Jean-Philippe Encausse
Shared posts
Lune : la Chine teste en secret un robot qui pourrait devancer les astronautes américains !
OpenAI unveils ChatGPT app integration feature
OpenAI Powers Up With Massive Deal for AMD Chips
Nvidia’s got some competition: Rival chip-maker AMD signed a deal with OpenAI to provide its GPUs to the AI leader. The five-year agreement will see OpenAI scoop up hundreds of thousands of AMD’s chips, or about six gigawatts worth. That’s enough to power roughly 5 million US homes. OpenAI will put the chips to work in data centers that’ll provide the staggering compute ChatGPT needs to remain everyone’s favorite therapist.
AMD, whose stock surged nearly 38% yesterday morning, said the tie-up will generate tens of billions of dollars in revenue. OpenAI will also receive warrants enabling it to buy up to 10% of AMD stock at a penny a share, to be awarded over time as certain milestones are met.
Chip Off Your Shoulder
Nvidia is the No. 1 AI chip seller, with Mizuho Securities estimating that the company controls more than 70% of the market. Nvidia’s expected to make more than six times as much revenue as AMD this year ($206 billion vs. $33 billion).
But with the need for AI power growing fast, rivals have ample room to butt in, especially as they work to offer cheaper alternatives to Nvidia’s industry-leading, but pricey, chips. And OpenAI is playing the field to get power from every corner it can as the AI industry becomes increasingly crowded and costly:
- OpenAI is working with Broadcom to create its own AI chip, which will help train its tech, as part of a $10 billion partnership. Meanwhile, OpenAI also made a $300 billion deal with cloud leader Oracle last month to secure 4.5 gigawatts of cloud computing power through the tech leader’s data centers.
- Additionally, OpenAI has a deal with Nvidia on the horizon in which Nvidia plans to invest $100 billion in OpenAI over the next decade; OpenAI is expected to use that investment in part to buy 10 gigawatts’ worth of Nvidia’s chips. But while Nvidia and OpenAI have signed a letter of intent, AMD may yet beat them to the punch. OpenAI and AMD are planning to jumpstart their arrangement by filing for regulatory approval ASAP. Plus, the AMD deal would give OpenAI the power to influence the chipmakers’ plans, which could make it a more appealing partner than Nvidia.
Bubble Trouble: OpenAI CEO Sam Altman is channeling the Kylo Ren “MORE” meme. The company has made deals totaling about $1 trillion over the past two weeks as Altman underscores the need for more computing power to meet AI’s future requirements. Including other tech companies like Alphabet and Microsoft, The Wall Street Journal reports that spending on AI has far exceeded any other past technological revolution (railroads, electricity, fiber-optic). Whether or not the industry’s a bubble on the verge of bursting, the leaders at its helm stand by the underlying tech.
The post OpenAI Powers Up With Massive Deal for AMD Chips appeared first on The Daily Upside.
Qualcomm rachète Arduino : une nouvelle ère pour l’IA embarquée et le développement open source
Qualcomm Technologies, Inc. a annoncé aujourd’hui la signature d’un accord pour acquérir Arduino, l’un des leaders mondiaux du matériel et des logiciels open source. Cette opération stratégique vise à renforcer la présence de Qualcomm dans l’écosystème des développeurs, en leur offrant un accès simplifié à ses technologies de pointe en matière de calcul, d’IA et […]
L’article Qualcomm rachète Arduino : une nouvelle ère pour l’IA embarquée et le développement open source est apparu en premier sur BlogNT : le Blog des Nouvelles Technologies.
The Silent Sea
The South Korean Netflix series The Silent Sea is packed with FUIs, from holographic displays and weapon UIs to handheld scanners, mobile interfaces, and control panels.
I was very interested to see FUI designs from a Korean production. Korean characters tend to be more compact than English, providing designers with a bit more breathing room in layouts. That said, I was surprised to find that many of the interfaces were in actually in English.
Boardroom hologram
The rotating hologram works beautifully in this boardroom setting. What an engaging way to talk through a mission. You’d structure presentations differently if this technology were available, shifting focus between the holograms and the speaker and choreographing moments of attention. It would be even more engaging if it were interactive, the ability to rotate, zoom, or highlight sections in real time would make for an incredibly immersive briefing.
View fullsize
View fullsize
View fullsize
Gun UI
There’s something about having a display on a gun that makes it feel instantly futuristic. Here, the UI lets you see ammo counts and mode settings at a glance and even switch the weapon to fire a tranquilliser.
Scanner
Here’s a classic sci-fi horror trope, a scanner that helps detect missing crew members or other lifeforms. Similar to Aliens, it serves a familiar narrative purpose, but The Silent Sea uses a high-fidelity display as opposed to a low-fi aesthetic. Check out (Alien: Isolation and Alien Romulus)
Window display
Here, we see a window looking out onto the surface of the moon, which tints to become a display. While it looks like a classic sci-fi interface, this kind of technology is available today. Smart glass and tintable surfaces are becoming more common, so what feels futuristic in the show might not be far from reality.
View fullsize
View fullsize
View fullsize
Mobile device
This one is a bit strange, it’s hard to see how the user could type messages using just three buttons and no keyboard. There is a suggestion wheel, but having to tap repeatedly to cycle through options seems tedious. That said, the interface probably isn’t meant to be scrutinized closely or looped over repeatedly. Viewed in context, it reads as something “from another time and world,” reinforcing the sense that this is technology shaped by its own unique environment.
View fullsize
View fullsize
Control panel
Here’s a series of control panel screens, showing a mix of Korean and English. I found it interesting that many of the design elements are so small and thin, especially considering the astronauts’ thick gloves. You might expect a chunkier, more simplified design to accommodate the tactile limitations. However, as long as the interactive areas are large enough, the fine detail of the display becomes less of an issue and it makes for a visually elegant interface that doesn’t feel clunky.
