Webinar presenting the findings of the AI Monitoring Committee of the Forum des images

For eleven months, philosophers, artists, educators, institutional leaders, and corporate experts worked together within the Artificial Intelligence Oversight Committee, launched by the Forum des Images and TUMO Paris. I facilitated this committee using collective intelligence methods. Here is what this experience taught me, how it shifted my perspective, and the paths it opens for the cultural and educational sectors.

The Forum des Images facing AI

The Forum des Images is an institution situated at the crossroads of several fields. It offers a program of moving images (cinema, video games, virtual images, comics) intended for the general public, families, and schools, with approximately 200,000 spectators per year. It also houses TUMO Paris, an inclusive and free digital creation school for 1,200 teenagers aged 12-18 per week. The Forum is thus at the intersection of culture, education, digital technology, and technological innovation.

When generative AI appeared in 2022, it suddenly democratized access to powerful creative tools—a technological leap that altered our visions of the future. Real hopes emerged: the democratization of knowledge, the availability of creative tools for the greatest number, the personalization of education, and the delegation of tedious tasks. However, delicate questions also surfaced regarding ethics, the risk of new inequalities, environmental impacts, the status of authors, trust in images, and the stereotypes conveyed by AI.

For nearly thirty years, the Forum des Images has embraced technological innovations to test them with artists, democratize them, and above all, learn to use them responsibly in the service of empowerment and the reduction of inequalities. But thinking through this responsibility takes time. More and more artists are taking this time, sometimes under difficult conditions. The cultural sector, composed of small and medium-sized organizations, does not always have the monitoring resources of large corporations. An economic divide around AI is emerging, which could weigh heavily on the diversity of perspectives.

It was to take this time that the Forum des Images created this oversight committee. In mid-2023, we began working together. First, the TUMO x AI event, where nearly 500 TUMO students created 200 short films in small groups using AI tools, preceded by an information session and followed by a debate. Then, participation in the international AI Teens conference organized by TUMO Yerevan. Finally, one piece was missing: the addition of long-term thinking partners. This gave birth to the AI Oversight Committee, which I was asked to lead, and for which I am synthesizing the work here.

Collective Intelligence vs. Artificial Intelligence

I would like to start by sharing something personal from my childhood. I discovered artificial intelligence in the 70s and 80s. At the time, it was called “expert systems.” I was faced with these machines, these computers, these keyboards, these screens, these programming languages (BASIC, Assembly, Pascal, C, etc.). I was completely fascinated. I felt a new world opening up before me, but I felt quite alone facing it. And I can tell you that my parents did not appreciate it at all! I felt something important, but it wasn’t recognized at all.

I think this is a recurring trait of innovation. When you sense things, when you have intuitions, you are often not recognized because criteria do not yet exist. There are perhaps paths of inspiration there too. I was programming games of life, imagining chatbots, creating programs to generate poetry automatically, and creating a cultural Minitel server from a computer left on at home (10 years before the appearance of the Internet). This subject was already there—even in Ada Lovelace’s 1842 text; when she invented computing, she spoke of musical and graphic creation for the future computing that would exist a century later.

I share this memory because it sheds light on the situation of today’s teenagers facing AI. They too sense something immense, and they too are sometimes poorly supported in this intuition. This is precisely what motivated the methodological gamble of the oversight committee. From the start, I set this conviction: face to face with artificial intelligence, the relevant response is collective intelligence. Keyword mapping, collaborative mind maps, open forums, cross-disciplinary exchanges, etc. All these techniques translate our humanity through their horizontality and the opportunities for contribution they open for each and every person.

The plurality of the committee members gathered by the Forum des Images team was a great asset. Bringing together around the same table a philosopher, a social affairs inspector, a visual artist, leaders of a digital creation school, and representatives from a major tech company and the City of Paris. AI simultaneously questions copyright, working conditions, pedagogical methods, cultural representations, and digital sovereignty. No one can embrace this complexity alone. And during the eleven months of the committee, artificial intelligences moved and evolved. Many uses that did not even exist at the beginning arrived along the way. What we learned is also a method: an openness and enrichment via others to be capable of following what is happening and making better decisions.

