05 nov Will Machines Ever Enjoy Listening to Stories? And Why Do We Need to Know? (Part 3)
An 8-part series published every 2 days
MACHINE AND HUMAN LEARNING
We learn like dogs
The human brain, like that of many other animals, largely operates based on a reward and punishment system mediated by neurotransmitters. From a behavioral perspective, when we conform to our social standards or expectations—like telling a joke that makes everyone laugh—our brain activates our reward system, releasing a dose of a neurotransmitter that provides a sensation of pleasure and satisfaction, reinforcing the behavior and encouraging us to repeat it in the future.
On the other hand, when we do something that goes against these standards or expectations, such as telling a joke that nobody laughs at, the physical discomfort we feel is associated with a neurotransmitter that prepares the body to deal with situations it considers threatening or uncomfortable, causing unpleasant physical sensations that discourage us from repeating that behavior.
This feedback system is fundamental for learning and adaptation. It helps us navigate the social world and understand what is considered acceptable or desirable, guiding our behavior to maximize rewards (like pleasure and social acceptance) and minimize punishments (like discomfort and rejection). This is why I refer to these physiological elements as Disciplinary Hormones. I will detail each of them later.
In addition to humans, many animals also respond to this system of reward and punishment. This is evident in dog training, where desired behaviors are rewarded with treats or affection, while undesirable behaviors are discouraged through corrections.
Learning or memorizing?
While the reward and punishment system works well for human learning, machines use a different mechanism known as deep learning, a subcategory of machine learning that uses artificial neural networks composed of multiple layers (hence the term “deep”). These neural networks are inspired by the structure and function of the human brain.
In deep learning, neural networks automatically learn hierarchical representations of data, from the most basic features to the most complex, without the need for human intervention. This is especially effective in complex tasks such as image recognition, natural language processing, and text generation.
Machine learning is revolutionizing the creative industry by allowing artists, musicians, and screenwriters to explore new forms of creation and co-create with technology. Tools like GANs help transform sketches into detailed landscapes, while models like MuseNet and Jukedeck allow for quick generation of original music. In cinema and advertising, AI like MidJourney creates visual images from textual descriptions, and deepfakes rejuvenate or replace actors’ faces. In games, dynamic dialogues and interactive stories are made possible with ChatGPT and AI Dungeon, providing immersive and personalized experiences for players. Regardless of any criticism one might have of artificial intelligence, it’s a fact that it expands the possibilities of human creativity and brings new forms of experimentation and production in art, music, and storytelling.
Can they, or not?
There are already artificial intelligence machines that can paint, compose music, write articles, and create stories. But this isn’t hard, I assure you. I’ve even seen an elephant painting with its trunk. What’s hard is creating an extraordinary painting, composing a memorable piece of music, or crafting breathtaking stories. That they can’t do. Some enthusiasts claim that one day they will be able to create anything we can, and I believe that may be true. However, I have the impression that this will still take a little while. But what does “taking a little while” mean today? That’s the question.
However, things aren’t moving at an elephant’s pace; quite the opposite. Practical examples show how AI is already influencing art. In recent years, some works created by artificial intelligence have had a significant impact, both for their technical merit and for the discussions on originality and artistic value.
“Portrait of Edmond de Belamy” (2018) – Created by an artificial intelligence model developed by the French collective Obvious, this painting is a blend of visual characteristics typical of classical portraits but with distorted and blurred details that hint at algorithmic processing. The piece was auctioned for over $430,000 at Christie’s, marking a milestone that brought visibility to AI-generated art and raised questions about value, authorship, and what defines a work of art.
“AICAN” (AI Creative Adversarial Network) – Created by Dr. Ahmed Elgammal and his team, AICAN is a neural network trained to generate images that resemble known artistic styles but with a touch of originality. The works created by AICAN have been exhibited in art galleries and compared to the production of contemporary human artists, sparking debates on AI’s ability to intuit new styles and the authenticity of such artistic expression.
“DeepDream” – This Google project, launched in 2015, uses convolutional neural networks to enhance and distort patterns in images, creating hallucinogenic and dreamlike visuals. Many of the works created with DeepDream have a surreal and even disturbing appearance, with faces, eyes, and animal-like shapes emerging in abstract ways. This unique aesthetic has inspired digital artists and highlighted the possibility of creating unique art through AI’s artificial perceptions.
“The Next Rembrandt” (2016) – This initiative reproduced Rembrandt’s style by training an AI with deep learning techniques on his works. The AI was programmed to create an original portrait as if painted by Rembrandt himself. The project was displayed as a tribute to the Dutch master, sparking discussions about AI’s role in preserving, replicating, and reinterpreting historical styles, as well as the boundary between tribute and authenticity.
“Memories of Passersby I” (2019) – Created by German artist Mario Klingemann, this installation uses a generative adversarial network (GAN) to produce human portraits that have never existed, in real time. The machine generates faces that gradually transform, without repetition. Exhibited at London’s Sotheby’s art gallery, this piece was praised for its innovation and criticized for challenging the limits of creativity, as its creator is an AI without memory or emotional references.
These examples show how AI-generated art has challenged the concept of authorship and authenticity, opening new possibilities while also sparking debates about the emotional and human value of artistic creation. Each work pushes, in some way, the boundaries of what we call art and invites the public to reconsider where human intuition ends and artificial innovation begins.