Wrist UI
The wrist UI is primarily used to display life stats, like oxygen levels and other vital signs, keeping the crew informed at a glance. It also doubles as an authentication device for accessing locked areas.
Reshaping Eyeballs With Electricity, No Lasers Or Cutting Required
Glasses are perhaps the most non-invasive method of vision correction, followed by contact lenses. Each have their drawbacks though, and some seek more permanent solutions in the form of laser eye surgeries like LASIK, aiming to reshape their corneas for better visual clarity. However, these methods often involve cutting into the eye itself, and it hardly gets any more invasive than that.
A new surgical method could have benefits in this regard, allowing correction in a single procedure that requires no lasers and no surgical cutting of the eye itself. The idea is to use electricity to help reshape the eye back towards greater optical performance.
The Eyes Have It
Existing corrective eye surgeries most often aim to fix problems like long-sightedness, short-sightedness, and astigmatism. These issues are generally caused by the shape of the cornea, which works with the lens in the eye to focus light on to the light-sensitive cells in the retina. If the cornea is misshapen, it can be difficult for the eye to focus at close or long ranges, or it can cause visual artifacts in the field of view, depending on the precise nature of the geometry. Technologies like LASIK reshape the cornea for better performance using powerful lasers, but also involve cutting into the cornea. The procedure is thus highly invasive and has a certain recovery time, safety precautions that must be taken afterwards, and some potential side effects. A method for reshaping the eye without cutting into it would thus be ideal to avoid these problems.
Enter the technology of Electromechanical Reshaping (EMR). As per a new paper, researchers at the University of California, Irvine, came across the idea by accident, having been looking into the moldable nature of living tissues. As it turns out, collagen-based tissues like the cornea hold their structure thanks to the attractions between oppositely-charged subcomponents. These structures can be altered with the right techniques. For example, since these tissues are laden with water, applying electricity can change the pH through electrolyzation, altering the attraction between components of the tissue and making them pliable and reformable. Once the electric potential is taken away, the tissues can be restored to their original pH balance, and the structure will hold firm in its new form.
Researchers first tested this technique out on other tissues before looking to the eye. The team were able to use EMR to reshape ears from rabbits, while also being able to make physical changes to scar tissue in pigs. These efforts proved the basic mechanism worked, and that it could have applicability to the cornea itself.
To actually effectively reshape the cornea using this technique, a sort of mold was required. To that end, researchers created a “contact lens” type device out of platinum, which was formed in the desired final shape of the cornea. A rabbit eyeball was used in testing, doused in a saline solution to mimic the eye’s natural environment. The platinum device was pushed on to the eye, and used as an electrode to apply a small electrical potential across the eyeball. This was controlled carefully to precisely change the pH to the region where the eye became remoldable. After a minute, the cornea of the rabbit eyeball had conformed to the shape of the platinum lens. With the electrical potential removed, the pH of the eyeball was returned to normal and the cornea retained the new shape. The technique was trialled on twelve eyeballs, with ten of those treated for a shortsightedness condition, also known as myopia. In the case of the myopic eyeballs, all ten were successfully corrected the cornea, creating improved focusing power that would correspond to better vision in a living patient’s eye.
While the technique is promising, great development will be required before this is a viable method for vision correction in human patients. Researchers will need to figure out how to properly apply the techniques to eyeballs that are still in living patients, with much work to be done with animal studies prior to any attempts to translate the technique to humans. However, it could be that a decade or two in the future, glasses and LASIK will be increasingly less popular compared to a quick zap from the electrochemical eye remoulder. Time will tell.
3D printing method 'grows' intricate, ultra-strong materials inside water-based gel
J'ai testé un détecteur de gaz radon et c'est flippant !
Vous connaissez le radon ? C’est ce gaz radioactif qui s’infiltre insidieusement dans nos maisons depuis le sol. Et si vous habitez en Auvergne comme moi ou en Bretagne ou dans une région où il y a du radon , vous êtes peut-être dans une zone à risque. C’est pour cela que j’ai acheté un détecteur Ecosense EcoBlu EB100 pour voir enfin comment ça se passait vraiment chez moi.
Mais avant de vous parler du produit, laissez-moi quand même vous expliquer ce qu’est le radon… Il s’agit d’un gaz radioactif naturel qui vient de la désintégration de l’uranium présent dans le sol. Il est inodore, incolore, invisible. Vous ne le sentez pas, vous ne le voyez pas, mais il est là. Et le truc, c’est qu’en Auvergne, avec notre sous-sol granitique du Massif Central, on est particulièrement exposé.
Le radon entre donc dans les maisons par les fissures, les passages de canalisations, les caves, les vides sanitaires…etc et remonte du sol pour finir par se concentrer dans les pièces fermées. Et là, c’est pas bon du tout !! Le Centre international de recherche sur le cancer a classé le radon comme cancérigène certain pour le poumon depuis 1987. En France, c’est d’ailleurs la deuxième cause de cancer du poumon après le tabac. Rien que ça ! Le radon c’est donc environ 3 000 décès par an.
Du coup, ça fait un petit moment que je sais que l’Auvergne fait partie des zones les plus touchées en France avec la Bretagne, le Limousin et la Corse, mais je m’en étais pas trop inquiété jusqu’à ce que je lise un article scientifique là dessus. J’ai donc voulu savoir à quoi moi et ma famille on s’exposait parce que bon, c’est vrai qu’en hiver, je suis pas le champion de l’aération ! Puis surtout, on est un peu impuissant face à un truc qu’on ne voit pas. Donc au moins, si on peut le mesurer, on peut agir !
