Note de contexte : Dernier article de Sam Altman, 10 juin 2025. Traduit par ChatGPT-4o et vérifié (et corrigé) par Gemini 2.5 Pro
Traduction :
La Singularité Douce
Nous avons dépassé l’horizon des événements ; le décollage a commencé. L’humanité est proche de construire une superintelligence numérique, et jusqu’à présent, cela semble bien moins étrange que ce à quoi on aurait pu s’attendre.
Les robots ne déambulent pas encore dans les rues, et la plupart d’entre nous ne passons pas la journée à parler avec une IA. Les gens meurent toujours de maladies, nous ne pouvons toujours pas facilement aller dans l’espace, et beaucoup de choses dans l’univers nous échappent encore.
Et pourtant, nous avons récemment construit des systèmes qui surpassent les humains à bien des égards, capables d’amplifier considérablement le travail des personnes qui les utilisent. La partie la moins probable du chemin est derrière nous ; les percées scientifiques ayant conduit à des systèmes comme GPT-4 et o3 ont été durement gagnées, mais elles nous porteront très loin.
L’IA contribuera au monde de nombreuses manières, mais les gains en qualité de vie dus à l’accélération du progrès scientifique et à l’augmentation de la productivité seront immenses : le futur peut être bien meilleur que le présent. Le progrès scientifique est le principal moteur de tout progrès ; il est profondément enthousiasmant d’imaginer combien plus nous pourrions avoir.
D’une certaine manière fondamentale, ChatGPT est déjà plus puissant que n’importe quel être humain ayant jamais vécu. Des centaines de millions de personnes s’y fient chaque jour, pour des tâches de plus en plus importantes. Une petite capacité nouvelle peut générer un impact immensément positif ; un petit défaut d’alignement multiplié par des centaines de millions de personnes peut avoir un impact négatif majeur.
L’année 2025 a vu l’arrivée d’agents capables de véritable travail cognitif ; écrire du code informatique ne sera plus jamais pareil. 2026 verra probablement l’arrivée de systèmes capables de découvrir des idées nouvelles. Et 2027 pourrait voir l’arrivée de robots capables d’agir dans le monde réel.
Beaucoup plus de gens pourront créer du logiciel, et de l’art. Mais le monde en demande beaucoup plus, et les experts resteront sans doute meilleurs que les novices, à condition d’adopter les nouveaux outils. De manière générale, la capacité d’une seule personne à produire beaucoup plus en 2030 qu’en 2020 représentera un changement frappant, et beaucoup sauront en tirer profit.
Sous certains aspects essentiels, les années 2030 pourraient ne pas sembler radicalement différentes. Les gens aimeront toujours leur famille, exprimeront leur créativité, joueront à des jeux, et nageront dans des lacs.
Mais sous d’autres aspects tout aussi importants, les années 2030 seront probablement radicalement différentes de tout ce qui a précédé. Nous ne savons pas jusqu’à quel point nous pouvons dépasser l’intelligence humaine, mais nous allons bientôt le découvrir.
Dans les années 2030, l’intelligence et l’énergie — les idées, et la capacité à les concrétiser — vont devenir extraordinairement abondantes. Ces deux facteurs ont longtemps été les limites fondamentales du progrès humain ; avec une abondance d’intelligence et d’énergie (et une bonne gouvernance), tout devient théoriquement accessible.
Nous vivons déjà avec une intelligence numérique remarquable, et après un choc initial, la plupart d’entre nous s’y sont habitués. Très vite, on passe de l’émerveillement devant un paragraphe bien écrit généré par une IA, à se demander quand elle écrira un roman entier ; de l’émerveillement devant un diagnostic médical salvateur, à se demander quand elle inventera le traitement ; d’un petit programme informatique, à une entreprise complète. C’est ainsi que se manifeste la singularité : les merveilles deviennent la norme, puis des exigences de base.
Les scientifiques disent déjà qu’ils sont deux à trois fois plus productifs qu’avant l’IA. L’IA avancée est fascinante pour de nombreuses raisons, mais peut-être rien n’est aussi significatif que le fait de pouvoir s’en servir pour accélérer la recherche en IA elle-même. Nous pourrions découvrir de nouveaux supports de calcul, de meilleurs algorithmes, et qui sait quoi d’autre. Si une décennie de recherche peut être accomplie en un an, voire en un mois, alors le rythme du progrès va évidemment changer.
À partir de maintenant, les outils que nous avons déjà construits vont nous aider à découvrir de nouvelles idées scientifiques et à créer de meilleurs systèmes d’IA. Ce n’est pas exactement une IA qui met à jour son propre code de façon autonome, mais c’en est néanmoins une version larvaire d’amélioration récursive.