A Non-Human Interlocutor

Among the findings of the committee, the one that surprised me most concerns what I would call the “psyche of civilization.” For the first time, we have a non-human interlocutor. It is not just a machine that does what it is told; it is an interlocutor in the true sense. Are we the only intelligent entities? Listening to the committee members, I realized this question was present in all fields, from artistic creation to education, from ethics to professions. I don’t have an answer, but it is a deeper question than I had initially imagined.

This observation aligns with a concept I am developing elsewhere: the “displaced us.” Artificial intelligence is no longer a machine built for a precise and delimited productive purpose, like a plow, a loom, or an assembly line. The contemporary machine models itself on us—on us as singular individuals and on us as a linguistic and cultural community—through its autonomous learning of our languages. It is “us,” philosophically, culturally, and cognitively. It is not a simple imitation. It is a “displaced us,” an “us” yet beside us in ontological terms, hence the great unease. And this proximity opens possibilities that the discourse of fear fails to perceive.

As humans, then, what characterizes us, and what should we cultivate? This is what I tried to experiment with in the facilitation method of this oversight committee. We experienced together in the committee, through our way of organizing and practicing collective intelligence, that what characterizes us is empathy, improvisation, the unexpected, and creativity. For me, this is central. And these are the things to be cultivated. Enriching ourselves through our otherness, listening to one another, accepting disagreement, and being enriched by our disagreements. I can tell you that we laughed a lot sometimes, and sometimes we argued too. And that is a good thing!

AI as a Mirror of Our Societies

The work of Karin Crona, a visual artist and committee member, made something visible that we knew only vaguely. By asking an AI to produce selfies from identical prompts but with different geographical origins, biases appeared flagrantly. A white Western woman will be represented smiling, while a woman of African or Asian origin will be systematically stereotyped, sometimes hypersexualized. AI does not create these biases; it reveals and amplifies them. It forces us to look our world in the face. Karin says it herself: “I am very enthusiastic and easily seduced by AI. Sometimes I don’t want to see the negative sides. The committee helped me develop my critical thinking.”

I had already experienced this mirror function in my own experiments with drones about ten years ago. These computer objects, devoid of consciousness but packed with algorithms, reflect an image of our own programming logic and technological blind spots, appearing before us as a non-human presence in our reality. With generative AI, this confrontation takes on a new dimension, because it is our collective representations, sedimented in billions of data points, that return to us in the form of images and texts.

As Claude Farge, Director General of the Forum des Images, stated in the webinar conclusion, AI holds up a mirror that is often a distorting one to our own reality. And it is perhaps the first time that a technology has questioned to such an extent, collectively and across all cultures, what it means to be human.

“Panser” (Healing/Thinking) Technology

The late philosopher Bernard Stiegler spoke of “panser” (a French pun on *penser* [to think] and *panser* [to heal/dress a wound]) technology, as in taking care of it. The committee confirmed this intuition for me. Dualism—being for or against, engaging in opinion debates—is not, in my view, the way forward. The feedback from committee members confirmed this. We are dealing with concrete things, and that is what matters.

This conviction found an echo in the concept of “philotechny” proposed by Vanessa Nurock, Professor of Philosophy at Université Côte d’Azur and committee member. A demanding love of technology, a critical love that falls into neither technophilia nor technophobia. She also introduced the notion of “polethics” (policy/politics + ethics + poetry), borrowed from the poet Michel Deguy, which aims to articulate politics, ethics, and poetry. The questions AI poses are neither purely technical, nor purely moral, nor purely political. They are all of these at once, which is why the presence of an artist in the committee was vital.

Vanessa Nurock also formulated a methodological proposal that guided our work: “ethics by design.” This means integrating ethical reflection at every step of the process, from conception onward, rather than as a veneer added later. This approach, linked to the idea of *care*, shifts the field of ethics toward concrete questions: what matters to us? Are we taking care of the most vulnerable, and notably future generations?