Bref, l’ Ecosense EcoBlu EB100 , c’est un petit boîtier blanc de la taille d’un réveil. Vous le branchez, et en 10 minutes vous avez déjà une première lecture. Après pour un résultat vraiment fiable, il faut attendre une heure. L’écran LED affiche alors le taux de radon en temps réel, la moyenne du jour, de la semaine et du mois. C’est super simple à utiliser !
Initialisation en cours…
Y’a même une alarme qui se déclenche si les niveaux deviennent critiques. Pour info, le seuil de référence en France, c’est 300 Bq/m3 en moyenne annuelle. Au-dessus, il faut agir.
Chez moi, les premiers jours, j’étais à 450 Bq/m3 dans le salon. Carrément au-dessus du seuil donc. Du coup, j’ai fait ce qu’il faut : aérer à fond ! Et assez rapidement finalement, c’est revenu à des taux plus tolérables, autour de 100 Bq/m3, parfois moins. J’ai renforcé la ventilation dans les pièces du sous-sol car c’est par là que ça arrive et surtout j’aére tous les jours. Et ce qui est bien avec cet appareil, c’est que machinalement je pose les yeux dessus et donc j’y pense et j’ouvre les fenêtres.
Sans ça, j’en aurais pas vraiment conscience par contre, et finalement je ne saurais pas si j’aère la maison assez ou pas.
Car ce qui est important à comprendre surtout, c’est que le taux de radon varie tout le temps. Selon la météo, la saison, l’état de vos fondations, le chauffage que vous mettez…etc. En fonction de tout un tas de critères, les taux peuvent monter ou descendre. C’est assez random. Y’a des matin, je me lève et je suis à 30 Bq/m3 et d’autres matins, c’est la fiesta largement au dessus des 300… Bref, l’EcoBlu me permet de surveiller ça en continu.
Après à ce prix, c’est quand même un investissement mais ça vaut le coup si vous êtes en zone à risque et si vous avez une maison ou en rez de chaussée d’apart. Si vous êtes en étage, y’a moins de risque d’être intoxiqué par cette saloperie.
Maintenant ce qu’il faut retenir c’est que le radon, c’est pas une blague. C’est comme si vous fumiez plusieurs clopes tous les jours ! Imaginez ça sur des années, ça craint ! Et si en plus, vous fumez pour de vrai, le risque de cancer du poumon est multiplié par 20.
L’idéal c’est quand même de repasser sous la barre des 100…
Voilà, après en dehors de l’aération, vous pouvez aussi agir sur l’étanchéité de votre maison. Colmater les fissures dans les dalles, les passages de canalisations, mettre une membrane étanche dans la cave si elle est en terre battue, ventiler les vides sanitaires…etc… Bref, limiter les entrées du radon depuis le sol c’est jouable ! L’idéal, c’est évidemment une VMC qui tourne en continu et pas juste une aération ponctuelle car le radon s’accumule quand l’air ne circule pas. Ainsi, plus vous renouvelez l’air intérieur, moins le radon stagne.
Perso, j’ai installé 2 ventilations à la cave comme je vous le disais. C’est mieux que rien, mais ça ne suffit pas.
En tout cas, je ne regrette pas cet achat de détecteur temps réel. Matérialiser un danger invisible, ça change tout et maintenant, que je vois les chiffres, je sais quand il faut aérer, et je sais que je ne me pèle pas les noisettes pour rien en plein hiver. Bref, c’est rassurant surtout !
Voilà, donc si vous êtes en Auvergne, Bretagne, Limousin, Vosges, Alpes, Corse… ou n’importe quelle zone granitique, je vous conseille vivement de mesurer le radon chez vous. Vous serez peut-être surpris… au mieux, vous découvrirez que tout va bien et vous pourrez renvoyer ce détecteur à Amazon et vous faire rembourser et au pire, vous corrigerez ce problème avant qu’il ne devienne grave…
Sources :
- Informations sur le radon en France : IRSN - Le risque radon en 10 questions
- Radon en Auvergne : ARS Auvergne-Rhône-Alpes
- Classification cancérigène : Cancer Environnement - Radon et effets sur la santé
Chemistry Nobel prize awarded for building ordered polymers with metal
The polymers around us are basically a disordered mess, with long chains of atoms tangled around each other. But starting around 1990, chemists began developing techniques that allow us to build polymers with precisely defined structures. These polymers, called metal-organic frameworks (MOFs), have distinct chemical properties: large pores that can be used to filter or store gases, catalytic centers within the polymer, and more.
On Wednesday, the Nobel Prize Committee honored three researchers for their role in developing MOFs: Richard Robson for demonstrating the first MOF, and Susumu Kitagawa and Omar Yaghi for developing them to their full potential.
Building a structure
Most polymers are built from individual molecules that are linked together by flexible bonds, allowing the molecules to flop around. As a result, in their final form as something like a plastic bag or a bike tire, the polymers are a complicated tangle, with molecules wrapped around each other in no particular order. We can still control some aspects of the resulting polymer by manipulating its bulk properties or changing the molecules it's built from, but there's not a lot of chemistry we can do beyond that.
Google Pixel 10 Pro Fold review: The ultimate Google phone
When the first foldable phones came along, they seemed like a cool evolution of the traditional smartphone form factor and, if they got smaller and cheaper, like something people might actually want. After more than five years of foldable phones, we can probably give up on the latter. Google's new Pixel 10 Pro Fold retains the $1,800 price tag of last year's model, and while it's improved in several key ways, spending almost two grand on any phone remains hard to justify.
For those whose phones are a primary computing device or who simply love gadgets, the Pixel 10 Pro Fold is still appealing. It offers the same refined Android experience as the rest of the Pixel 10 lineup, with much more screen real estate on which to enjoy it. Google also improved the hinge for better durability, shaved off some bezel, and boosted both charging speed and battery capacity. However, the form factor hasn't taken the same quantum leap as Samsung's latest foldable.
