, this evolution raises important questions about the nature of creation, authorship, and the creative process itself.

The question of ethics and authenticity emerges as the first significant challenge. It is necessary to find a balance between the use of AI and the authenticity of the creative process, addressing the complex ethical implications linked to the use of AI systems trained on works by other creatives.

Copyright and intellectual property represent a second crucial area of reflection. The evolution of copyright and intellectual property concepts in the context of human-AI hybrid creation requires a rethinking of traditional paradigms.

The theme of education and training arises as a third area of interest. Preparing the next generation of hybrid authors raises questions about the essential skills to develop and the most effective training methodologies.

Critical evaluation constitutes a fourth element of reflection. Evaluation criteria will need to evolve to integrate and adequately consider the role of AI in the creative process.

Finally, the preservation and documentation of works that incorporate AI-generated elements presents specific challenges, especially considering the rapid evolution of the technologies involved.

The hybrid author combines human creative vision with the technological potential of AI, also implying the need to remain open to the possibilities offered by this synergy along with a critical and reflective approach on the role of the creative and on the meaning of the work in the era of artificial intelligence.

2.5. Is technical knowledge of AI necessary?¶

It is important to consider that we are facing emerging technologies, still in a phase of extraordinary development, not only in terms of the number of applications but also of continuous technical innovations. Consequently, I believe it is necessary to deepen the understanding of the operating mechanisms not only of the various AI but also of the ecosystem that surrounds them. To illustrate this point, consider the discussions on AI training. Without an in-depth knowledge of how the datasets were formed, there is a risk of reaching erroneous conclusions, which can lead to inappropriate decisions.

This consideration applies to all the actors involved. Personally, I find it frustrating to hear unfounded statements about the alleged "theft" of images for AI training, often coming from professionals in other sectors or from self-proclaimed experts who talk about it as if all generative AI for images had been trained with images taken without authorization. Characters who, despite presenting themselves as experts in the sector, show superficial knowledge of the subject: they are not able to distinguish between open and proprietary datasets, they ignore the differences between the licenses that regulate the use of images for training, and they often don't even know the names of the main datasets used in the sector, let alone their technical characteristics or the ways in which they were assembled.

Returning to the main theme, if AI were to be recognized as co-authors, new questions would naturally arise; for example, in case of sale of the work, should part of the proceeds go to whoever "built" the AI?

From my point of view as a photographer, I think not. The parallel that comes to mind is that of Photoshop, the well-known photo editing software developed by Adobe. Adobe does not claim any rights on works created with Photoshop and then sold by the authors; correctly I would say, given that Adobe's purpose is to develop software for creatives, not to participate in the creation of images. Bringing the reasoning to AI, those who developed them did not aim at generating specific works but at completing a "machine."

Looking at the contracts that regulate the use of these tools (i.e., the so-called Terms of Service or TOS), it is generally noted that the generated images are the property of the user, or at least the rights of use are attributed to the user (however, it is always advisable to verify specific cases). Stable Diffusion also follows this line, stating that the generated images belong to the user. A difference between proprietary services and Stable Diffusion lies in the fact that the former usually reserve a right to use the generated/uploaded materials without economic requests, while Stable Diffusion explicitly renounces any claim.

Verification of the Terms of Service (TOS) is an essential phase in using generative AI systems. This analysis is particularly important for understanding the limits and possibilities of using generated images. The various AI services, in fact, provide for differentiated conditions of use: some allow full commercial use, others place specific restrictions. Understanding these terms is fundamental to ensure appropriate and legally compliant use of content generated through each platform.

2.6. Can you have copyright on works created with AI?¶

The use of AI in creative processes raises important legal questions, particularly in the area of copyright and intellectual property of works created with their support.

Although many legal systems have not yet developed a specific regulatory framework for this new reality, jurisprudence is already outlining significant orientations. Various rulings and official documents have in fact begun to recognize the possibility of protecting through copyright the works generated with the assistance of AI, establishing important precedents for the regulation of this rapidly transforming sector.

For example, the U.S. Copyright Office, in its guidelines dedicated specifically to this new phenomenon, recognizes the possibility of claiming exclusive rights for the human part of the work in the presence of a sufficient level of human authorship. In section III it states:

In other cases, however, a work containing AI-generated material will also contain sufficient human authorship to support a claim to copyright. [Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, March 16, 2023]

Furthermore, there are rulings in other parts of the world that follow a similar approach. In these cases, copyright was granted when the person who used the AI was able to demonstrate sufficient creative commitment in the process of creating the work.

We will return to these aspects in the second part.