I advocate for a position that joins this one. Rather than enthusiastic and blind adoption, or conscientious objection that keeps its distance, I believe in engaging in a critical and creative practice. Learning to untangle the web so it doesn’t stifle us requires getting our hands in it—experiencing these tools concretely within collective frameworks that allow for reflection.

Thinking by Tasks Rather than by Jobs

Faced with fears of job destruction, the committee developed a finer analytical grid. Louis-Charles Viossat, Inspector General of Social Affairs, proposed reasoning in terms of tasks rather than jobs. Some tasks are automatable, others can be augmented by AI, and still others remain intrinsically human. One cannot say, for example, that the job of a screenwriter will disappear. One must consider that certain research tasks within that job will be transformed or augmented by generative AI.

This observation leads to an important point. The effects of AI will likely be more significant on working conditions than on the quantity of jobs. Algorithmic surveillance, the weakening of interpersonal relationships, and management by machines are all concerns we must be vigilant about. Louis-Charles Viossat also pointed out that AI can offer more autonomy through skill upgrades while simultaneously creating new forms of dependence. His conclusion is a political question: it is not a matter of whether AI will liberate or alienate us, but who decides what, and how we create the conditions for a real choice.

This task-based approach aligns with my reflections on “disseminated intelligences.” Rather than a centralized super-brain replacing humans, the future is shaping up as an articulation between multiple forms of intelligence—local, specialized, and in permanent dialogue. Human work does not disappear; it is recomposed. And this recomposition requires people to acquire new, specifically human skills that escape automation.

Thomas Scotto d’Abusco, a programming trainer at TUMO Paris, found a good metaphor: the chef in a great restaurant. To become a chef, you have to start by washing dishes, chopping carrots, making sauces. Someone who takes the shortcut of delegating everything to AI might quickly get prototypes that work but would find themselves stuck when faced with technical errors. Whereas someone who has learned the basics can take over when the AI makes a mistake, hallucinates, or loses context. The human becomes the guarantor of quality, the corrector of algorithmic errors. And Thomas raises a point I find very true: if our only skill is writing prompts, as soon as another “prompter” arrives, they will know how to do exactly the same thing. What is important is building know-how that makes us different from one another.

The “Paranoid Optimism” of Teenagers

Delphine Stucchi, representing the French National Education within the committee, noted that according to a Senate report, 90% of high school sophomores already use generative AI, compared to only 20% of teachers. This gap poses a real challenge. The committee identified a necessary dual approach: teaching ABOUT AI (developing critical thinking, identifying biases, understanding mechanisms) and teaching WITH AI (using these tools to personalize learning, support students in difficulty, differentiate pathways). But in all cases, as Delphine Stucchi reminded us, no technology can supplant the art of pedagogy.

What struck me most in this adventure was the lucidity of the teenagers. Pegor Papazian, co-founder and head of development at TUMO, shared the results of the international AI/Teens conference, which in 2025 brought together 90 teenagers from 16 cities worldwide for 24 hours, following the sun from Tokyo to Los Angeles. What the organizers discovered surprised them. These teens neither rejected AI nor glorified it. They were very rational, showing a maturity often absent in adult discourse.

A participant from Mumbai perhaps summarized their position best: “Using AI is like using a cheat code in a video game. But that doesn’t mean I’m letting go of the controller.” And when they asked themselves if AI could replace friends, their answer was clear: “AI can alleviate loneliness, but it cannot offer presence—being there in discomfort, without a solution, just with someone.” A friend of Pegor Papazian found the right phrase: “paranoid optimism”—an acute awareness of the risks coupled with a will to learn and master these technologies. These young people are not fooled. This expertise must be recognized and valued.

Pegor Papazian also asked a very pertinent question. The better AI becomes at solving problems, designing buildings, and writing texts, the less motivated we are to learn those things. If AI can fix your bike, you are less motivated to learn how to do it yourself. School systems have always relied on motivation mechanisms like diplomas, grades, and praise. All of this is threatened by AI. It is an educational challenge that calls for creative responses rather than defensive stances.