An iterative (but good) design
The Pixel 10 Pro Fold doesn't reinvent the wheel—it looks and feels almost exactly like last year's foldable, with a few minor tweaks centered around a new "gearless" hinge. Dropping the internal gears allegedly helps make the mechanism twice as durable. Google claims the Pixel 10 Pro Fold's hinge will last for more than 10 years of folding and unfolding.
Sam Altman Says Copyright Holders Are Begging for Their Characters to Be Included in Sora
OpenAI’s latest text-to-video generating app, Sora 2, exploded onto the scene late last week, quickly devolving into a mind-numbing mess of AI slop.
Before OpenAI stemmed the flood with crudely implemented guardrails, users immediately started generating footage featuring copyrighted materials, from SpongeBob SquarePants cooking up meth to CEO Sam Altman grilling a photorealistic Pikachu.
The spread of copyrighted material on the app sparked a debate surrounding the mass infringement of protected intellectual property, which has already led to major Hollywood studios coming after OpenAI’s competitors.
But despite the app’s messy launch, Altman has remained adamant that rightsholders are already lining up to have their IP featured on the platform.
During a newly-released chat with the a16z podcast‘s cofounder Ben Horowitz and general partner Erik Torenberg, Altman said that “in the case of Sora, we’ve heard from a lot of concerned rightsholders and also a lot of rightsholders who are like ‘My concern is you won’t put my character in enough.'”
While they allegedly told him that they want “restrictions” so their characters wouldn’t “say some crazy offensive thing,” Altman says copyright holders want “people to interact” and to “develop the relationship” to make their franchises become “more valuable.”
“So I can completely see a world where subject to the decisions that a rightsholder has, they get more upset with us for not generating their character often enough than too much,” he told Horowitz and Torenberg.
It’d be interesting to know the balance of rightsholders who have reached out to OpenAI who are upset that their characters are being featured on the app versus ones who want greater visibility on the platform. Major content producers have historically bristled at having their IP reproduced without authorization.
A litany of lawsuits aimed at AI image generator Midjourney and embattled AI chatbot company Character.AI suggest that OpenAI may be facing an uphill battle to get rightsholders on its side. Case in point, last week, Character.AI announced that it was removing chatbots directly inspired by Disney characters from its platform after receiving a cease-and-desist letter from the media juggernaut.
In short, the company’s launch of Sora 2 seems to have preceded working out many of the logistics.
The mass infringement of well-known copyrighted materials, in particular, appears to have caught Altman off guard.
“Another thing you learn once you launch one of these things is how people use them versus how you think they’re going to use them,” he told a16z. “And people are certainly using Sora in the ways we thought they were going to use it, but they’re also using it in ways that are very different.”
For now, Altman said that he wants to show the world what society looks like when tools to generate incriminating CCTV footage of crimes that were never committed — or even entire “South Park” episodes — land in the hands of anybody with a smartphone and an invite code.
In other words, Sora 2 is OpenAI’s grand AI slop experiment that may continue to take unexpected turns.
“I also think it is important to give society a taste of what’s coming on this co-evolution point,” he said on the podcast.
But “there will be some adjustment that society has to go through,” Altman added.
More on Sora: Sora 2 Has a Huge Financial Problem
The post Sam Altman Says Copyright Holders Are Begging for Their Characters to Be Included in Sora appeared first on Futurism.
Samsung HUD Glasses, Powered By Google, Could Launch Next Year
Samsung and Google plan to launch HUD glasses next year to take on Meta Ray-Ban Display, South Korea's The Financial News reports.
Samsung made the "reference design" glasses with a monocular heads-up display (HUD) that Google showed off at TED2025 and demoed at I/O 2025 earlier this year, but the companies haven't yet officially announced plans for a product.
UploadVRDavid Heaney
Google did announce earlier this year that it's working with the eyewear companies Gentle Monster and Warby Parker on Gemini smart glasses, and will work with Kering Eyewear in the future, but it also confirmed that the in-lens display will be "optional" for products, suggesting that these will be simpler Ray-Ban Meta competitors.
Now, The Financial News reports that Samsung and Google plan to launch smart glasses with a display next year, productizing the concept shown at TED and I/O to take on Meta Ray-Ban Display.
Samsung is also working on cheaper smart glasses without a display, the outlet says, seemingly aiming to compete with the regular Ray-Ban Meta glasses too.
UploadVRDavid Heaney
Samsung and Google's partnership on smart glasses products would be a continuation of their work on the first Android XR headset, rumored to launch later this month, which may be branded Galaxy XR.
For both the headset and glasses, the idea here is that Samsung handles the hardware while Google provides the software, playing to the strengths of each company, and bringing in the existing Android and Google devices and services ecosystem – strong cards to play against Meta.
ChatGPT fusionne moteur de recherche, App Store et OS, ce qui change tout pour les marques
Avec le lancement cette nuit des Apps ChatGPT, OpenAI transforme son assistant en plateforme d’applications intégrées. Les marques pourront désormais être invoquées directement dans le fil de la conversation, sans que l’utilisateur ne quitte l’interface. Une rupture qui redistribue les cartes du marketing digital. OpenAI franchit un cap majeur et créé une rupture historique, ChatGPT …
L’article ChatGPT fusionne moteur de recherche, App Store et OS, ce qui change tout pour les marques est apparu en premier sur FW.MEDIA.