Artificial intelligence is us
By seeking references without critical thinking, AI ends up producing worrying biases. In human learning, the reward and punishment system reinforces desirable social behaviors and discourages actions considered inappropriate, promoting conformity with cultural norms and expectations. Artificial intelligence absorbs and reproduces these social patterns from the data on which it is trained, resulting in algorithmic biases that reflect—and often reinforce—preexisting cultural stereotypes. These algorithmic biases, harder to identify and correct, can reproduce societal patterns unquestioningly, intensifying or perpetuating social norms and prejudices. For this reason, it’s essential to distinguish between intentional human content, such as our preferences and values, and the content embedded in AI through training data.
The site Rest of World analyzed 3,000 AI-generated images from different countries and cultures and found that they portray the world in a profoundly stereotyped way. Unsurprisingly, this visual piece clearly shows how biases are deeply rooted in AI systems.
STORYTELLING IS DIFFERENT
The limits imposed by logic
I don’t know if you noticed, but all of the examples above are based on visual arts, which is a much more open field, so to speak. Visual artists can produce practically anything: from structured narratives to works that show a complete absence of logic or meaning, and still find resonance with the audience. Meanwhile, a storyteller must follow a set of rigid rules for the narrative to be, first of all, understood. And then, if well-constructed, appreciated. In other words, a storyteller needs to use language (or recognizable codes), must present a logical sequence, and create an arc with a beginning, middle, and end. This is very different from painting a picture, for example. It’s not more or less important; it simply requires a greater confrontation with boundaries. And in this area, AI still lacks originality.
Logic alone doesn’t create good stories
While machines can structure narratives based on congruent processes, they lack the intuition and emotional sensitivity needed to create something that resonates deeply with humans. What it does, for now, is structure the generative process using references to what has already been done. This, of course, is also part of the human process. But great storytellers go far beyond this. Information, facts, and logical thinking are very important, but on their own, they don’t have the same impact on our brains as a well-told story.
Good stories, therefore, aren’t simply logical sequences of events but complex combinations of feelings, contexts, and experiences. Human creativity often defies logic, and this is what allows stories to have a lasting emotional impact.
An interesting example is the series of stories created by the GPT language model, which, despite its ability to generate coherent and logically structured narratives, often results in stories that fail to create an emotional connection. For example, when trying to write a mystery story, GPT can organize the plot and twists logically and convincingly, but the characters lack emotional depth. They move through the story predictably, without the psychological complexity or subtle nuances that would make the reader genuinely care about their fate.
In contrast, for example, Dostoevsky’s novel Crime and Punishment explores Raskolnikov’s torturous psychological journey after committing murder, in a narrative that prompts deep reflection on morality, guilt, and redemption. The novel engages the reader because it captures the protagonist’s anguish and inner conflict with a human intensity that transcends logic. It is precisely this ability to explore psychological and existential nuances that many AI-created stories still fail to replicate, leaving them technically well-executed but emotionally distant—and thus harder to connect with.
Why do we humans enjoy listening to stories?
The reasons are diverse, but here are a few key ones that I believe are fundamental to explain our appreciation for this special form of human communication:
1) Listening to stories is a learning process. We learn through metaphors and associations, and good stories are filled with both.
2) Stories organize our thoughts. Since our mind is a wild horse in a birdcage, thoughts overlap to form a tangle of concepts. A story we connect with and identify with doesn’t simply bring us new information or emotions; it organizes, summarizes, and simplifies a feeling we could easily produce if our mind weren’t the confusion it is.
3) Stories carry identification and empathy. We see ourselves in stories. We see our ability (or lack thereof) to face life’s obstacles. We encounter possibilities and limitations of the human species. Stories offer us the chance to see our paradise and our hell. And this helps us face the emotional explosion that is life.
4) We share emotions with the characters: love, anger, envy, affection, joy, etc. This happens because, through a cerebral process, we mirror the emotions we see represented in stories. We root for or against the characters. But not for them. For ourselves. It’s a kind of emotional selfishness.
5) We simulate intense experiences without having actually lived them. We travel through the range of possibilities that life can offer us, even without our physical presence—that is, without paying the price that life’s protagonism demands. We deal with challenges, embark on epic journeys, love, face powerful enemies, all from the comfort of being distant, protected observers. We also suffer, it’s true. But the most appreciated stories always end with some kind of redemption, a comeback, or an overcoming.
6) We represent concepts playfully. It’s the famous moral of the story. Every good story has a message that, even without us realizing it, works on our unconscious, forming a framework of beliefs related to the lessons received. Little Red Riding Hood teaches children not to talk to strangers. And, in my opinion, it also serves to show children the importance of developing observational skills so that they can easily recognize the difference between a grandmother and a wolf.
7) Religions, myths, and rituals have, over the millennia, helped humans come to terms with nature’s fierce dynamics. What we don’t understand provokes a deep sense of insecurity and anxiety, especially regarding life’s unfathomable mysteries. Good stories told using the fantastic, the mystical, the divine, or the sacred spirit have the power to alleviate our tensions by providing answers to our anxieties and suggesting purposes for our lives. They show, at the same time, our smallness and grandeur.
READ IN THE NEXT PART
– And what does Darwin have to do with it?
– Simple and profound
– A relevant information
– Reaching the collective unconscious
– Collective intelligence