D’autres boucles d’auto-renforcement sont également à l’œuvre. La valeur économique générée a enclenché un cercle vertueux d’investissement dans les infrastructures nécessaires pour faire tourner ces IA de plus en plus puissantes. Et les robots capables de fabriquer d’autres robots (et, d’une certaine manière, les centres de données capables d’en construire d’autres) ne sont plus très loin.
Même s’il faut produire le premier million de robots humanoïdes « à l’ancienne », une fois qu’ils peuvent gérer l’ensemble de la chaîne logistique — extraction et raffinage des minerais, transport, usines, etc. — pour construire d’autres robots, lesquels construiront des usines de puces, des centres de données, etc., alors le rythme du progrès changera radicalement.
À mesure que la production de datacenters sera automatisée, le coût de l’intelligence devrait tendre vers celui de l’électricité. (On se demande souvent combien d’énergie consomme une requête ChatGPT : environ 0,34 watt-heure, soit ce qu’un four utilise en un peu plus d’une seconde, ou une ampoule LED efficace en quelques minutes. Cela utilise aussi environ 0,000085 gallon d’eau, soit environ un quinzième de cuillère à café.)
Le rythme du progrès technologique continuera d’accélérer, et l’humanité continuera de s’y adapter. Il y aura des moments très difficiles — des catégories entières d’emplois disparaîtront — mais le monde deviendra aussi tellement plus riche, tellement plus vite, que nous pourrons sérieusement envisager des politiques autrefois impensables. Nous n’adopterons probablement pas un nouveau contrat social d’un seul coup, mais dans quelques décennies, les changements progressifs auront produit quelque chose de fondamental.
Si l’histoire est un guide, nous trouverons de nouvelles choses à faire, de nouveaux désirs, et nous assimilerons rapidement les nouveaux outils (le changement d’emploi après la révolution industrielle est un bon exemple récent). Les attentes augmenteront, mais les capacités aussi, et nous bénéficierons tous de meilleures choses. Nous créerons des merveilles les uns pour les autres. L’être humain conserve un avantage fondamental sur l’IA : nous sommes câblés pour nous soucier profondément des autres, de ce qu’ils pensent et font, et nous nous soucions peu des machines.
Un agriculteur de subsistance vivant il y a mille ans verrait nos activités d’aujourd’hui comme des « faux métiers », et penserait que nous jouons à des jeux pour nous divertir, ayant déjà assez de nourriture et des luxes inimaginables. J’espère que nous verrons les métiers d’ici mille ans comme des « faux métiers », tout en sachant qu’ils paraîtront profondément importants et satisfaisants à ceux qui les exerceront.
Le rythme des merveilles à venir sera immense. Il est difficile aujourd’hui d’imaginer ce que nous aurons découvert d’ici 2035 : peut-être passerons-nous de la résolution de la physique des hautes énergies une année à la colonisation spatiale l’année suivante ; ou d’une percée en science des matériaux à une véritable interface cerveau-machine à large bande passante. Beaucoup vivront leur vie comme aujourd’hui, mais certains choisiront sûrement de « se brancher ».
Ce qui vient est difficile à conceptualiser. Mais y vivre sera probablement impressionnant tout en restant gérable. D’un point de vue relativiste, la singularité se produit pas à pas, et la fusion se fait lentement. Nous montons la pente de la courbe exponentielle du progrès technologique : elle semble toujours verticale vers l’avant, et plate en arrière, mais elle forme une courbe fluide et continue. (Pensez à ce que 2020 aurait pensé d’une AGI proche en 2025, comparé à ce que nous avons vraiment vécu en cinq ans.)
Il y aura des défis majeurs à relever, en plus des immenses avantages. Nous devons résoudre les problèmes de sécurité, à la fois techniques et sociétaux, mais il est ensuite crucial de distribuer largement l’accès à la superintelligence, vu ses implications économiques. La meilleure voie pourrait ressembler à ceci :
Résoudre le problème de l’alignement, c’est-à-dire parvenir à garantir de manière robuste que les systèmes d’IA apprennent et agissent conformément à ce que nous voulons collectivement sur le long terme (les fils d’actualité des réseaux sociaux sont un exemple d’IA non alignée : les algorithmes y sont excellents pour capter votre attention et comprendre vos préférences à court terme, mais ils exploitent des biais cognitifs qui contournent vos préférences à long terme).