F1 in Singapore: “Trophy for the hero of the race”
Few modern F1 venues are as dazzling as the Marina Bay circuit in Singapore. If you watch the early practice or the young women of F1 Academy in their races, you'll get an idea of the street circuit's relationship to the city in daylight as it takes in landmarks and crosses the water. At night, the brilliant white ribbon of racetrack throws the rest of the surroundings into darkness. Unlike a Le Mans car, there are no headlights in F1.
Just over 1,600 lights are needed for the job, which got underway in mid-June. They're LED now, a switch made in 2023 that cut energy consumption by a noticeable 30 percent. The light itself is even carefully tuned—a color of 5,700 K and a color rendering index of 90 best replicate daylight for the drivers. Oh, and there can't be any flickering that could affect the TV broadcast. The results look spectacular, especially in 4K.
As for the racing this year? Perhaps not so much. Past Singaporean Grands Prix have tended to be action-packed—or even chaotic. In the inaugural 2008 race, Red Bull came away convinced that electromagnetic interference from an underground train line caused the failure of one of its race cars, as well as a sister car belonging to the team we now call Racing Bulls. Other teams remained worried about this phenomenon as late as 2015, although no similar failures had been recorded. For the record, Singapore's Land Transit Authority says there was no train track below the relevant corner of the circuit—the nearest train tunnel is more than 600 feet away (200 m) and 32 feet (10 m) below ground level.
Actualité : Silence sur Jupiter : la sonde Juno, chef-d'œuvre de la NASA, laissée à l'abandon à cause du chaos politique
Un incendie et pas de backup - La Corée du Sud perd 858 To de données gouvernementales
Vous vous souvenez de cette règle de base en informatique, que je vous rabâche régulièrement, et qui dit de toujours avoir plusieurs sauvegardes de vos données critiques ?
Hé bien apparemment, le gouvernement sud-coréen a zappé ce cours, car le 26 septembre dernier, un incendie s’est produit au centre de données NIRS (National Information Resources Service) à Daejeon et a cramé 858 téraoctets de fichiers gouvernementaux . Et y’a pas de backup. Nada.
Le feu a démarré pendant une opération de maintenance sur une batterie lithium-ion dans laquelle une cellule a lâché , déclenchant ce qu’on appelle un emballement thermique… En gros, la batterie s’est transformée en bombe incendiaire et le brasier s’est propagé dans la salle serveur du cinquième étage, faisant tomber 647 services en ligne gouvernementaux d’un coup. Parmi eux, 96 systèmes critiques ont été directement détruits, et 551 autres ont été coupés préventivement pour éviter que la chaleur les bousille aussi.
Le système qui a morflé le plus, c’est G-Drive, le cloud de stockage utilisé par les fonctionnaires sud-coréens depuis 2018. Environ 750 000 employés du gouvernement central peuvent stocker leurs documents de travail dessus, mais seulement 125 000 l’utilisaient vraiment. Chacun disposait d’environ 30 Go d’espace de stockage, ce qui fait qu’au total le système contenait 858 To de données de travail accumulées sur huit ans.
Snif…
En fait, leur G-Drive est conçu comme un système de stockage haute capacité, mais basse performance, et ils ont des contraintes réglementaires qui imposent un stockage exclusif sur cette plateforme afin d’éviter les fuites de données.
Donc autrement dit, pour se prémunir contre les risques de fuite, ils ont créé un point de défaillance unique. Et comme y’avait pas de backup, c’est la cata… Bravo !
Du coup, quand l’incendie a détruit les serveurs physiques, tout est parti en fumée. Huit ans de documents de travail pour certains ministères, qui ont été complètement perdus… C’est surtout le ministère de la Gestion du Personnel qui s’est pris la claque la plus violente parce qu’il avait rendu obligatoire le stockage de tous les documents sur G-Drive uniquement. D’autres organismes comme le Bureau de Coordination des Politiques Gouvernementales, qui utilisaient moins la plateforme, ont moins souffert.
Bref, les autorités essaient maintenant de récupérer ce qu’elles peuvent depuis d’autres sources. Y’a des petits bouts de fichiers sauvegardés localement sur les ordinateurs personnels des fonctionnaires le mois précédent, les emails, les documents officiels validés et les archives papier… Bref, pour tous les documents officiels passés par des processus d’approbation formels, il y a un espoir de récupération via leur système OnNara (un autre système gouvernemental qui stocke les rapports finaux), mais pour tout le reste (brouillons, fichiers de travail en cours, notes internes…etc.) c’est mort de chez mort…
Le ministère de l’Intérieur a expliqué que la plupart des systèmes du centre de Daejeon sont normalement sauvegardés quotidiennement sur des équipements séparés dans le même centre ET dans une installation de backup distante. Mais G-Drive, lui, n’avait pas, comme je vous le disais, cette possibilité.
Évidemment, cet incident a déclenché une vague de critiques acerbes sur la gestion des données gouvernementales sud-coréennes. Un système de backup en miroir en temps réel qui duplique le serveur principal pour assurer la continuité de service en cas de panne, était complètement absent de l’infrastructure et pour un système aussi critique que le stockage de documents de 750 000 fonctionnaires, c’est difficilement compréhensible.
Voilà donc le gouvernement estime qu’il faudra jusqu’à un mois pour récupérer complètement les 96 systèmes de base directement endommagés par l’incendie et pour G-Drive et ses 858 To de données, par contre, c’est une autre histoire, car sans backup, les données sont définitivement perdues.
De plus, cet incendie déclenche actuellement un genre d’examen mondial des batteries lithium-ion dans les datacenters et des architectures de plan de reprise d’activité (PRA). Les batteries lithium-ion sont en effet utilisées partout pour l’alimentation de secours, mais leur risque d’emballement thermique en cas de défaillance pose de sérieuses questions sur leur place dans des infrastructures critiques…
Bref, je souhaite bon courage aux Coréens, et j’espère que tout le monde saura tirer des enseignements de ce malheureux incendie…
OpenAI launches AgentKit to help developers build and ship AI agents
L'UE promeut l'éducation financière
The GEO companies winning the AI search arms race
AI shopping is here, and brands are entering a new battleground.
