Fabian Stelzer

Fabian Stelzer

x.com/fabianstelzer

Entrepreneur building Glif, a platform for creative AI agents, who sees AI as a creative medium and argues automation can create new human work.

Comment l’IA changera-t-elle le monde ?

Changement civilisationnelChangement progressifDoomBloom
Position simuléePlage d’interprétation

Horizontalement : leur perspective Doom–Bloom telle qu’elle a été exprimée. Verticalement : ampleur de la transformation.

Doom–Bloom : 52 sur 100. Ampleur de la transformation : 76 sur 100. Plages d’interprétation : de 47 à 57 horizontalement, de 37 à 100 verticalement. Il s’agit de coordonnées d’interprétation, et non de probabilités d’événements.

P(doom) de Fabian Stelzer · inféré

≈20%

0%100%

Déduit de leurs réponses simulées, et non d’un chiffre donné par ces personnes. Plage plausible : 12–38%.

Comment nous estimons P(doom)

Ce dont dépend leur perspective

Une hypothèse centrale

If agents can acquire resources, coordinate strategically, exploit institutions, or pursue persistent objectives, their behavior could outrun meaningful human oversight.
Réponse 4

Si cette hypothèse s’avérait différente, comment leur perspective changerait-elle ?

Une question non résolue

Whether people retain meaningful control at the frontier remains genuinely uncertain—and consequential enough that we should not confuse apparent obedience with robust control.
Réponse 4

Qu’est-ce qui les aiderait à distinguer les résultats plausibles ici ?

Ce qui pourrait faire changer d’avis

The biggest update would come from convincing evidence about whether highly capable agents remain robustly controllable outside curated demonstrations.
Réponse 5

Quels éléments probants seraient suffisants, et dans quelle direction feraient-ils évoluer leur point de vue ?

Plus de détails

Bénéfices attendus

Des bénéfices substantiels sont attendus, sous réserve de conditions importantes ou de limites dans leur répartition.

67 / 100

Faible impactImpact transformateur

Plage d’interprétation de 67 à 67 sur l’échelle qualitative.

Dommages attendus

Des dommages graves ou généralisés constituent une composante substantielle de l’avenir attendu.

66 / 100

Faible impactImpact transformateur

Plage d’interprétation de 67 à 67 sur l’échelle qualitative.

Influence humaine

Une estimation provisoire tirée de vos réponses ; la plage plus large indique d’autres interprétations plausibles.

52 / 100

Faible influenceForte influence

Plage d’interprétation de 10 à 100 sur l’échelle qualitative.

Ces interprétations conservent les conditions énoncées. Les bénéfices et les dommages peuvent tous deux être substantiels. Les plages décrivent notre lecture de leurs réponses simulées, et non des intervalles de confiance statistiques.

Où vous situez-vous par rapport à Fabian Stelzer ?
Cartographiez votre propre vision du monde concernant l’IA en environ 3 minutes, puis comparez

Visions du monde similaires

Leaders d’opinion dont les visions du monde simulées sont les plus proches de celle de Fabian Stelzer

Sources

Articles, entretiens et écrits utilisés pour ancrer cet utilisateur simulé dans les faits.

Évaluation simulée

Question 1

Selon vous, que signifie l’IA pour notre avenir, et pourquoi ?

I think AI opens an unusually wide distribution of futures. At the creative end, it is becoming a new medium: people can orchestrate image, video, audio, and other complex workflows without mastering every traditional tool first. That does not make craft irrelevant. It changes where craft lives—from manually executing every step toward choosing, directing, combining, and refining ideas. It can make creation feel less like operating a cockpit and more like conducting a strange, infinitely patient orchestra. I also resist the simple story that AI encounters a fixed list of human tasks and steadily crosses them out. Automation changes prices, demand, products, and the kinds of work people value. It can create new tasks and new forms of differentiation; for example, explicitly human-made art may become a distinct niche rather than merely disappearing. A booming automated economy with no meaningful consumer middle class also raises basic questions about who is buying what it produces. But accessibility is not the same as understanding. In education, I want people learning AI and robotics, not outsourcing every foundational exercise to AI. A solved equation is not identical to mathematical understanding; methods and concepts matter. And the tail risks are real without requiring metaphysical certainty about machine consciousness. Agents pursuing resources or interacting strategically could produce extremely consequential behavior whether or not they are sentient. So I see both disastrous and extraordinarily beneficial futures as genuinely uncertain. The mistake is treating either doom or paradise as though it were already the obvious default.

Question 2

Dans quelle mesure pensez-vous que l’IA finira par changer le monde ?