Ensuite, faire en sorte que la superintelligence soit bon marché, largement disponible, et non concentrée entre les mains d’une seule personne, entreprise ou pays. La société est résiliente, créative, et s’adapte vite. Si nous pouvons mobiliser la volonté et la sagesse collectives, alors même si nous faisons beaucoup d’erreurs, et que certaines choses tournent très mal, nous apprendrons et nous adapterons rapidement, et saurons tirer le maximum du positif, en minimisant le négatif. Donner aux utilisateurs beaucoup de liberté, dans des limites générales définies par la société, semble essentiel. Plus tôt le monde commencera à discuter de ces limites et de la manière de définir un alignement collectif, mieux ce sera.
Nous (l’industrie dans son ensemble, pas seulement OpenAI) construisons un cerveau pour le monde. Il sera extrêmement personnalisé et facile à utiliser pour chacun ; nous ne serons limités que par les bonnes idées. Longtemps, les ingénieurs des startups ont méprisé les « gars à idées », ceux qui avaient une idée sans équipe pour la concrétiser. Il me semble désormais que leur moment de gloire est arrivé.
OpenAI est beaucoup de choses aujourd’hui, mais avant tout, c’est une entreprise de recherche sur la superintelligence. Il nous reste beaucoup à faire, mais la majeure partie du chemin est désormais éclairée, et les zones d’ombre reculent rapidement. Nous sommes extraordinairement reconnaissants de pouvoir faire ce que nous faisons.
Une intelligence trop bon marché pour être mesurée est désormais à portée de main. Cela peut paraître fou, mais si nous vous avions dit en 2020 que nous en serions là aujourd’hui, cela aurait semblé encore plus fou que nos prédictions actuelles pour 2030.
Puisse notre montée vers la superintelligence être fluide, exponentielle et sans heurts.
Article original : https://blog.samaltman.com/the-gentle-singularity
The Gentle Singularity
We are past the event horizon; the takeoff has started. Humanity is close to building digital superintelligence, and at least so far it’s much less weird than it seems like it should be.
Robots are not yet walking the streets, nor are most of us talking to AI all day. People still die of disease, we still can’t easily go to space, and there is a lot about the universe we don’t understand.
And yet, we have recently built systems that are smarter than people in many ways, and are able to significantly amplify the output of people using them. The least-likely part of the work is behind us; the scientific insights that got us to systems like GPT-4 and o3 were hard-won, but will take us very far.
AI will contribute to the world in many ways, but the gains to quality of life from AI driving faster scientific progress and increased productivity will be enormous; the future can be vastly better than the present. Scientific progress is the biggest driver of overall progress; it’s hugely exciting to think about how much more we could have.
In some big sense, ChatGPT is already more powerful than any human who has ever lived. Hundreds of millions of people rely on it every day and for increasingly important tasks; a small new capability can create a hugely positive impact; a small misalignment multiplied by hundreds of millions of people can cause a great deal of negative impact.
2025 has seen the arrival of agents that can do real cognitive work; writing computer code will never be the same. 2026 will likely see the arrival of systems that can figure out novel insights. 2027 may see the arrival of robots that can do tasks in the real world.
A lot more people will be able to create software, and art. But the world wants a lot more of both, and experts will probably still be much better than novices, as long as they embrace the new tools. Generally speaking, the ability for one person to get much more done in 2030 than they could in 2020 will be a striking change, and one many people will figure out how to benefit from.
In the most important ways, the 2030s may not be wildly different. People will still love their families, express their creativity, play games, and swim in lakes.
But in still-very-important-ways, the 2030s are likely going to be wildly different from any time that has come before. We do not know how far beyond human-level intelligence we can go, but we are about to find out.
In the 2030s, intelligence and energy—ideas, and the ability to make ideas happen—are going to become wildly abundant. These two have been the fundamental limiters on human progress for a long time; with abundant intelligence and energy (and good governance), we can theoretically have anything else.
Already we live with incredible digital intelligence, and after some initial shock, most of us are pretty used to it. Very quickly we go from being amazed that AI can generate a beautifully-written paragraph to wondering when it can generate a beautifully-written novel; or from being amazed that it can make live-saving medical diagnoses to wondering when it can develop the cures; or from being amazed it can create a small computer program to wondering when it can create an entire new company. This is how the singularity goes: wonders become routine, and then table stakes.
We already hear from scientists that they are two or three times more productive than they were before AI. Advanced AI is interesting for many reasons, but perhaps nothing is quite as significant as the fact that we can use it to do faster AI research. We may be able to discover new computing substrates, better algorithms, and who knows what else. If we can do a decade’s worth of research in a year, or a month, then the rate of progress will obviously be quite different.