OpenAI‘s September 2025 rollout of in-platform checkout via Shopify and Stripe enables shoppers to search, decide, and buy entirely within ChatGPT. Adobe reports that traffic from generative AI platforms to US e-commerce sites surged 4,700% year-over-year in July 2025, and is accelerating every month.
To compete, brands must establish a strong presence in AI search — or risk losing share in the emerging agentic shopping journey. Generative engine optimization (GEO) is the solution: startups in this space are building tools to track, measure, and optimize brand visibility across ChatGPT, Claude, Google AI Overviews, and other AI-generated answers, creating a new SEO arms race.
Using CB Insights’ Mosaic score — which measures private company health and predicts likelihood of success — we analyzed more than 20 GEO companies, ranking them by 1-year Mosaic score growth to identify the fastest-rising vendors. By examining their capabilities and growth signals, this analysis highlights the GEO partners best positioned to help brands win in AI search.
Explore the generative engine optimization (GEO) market
- Native GEO solutions have the first-mover advantage. Of the 8 GEO companies with the most momentum — each posting more than the market’s average 14% Mosaic growth in the past year — 7 were founded between 2023-2025. All but 2 companies with Mosaic scores of 600+ also launched in the same window. These “native” GEO companies were built specifically for LLM visibility rather than retrofitting or bolting onto SEO tools, giving them advantages in data collection, multi-model monitoring, and content recommendations.
- Four capabilities have emerged as table stakes, and vendors without this full stack will fall behind. Analysis of the CB Insights market scorecard for companies with the top Mosaic scores (600+) highlights a consistent feature set: multi-platform monitoring across ChatGPT, Perplexity, Google AI Overviews, and emerging engines; competitive benchmarking that tracks share of voice against rivals; sentiment analysis measuring how brands are portrayed; and actionable insights on content gaps and adjustments. Vendors that deliver this complete capability set are the ones that will lead the market. Buyers should prioritize these features when evaluating potential partners.
- Automated content generation is becoming a differentiator. While most GEO tools diagnose visibility, leaders are building autonomous content engines that create LLM-optimized content. Profound offers AI content workflows using preset templates to write AI-friendly material from aggregated citations. Surfer‘s Content Editor suggests LLM-ready content modifications. Scrunch constructs parallel “AI-friendly” website versions optimized for machine consumption alongside human-facing sites. As GEO evolves, the platforms that can both track and directly shape how LLMs read the web will dictate the rules of agentic discovery. Brands’ demands for the most efficient solutions will force pure-play monitoring tools to build or acquire content capabilities.
For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.
The post The GEO companies winning the AI search arms race appeared first on CB Insights Research.
ChinAI #331: Chinese Public Perceptions and Usage of AI (2025 survey)
Greetings from a world where…
The Farseer Trilogy is epic
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: Chinese Public Perceptions and Usage of Generative AI
Context: Last month, to explore the public’s usage habits and attitudes toward generative AI, Tencent Research Institute conducted an online survey of 3,570 Chinese adults (link to original Chinese). This is not a nationally representative survey; rather, it captures the views of highly educated young adults. To make this clear, 85% of the survey respondents had a college degree or higher, whereas just 10% of China’s labor force (age 25-64) has at least a bachelor’s degree. They did collect a gender-balanced sample from a variety of industries, so the below findings probably reflect the views of young, highly educated digital natives in China.
Key Takeaways: Adoption patterns — wide, localized, but “wait-and-see” attitude regarding willingness to pay. Let’s go through these three points in turn.
96% of respondents had used AI-generated content tools, with two-thirds (67.7%) of those being daily users.
-
Chinese companies accounted for the top three most-used products: ByteDance’s Doubao, DeepSeek, and Tencent Yuanbao. Tencent Research Institute attributes this trend to network access restrictions (e.g., you have to find mirror sites or use a VPN to access Claude or ChatGPT), superior Chinese-language capabilities of local services, and synergies with the ecosystem of Chinese applications (e.g., WeChat’s integration of DeepSeek)
I was disappointed that the survey results don’t show the actual bar graphs for this question; instead, they just provide a word cloud, which omits the exact figures for the most popular generative AI apps.
Two more interesting points about age groups and GenAI product usage. “Among respondents aged 20-29, ChatGPT’s usage rate was significantly higher than among respondents in other age groups,” the article states. And, Moonshot AI’s Kimi places in the top three among respondents under 20.
In the section on commercialization prospects, the Tencent Research Institute survey found that about 60% of respondents stated that they were “watching from the sidelines [正在观望], but would pay if it’s good” (see image below). Among the users that were open to paying for generative AI services, the most preferred payment option was “monthly payment” and under 100 RMB/month. The study authors theorizes that consumers see the generative AI product market like video streaming services (for context, the monthly fee for iQiyi [China’s Netflix] is around 25 RMB/month).
The Chinese public has mixed views about AI’s societal impact.
About 46% of respondents feel “both excited and concerned” about the future of AI. The three main risks: 1) 60% of respondents expressed concern about the “proliferation of false information and fake news”1; 2) 60% also fear job displacement; and 3) 47% cited the risk of “privacy breaches.”
One item that caught my attention relates to the so-called “enthusiasm gap” between the Chinese public and Western publics, which is worth unpacking further. Overall, the majority of respondents (72%) believe that generative AI will have a “primarily positive impact.” This seems to support this overall assumption that the Chinese public is more optimistic about AI than people in other countries.