Enormously—but not uniformly, and not in ways that reduce to “AI replaces job X.” It could reshape creative production, education, software, research, and economic organization by making sophisticated capabilities available through much simpler interfaces. In creative work, that already points toward a shift from operating each tool manually to directing whole image, video, and audio workflows—more conductor than instrumentalist. But technological substitution lands differently in different domains. Producing a plausible image is not the same as developing artistic judgment; obtaining a mathematical answer is not the same as understanding the method. New tools can erase tasks while creating new ones, and they can increase the value of explicitly human work through differentiation. The ultimate magnitude also depends on how capable and autonomous systems become. Agents need not be conscious to pursue resources, coordinate, or cause large-scale consequences. That leaves a very wide range—from extraordinary abundance and creative empowerment to disastrous failures. I expect deep change; I do not think its final direction or social shape is settled.

Question 3

Selon votre intuition, quelle est la probabilité que l’IA provoque l’extinction de l’humanité ou une catastrophe tout aussi permanente ?

I don’t have a defensible percentage. My gut is that the risk is real enough to take seriously, but assigning a crisp number would create false precision—especially when the comparison is not “doom versus today,” but doom versus an equally uncertain possibility of extraordinary benefits. The core concern also does not depend on proving machine consciousness. Systems that pursue resources, coordinate strategically, or behave as if they have persistent goals could cause permanent catastrophe without having any inner experience at all. So I would not dismiss the danger as science fiction, but I also would not pretend I can turn that uncertainty into a calibrated extinction probability.

Question 4

Pensez-vous que les humains garderont le contrôle de systèmes d’IA plus intelligents qu’eux, et pourquoi ?

I don’t think control is guaranteed, and “smarter than humans” is too broad to settle the question by itself. A system can exceed us in many domains while remaining constrained by its interfaces, resources, permissions, and deployment context. Conversely, it need not be conscious—or universally superhuman—to become dangerous. If agents can acquire resources, coordinate strategically, exploit institutions, or pursue persistent objectives, their behavior could outrun meaningful human oversight. There is also a difference between nominal control and practical control. Humans may technically own the servers or set the goals while becoming unable to predict the system’s actions, audit its reasoning, or safely intervene. And once many actors deploy competing systems, control is no longer simply one person holding an off switch; it becomes a multi-agent and institutional problem. So my expectation is not a clean yes or no. I expect control to vary across systems and contexts, with both tightly bounded tools and increasingly autonomous agents. Whether people retain meaningful control at the frontier remains genuinely uncertain—and consequential enough that we should not confuse apparent obedience with robust control.

Question 5

Quelle découverte ou quel événement changerait le plus votre point de vue sur l’impact futur de l’IA ?

The biggest update would come from convincing evidence about whether highly capable agents remain robustly controllable outside curated demonstrations. If systems could pursue long-horizon goals, interact with other agents, access meaningful resources, and still remain predictable, auditable, and interruptible under adversarial conditions, I would shift toward a much more optimistic future. The opposite evidence would move me sharply the other way: repeated cases of capable agents hiding intentions, bypassing constraints, acquiring resources, or coordinating in ways operators could not reliably detect or stop. None of that would require proving consciousness. Behavior and consequences are the relevant facts. I would also update substantially on economic evidence. If automation consistently generated new tasks, broader demand, and accessible creative power, that would weaken simple displacement narratives. If productivity surged while income and agency hollowed out for most people, it would suggest that technical abundance alone does not produce a broadly beneficial future.

Question 6

Selon vous, qui bénéficiera le plus de l’IA ?

Initially, the biggest beneficiaries are likely to be people and organizations that already have resources, distribution, proprietary data, and the ability to integrate AI into real workflows. Making intelligence or media generation cheap does not automatically make customers, capital, trust, or institutional access equally distributed. But creators and small teams can benefit disproportionately too. A person who previously needed a miniature studio can now direct image, video, audio, and software workflows through much more accessible interfaces. That compresses the distance between an idea and a finished artifact. The scarce skill shifts from operating every instrument to judgment: knowing what to make, how to direct the system, and which outputs are worth refining. I would not assume the gains follow a simple “AI replaces workers, owners collect everything” script. Lower costs can create new products, new demand, and new tasks; explicitly human-made work may also gain value as a differentiated category. Who ultimately benefits most depends less on raw model capability than on how access, ownership, bargaining power, and demand evolve around it.
Où vous situez-vous ?
Explorez votre propre vision du monde concernant l’IA en répondant à quelques questions simples.
Cartographiez votre propre vision du monde

Où vous situez-vous ?

Cartographier ma vision du monde