Imperfection as a Path to Creativity

Claude Farge said it from the start of the committee: he preferred the imperfections of the early DALL-E over the perfect results of Midjourney. And I believe this dimension of imperfection, of trial and error, matters a great deal. Biologist Olivier Hamant, in his book *Against the Cult of Performance*, speaks of “robustness” through recycling, trial and error, experimentation, and stepping aside—things where we try, we fail, and we learn from one another.

Claude Farge made a charming confession during the final presentation. Having forgotten to thank the technical team in his introduction, he noticed that if he had entrusted his speech to an AI, it would have been perfect. He then continued the thought: if we replaced management with an AI, who would employees turn to when they are dissatisfied? Who would be responsible? We know that among humanity’s greatest inventions, it was errors that led the way. Generative models are becoming more and more perfect, and therefore perhaps less and less innovative. We will have to learn to use them, or even subvert them, so they generate disruption and inventiveness.

This reflection joins my own work on the subject. As Sophie Nordmann writes in *The Philosopher’s Vocation* (2025), what distinguishes human thought from other forms of intelligence is not a “positive” something, but precisely its capacity to open breaches rather than combine, manipulate, or produce data. Artificial intelligence can structure a strategy, but it does not feel the consequences of that strategy. It can produce an answer, but it bears no responsibility. This limit is not a defect to be corrected; it is constitutive of what these machines are.

From Cultural Democratization to Cultural Democracy

To conclude, I would like to share something I find very important in the field of culture, which the committee reinforced in me. There is an opposition we can now formulate quite clearly between *cultural democratization*—in the tradition of Malraux, bringing the great works of humanity to the greatest number—and *cultural democracy*—that is, more horizontal cultural spaces where we enrich one another with our respective skills, with less hierarchy and more openness to others. This is what is carried by “cultural rights.”

Artificial intelligence, given the capabilities it can bring to everyone if used well, can be a lever at the service of cultural democracy and therefore democracy itself. This power to act passes, once again, through empathy and creativity. The anthropological challenge represented by AI—this mutation I call a “displaced us”—requires from us a critical presence, not a withdrawal. Faced with a technology that claims to centralize intelligence, let us cultivate ecosystems of disseminated thought.

Concrete Commitments

Claude Farge announced several concrete institutional commitments for the Forum des Images:

• Sustain this oversight committee to be supported over the long term,
• Explore the creation of a local AI server using open-source solutions that are less energy-intensive,
• Train mediators and staff, particularly on critical aspects,
• Create a sort of “NutriScore” indicating to the public the use of AI in cultural offerings,
• And develop round tables and participatory formats with the public around these issues.
• He also insisted on the importance of social dialogue, as it is impossible and undesirable to force such changes through.

Among the most important lessons from the committee, I take away the need to create multidisciplinary ethics committees in every cultural and educational organization—spaces for dialogue bringing together internal and external stakeholders to question practices without judgment. I recommend that these sessions notably NOT be prepared with the help of AI, to guarantee authentically human reflection. It also seems essential to give young people the floor so they can formulate their own recommendations, rather than always speaking for them. And to develop “soft skills”—creativity, empathy, critical thinking—which remains the human prerogative and becomes all the more precious in the age of automation.

It is not intelligence that makes our humanity

It is not intelligence that makes our humanity; it is our ability to be moved together. Faced with artificial intelligences, our humanity is no longer defined by our cognitive capacities but by our capacity to create links, to welcome the other, to resist domination together, and to collectively invent life experiences that transform us.

AI is a mirror, but it can also be a lever. On the condition that we never lose sight of what distinguishes us from machines: empathy, the capacity for connection, attention to others. Humanizing AI rather than robotizing humans—that is the formula that perhaps best summarizes these eleven months of shared reflection.

Watch the video of the presentation on the Forum des images website: https://www.forumdesimages.fr/restitution-du-comite-de-veille-ia

Appendix: Summary of the Restitution on January 29, 2026

What follows is a factual summary of the interventions by each committee member during the restitution webinar, in the order of speaking.