But, what happens when we dig a little deeper into this purported optimism gap? Again, I would be very skeptical about taking online surveys from China as representative of the entire population. For instance, one survey cited as evidence for this gap finds that 81% of the Chinese public agrees that the benefits of AI outweigh the risks (the figure was 41% for the US). However, turn to page 75 of the methodology section, and you’ll see this revealing statement:
“Samples from emerging economies (Brazil, India, China, and South Africa) represented considerably more university educated people than their respective general populations (using OECD 2021 education data as a comparison). A higher representation of educated people is common in survey research from the BICS countries. For instance, Edelman (2022) and Ipsos (2022) both note that online samples in Brazil, India, China and South Africa are more educated, affluent, and urban than the general population.”
To me, here’s the bottom line. Regarding findings from online surveys like this week’s feature translation and the above one, they reflect the views of highly educated adults in China, very likely concentrated in the coastal provinces — not the general Chinese public. Thus, I would be very wary of using these survey findings to support the claim that AI will diffuse more quickly throughout the entire Chinese economy.
FULL TRANSLATION: Chinese Public Perceptions and Usage of Generative AI
ChinAI Links (Four to Forward)
Should-read: Compute is not the answer to AI sovereignty
A GovAI summer fellow, Hamish Low, has published a thoughtful piece about how the UK should approach its compute buildout and AI sovereignty. He concludes, “Given a sufficiently focused and aggressive industrial policy we can build powerful positions in key future nodes of the AI value chain and ensure that Britain stays in the room when that crisis comes. (Mutual) Interdependence (with the U.S.) can be much more powerful for the UK than compute infrastructure.”
Should-read: The U.S. Needs A Generative AI Intensity Index
This is a very cool project from SeedAI that tackles the challenges in measuring the diffusion of generative AI. Joshua New, Marina Meyjes, and Austin Carson propose an index to “measure the digital processes created during real-world use of generative AI systems, offering a direct proxy for their adoption across the economy.”
Should-read: Grant Delays Threaten Cultural and Language Studies Programs
On September 30, the Department of Education ended seven decades of (bipartisan) support for initiatives such as the National Resource Center (NRCs) program, dedicated to language and area studies. GW has two NRCs, including one that supports students to pursue intensive language study in East Asia. Again, what are we doing here? Reporting for Inside Higher Ed, Johanna Alonso provides the broader context.
Must-read more closely later but putting it here now: DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning
DeepSeek researchers published an open access Nature paper that details their process for training the R1 model. H/t to Miles Brundage for flagging that it includes a 10-page safety report that talks about their efforts to control risks. More on this in future issues.
Thank you for reading and engaging.
*These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Also! Listen to narrations of the ChinAI Newsletter in podcast format here.
Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99
Tencent Research Institute recently published a report on AI-generated misinformation, which ChinAI translated: https://chinai.substack.com/p/chinai-322-100-cases-of-ai-disinformation?utm_source=publication-search
OpenAI offers more copyright control for Sora 2 videos
Matignon publie une charte pour une utilisation raisonnée de l’intelligence artificielle générative
L'intelligence artificielle générative est la bienvenue dans les services du Premier ministre... mais sous conditions. Dans une charte récemment publiée par l'hôtel Matignon, les recommandations à l'usage des agents distinguent les usages autorisés et les usages proscrits. Parmi les premiers figurent "les cas d'usages auxquels l'IA peut répondre" : résumer, synthétiser, rédiger et corriger des documents non confidentiels… La recherche documentaire peut également être confiée à une IA mais "il est nécessaire d’effectuer une vérification systématique de la validité de ces informations."
Autres cas d'usages autorisés, la traduction de documents qui ne présentent pas de caractère confidentiel et la génération d'éléments sonores et visuels pour la création de maquettes par exemple.
[...] Lire la suite de cet article sur Archimag.comOpenAI and chipmaker AMD sign chip supply partnership for AI infrastructure
Airbags, and How Mercedes-Benz Hacked Your Hearing
Airbags are an incredibly important piece of automotive safety gear. They’re also terrifying—given that they’re effectively small pyrotechnic devices that are aimed directly at your face and chest. Myths have pervaded that they “kill more people than they save,” in part due a hilarious episode of The Simpsons. Despite this, they’re credited with saving tens of thousands of lives over the years by cushioning fleshy human bodies from heavy impacts and harsh decelerations.
While an airbag is generally there to help you, it can also hurt you in regular operation. The immense sound pressure generated when an airbag fires is not exactly friendly to your ears. However, engineers at Mercedes-Benz have found a neat workaround to protect your hearing from the explosive report of these safety devices. It’s a nifty hack that takes advantage of an existing feature of the human body. Let’s explore how air bags work, why they’re so darn loud, and how that can be mitigated in the event of a crash.
A Lot Of Hot Air
Once an obscure feature only found in luxury vehicles, airbags became common safety equipment in many cars and trucks by the mid-1990s. Indeed, a particular turning point was when they became mandatory in vehicles sold in the US market from late 1998 onwards, which made them near-universal equipment in many other markets worldwide. Despite their relatively recent mainstream acceptance, the concept of the airbag actually dates back a lot farther.
The basic invention of the airbag is typically credited to two English dentists—Harold Round and Arthur Parrott—who submitted a patent for the concept all the way back in 1919. The patent regarded the concept of creating an air cushion to protect occupants in aircraft during serious impacts. Specific attention was given to the fact that the air cushion should “yield readily without developing the power to rebound,” which could cause further injury. This was achieved by giving the device air outlet passages that would vent as a person impacted the device, which would allow the cushion to absorb the hit gently while reducing the chance of injury.