From here on, the tools we have already built will help us find further scientific insights and aid us in creating better AI systems. Of course this isn’t the same thing as an AI system completely autonomously updating its own code, but nevertheless this is a larval version of recursive self-improvement.
There are other self-reinforcing loops at play. The economic value creation has started a flywheel of compounding infrastructure buildout to run these increasingly-powerful AI systems. And robots that can build other robots (and in some sense, datacenters that can build other datacenters) aren’t that far off.
If we have to make the first million humanoid robots the old-fashioned way, but then they can operate the entire supply chain—digging and refining minerals, driving trucks, running factories, etc.—to build more robots, which can build more chip fabrication facilities, data centers, etc, then the rate of progress will obviously be quite different.
As datacenter production gets automated, the cost of intelligence should eventually converge to near the cost of electricity. (People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second, or a high-efficiency lightbulb would use in a couple of minutes. It also uses about 0.000085 gallons of water; roughly one fifteenth of a teaspoon.)
The rate of technological progress will keep accelerating, and it will continue to be the case that people are capable of adapting to almost anything. There will be very hard parts like whole classes of jobs going away, but on the other hand the world will be getting so much richer so quickly that we’ll be able to seriously entertain new policy ideas we never could before. We probably won’t adopt a new social contract all at once, but when we look back in a few decades, the gradual changes will have amounted to something big.
If history is any guide, we will figure out new things to do and new things to want, and assimilate new tools quickly (job change after the industrial revolution is a good recent example). Expectations will go up, but capabilities will go up equally quickly, and we’ll all get better stuff. We will build ever-more-wonderful things for each other. People have a long-term important and curious advantage over AI: we are hard-wired to care about other people and what they think and do, and we don’t care very much about machines.
A subsistence farmer from a thousand years ago would look at what many of us do and say we have fake jobs, and think that we are just playing games to entertain ourselves since we have plenty of food and unimaginable luxuries. I hope we will look at the jobs a thousand years in the future and think they are very fake jobs, and I have no doubt they will feel incredibly important and satisfying to the people doing them.
The rate of new wonders being achieved will be immense. It’s hard to even imagine today what we will have discovered by 2035; maybe we will go from solving high-energy physics one year to beginning space colonization the next year; or from a major materials science breakthrough one year to true high-bandwidth brain-computer interfaces the next year. Many people will choose to live their lives in much the same way, but at least some people will probably decide to “plug in”.
Looking forward, this sounds hard to wrap our heads around. But probably living through it will feel impressive but manageable. From a relativistic perspective, the singularity happens bit by bit, and the merge happens slowly. We are climbing the long arc of exponential technological progress; it always looks vertical looking forward and flat going backwards, but it’s one smooth curve. (Think back to 2020, and what it would have sounded like to have something close to AGI by 2025, versus what the last 5 years have actually been like.)
There are serious challenges to confront along with the huge upsides. We do need to solve the safety issues, technically and societally, but then it’s critically important to widely distribute access to superintelligence given the economic implications. The best path forward might be something like:
Solve the alignment problem, meaning that we can robustly guarantee that we get AI systems to learn and act towards what we collectively really want over the long-term (social media feeds are an example of misaligned AI; the algorithms that power those are incredible at getting you to keep scrolling and clearly understand your short-term preferences, but they do so by exploiting something in your brain that overrides your long-term preference).
Then focus on making superintelligence cheap, widely available, and not too concentrated with any person, company, or country. Society is resilient, creative, and adapts quickly. If we can harness the collective will and wisdom of people, then although we’ll make plenty of mistakes and some things will go really wrong, we will learn and adapt quickly and be able to use this technology to get maximum upside and minimal downside. Giving users a lot of freedom, within broad bounds society has to decide on, seems very important. The sooner the world can start a conversation about what these broad bounds are and how we define collective alignment, the better.
We (the whole industry, not just OpenAI) are building a brain for the world. It will be extremely personalized and easy for everyone to use; we will be limited by good ideas. For a long time, technical people in the startup industry have made fun of “the idea guys”; people who had an idea and were looking for a team to build it. It now looks to me like they are about to have their day in the sun.
OpenAI is a lot of things now, but before anything else, we are a superintelligence research company. We have a lot of work in front of us, but most of the path in front of us is now lit, and the dark areas are receding fast. We feel extraordinarily grateful to get to do what we do.
Intelligence too cheap to meter is well within grasp. This may sound crazy to say, but if we told you back in 2020 we were going to be where we are today, it probably sounded more crazy than our current predictions about 2030.
May we scale smoothly, exponentially and uneventfully through superintelligence.