Claude Farge, Director General of the Forum des Images

Claude Farge opened the restitution by reminding us of the committee’s ambition: to reflect on the major challenges of AI while translating them into concrete actions for the Forum des Images. He thanked the institutional partners, the City of Paris, Salesforce (which supported the committee), and TUMO Yerevan. In conclusion, he presented six avenues of work: sustaining the ethics committee, growing AI workshops in new formats, installing a local server with open-source AIs, training staff on critical aspects, creating an AI NutriScore, and developing participatory formats. He stressed the need for social dialogue (CSE) in these transformations and reflected on error and creativity, noting that models are becoming too perfect and perhaps less innovative, requiring us to subvert them for inventiveness.

Séverine Le Bescond, Deputy Director General of the Forum des Images

Séverine Le Bescond presented the institutional context. She described the hopes (democratization, personalization) and concerns (ethics, environmental impact, copyright) sparked by generative AI. She traced the Forum’s steps: staff training in 2024, the TUMO x IA event with 500 students, participation in AI Teens, and the progressive integration of AI into cinema programming. She emphasized that the cultural sector needs “long time” for reflection, which the committee provided.

Benoît Labourdette, Filmmaker, Pedagogue, Committee Facilitator

I presented the collective intelligence method and the four themes explored (art, ethics, jobs, education). I shared a personal memory of 80s programming to highlight the situation of today’s teens. I argued that having a non-human interlocutor forces us to define what makes us human—specifically empathy and improvisation. I cited Stiegler’s “panser la technique” and proposed AI as a tool for cultural democracy.

Vanessa Nurock, Professor of Philosophy, UNESCO Chair

Vanessa Nurock stated that philosophy helps structure questions rather than just dictate rules. she identified three difficulties: the Collingridge dilemma, the blurring of concepts (natural/artificial), and non-consensual social experimentation. She proposed “polethics” (politics + ethics + poetry) and “ethics by design” (theoretical/practical back-and-forth, integration of care, reversibility for future generations), concluding with “philotechny”—a critical love of technology.

Delphine Stucchi, Representative of National Education

Delphine Stucchi noted the gap between students (90% use AI) and teachers (20%). She identified two challenges: pedagogical fraud vs. pedagogical integrity. She presented the Ministry’s response: teaching ABOUT AI (critical thinking, bias) and teaching WITH AI (GDPR-compliant tools for evaluation and supporting students with disabilities). She affirmed that pedagogy must always come before technology.

Pegor Papazian, Co-founder and Head of Development at TUMO

Pegor Papazian compared humanity facing AI to a lottery winner whose identity is threatened. He pointed out the motivation paradox: if AI can do it, why learn it? He shared the “paranoid optimism” of teens from the AI Teens conference—rational, mature, and aware that while AI can alleviate loneliness, it cannot offer human presence in discomfort.

Thomas Scotto d’Abusco, Programming Trainer, TUMO Paris

Thomas addressed the question: why code if AI can? He used the chef analogy: you need the basics to correct AI when it “hallucinates.” He noted positive uses (translation, simplification) but also the ecological cost. He warned against the “all prompters” risk—if we only prompt, we become interchangeable.

Louis-Charles Viossat, Inspector General of Social Affairs

Louis-Charles focused on work, suggesting we look at tasks rather than jobs. He warned that AI’s impact on working conditions (algorithmic surveillance) might be greater than job losses. He cited the Hollywood strikes as an example of social dialogue and concluded that the real question is who decides how these tools are used.

Yang Liu-Riess, Solution Architect, Salesforce

Yang shared practical experience on the “augmented company.” AI shouldn’t replace employees but equip them to spend less time on repetitive tasks (e.g., automatic meeting notes). She emphasized that tools must meet safety standards and that the future lies in complementarity between humans, tools, and clients.

Karin Crona, Multidisciplinary Artist

Karin has used Midjourney since 2022. Rather than being angry at AI stereotypes, she used humor to “break the biases” creatively. She documented the evolution of biases (the identical CEO, wheelchairs representing all disabilities) and concluded that AI is an assistant that should not be used alone.