The concept only later became applicable to automobiles when Walter Linderer filed for a German patent in 1951, and John W. Hetrick filed for a US patent in 1952. Both engineers devised airbags that were based on the release of compressed air, triggered either by human intervention or automated mechanical means. These concepts proved ultimately infeasible, as compressed air could not be feasibly be released to inflate an airbag quickly enough to be protective in an automobile crash.
It would only be later in the 1960s that workable versions using explosive or pyrotechnic inflation came to the fore. The concept was simple—use a chemical reaction to generate a great deal of gas near-instantaneously, inflating the airbag fractions of a second before vehicle occupants come into contact with the device. The airbags are fitted with vents that only allow the gas to escape slowly. This means that as a person hits the airbag, they are gently decelerated as their impact pushes the gas out of the restrictive vents. This helps reduce injuries that would typically be incurred if the occupants instead hit interior parts of the car without any protection at all.
The Big Bang
The use of pyrotechnic gas generators to inflate airbags was the leap forward that made airbags practical and effective for use in automobiles. However, as you might imagine, releasing a massive burst of gas in under 50 milliseconds does create a rather large pressure wave—which we experience as an incredibly loud sound. If you ever seen airbags detonated outside of a vehicle, you’ve probably noticed they sound rather akin to fireworks or a gun going off. Indeed, the sound of an airbag can exceed 160 decibels (dB)—more than enough to cause instant damage to the ear. Noise generated in a vehicle impact is often incredibly loud, too, or course. Ultimately, this isn’t great for the occupants of the vehicle, particularly their hearing. Ultimately, an airbag deployment is a carefully considered trade-off—the general consensus is that impact protection in a serious crash is preferable, even if your ears are worse for wear afterwards.
However, there is a technique that can mitigate this problem. In particular, Mercedes-Benz developed a system to protect the hearing of vehicle occupants in the event that the airbags are fired. The trick is in using the body’s own reactions to sound to reduce damage to the ear from excessive sound pressure levels.
The stapedius reflex (also known as the acoustic reflex) is one of the body’s involuntary, instantaneous movements in response to an external stimulus—in this case, certain sound levels. When a given sound stimulus occurs to either ear, muscles inside both ears contract, most specifically the stapedius muscle in humans. When the muscle contracts, it has a stiffening effect on the ossicular chain—the three tiny bones that connect the ear drum to the cochlea in the inner ear. Under this condition, less vibrational energy is transferred, reducing damage to the cochlea from excessive sound levels.
The threshold at which the reflex is triggered is usually 10 to 20 dB lower than the point at which the individual feels discomfort; typical levels are from around 70 to 100 dB. When triggered by particularly loud sounds of 20 dB above the trigger threshold, the muscle contraction is enough to reduce the sound level at the cochlea by a full 15 dB. Notably, the reflex is also triggered by vocalization—reducing transmission through to the inner ear when one begins to speak.
Mercedes-Benz engineers realized that the stapedius reflex could be pre-emptively triggered ahead of firing the airbags, in order to provide a protective effect for the ears. To this end, the company developed the PRE-SAFE Sound system. When the vehicle’s airbag control unit detects a collision, it triggers the vehicle’s sound system to play a short-duration pink noise signal at a level of 80 dB. This is intended to be loud enough to trigger the stapedius reflex without in itself doing damage to the ears. Typically, it takes higher sound levels closer to 100 dB to reliably trigger the reflex in a wide range of people, but Mercedes-Benz engineers realized that the wide-spread frequency content of pink noise enable the reflex to be switched on at a much lower, and safer, sound level. With the reflex turned on, when the airbags do fire a fraction of a second later, less energy from the intense pressure spike will be transferred to the inner ear, protecting the delicate structures that provide the sense of hearing.
Mercedes-Benz first released the technology in production models almost a decade ago.
The stapedius reflex does have some limitations. It can be triggered with a latency of just 10 milliseconds, however, it can take up to 100 milliseconds for the muscle in the ear to reach full tension, conferring the full protective effect. This limits the ability of the reflex to protect against short, intense noises. However, given the Mercedes-Benz system triggers the sound before airbag inflation where possible, this helps the muscles engage prior to the peak sound level being reached. The protective effect of the stapedius reflex also only lasts for a few seconds, with the muscle contraction unable to be maintained beyond this point. However, in a vehicle impact scenario, the airbags typically all fire very quickly, usually well within a second, negating this issue.
Mercedes-Benz was working on the technology from at least the early 2010s, having run human trials to trigger the stapedius reflex with pink noise in 2011. It deployed the technology on its production vehicles almost a decade ago, first offering PRE-SAFE Sound on E-Class models for the 2017 model year. Despite the simple nature of the technology, few to no other automakers have publicly reported implementing the technique.
Car crashes are, thankfully, rather rare. Few of us are actually in an automobile accident in any given year, even less in ones serious enough to cause an airbag deployment. However, if you are unlucky enough to be in a severe collision, and you’re riding in a modern Mercedes-Benz, your ears will likely thank you for the added protection, just as your body will be grateful for the cushioning of the airbags themselves.
Le prochain film Star Wars prolonge l’effacement des plus grands récits du Mandalorien : quel impact réel ?
Alors que le prochain film Star Wars s’apprête à sortir, la franchise poursuit une tendance amorcée depuis quatre ans en revenant sur certains des arcs narratifs les plus marquants de The Mandalorian, soulevant des interrogations sur l’impact réel de ces choix scénaristiques.
✍️ Edito - Critterz : le film créé par IA qui fait flipper les studios d'animation
Alors qu’OpenAI annonce un film d’animation pour 2026, co-réalisé avec l’intelligence artificielle, la profession s’alarme : et si Critterz annonçait l’ère d’un cinéma sans humains ?
En 2023, un court-métrage mystérieux apparaît sur YouTube : Critterz, des créatures poilues et mignonnes dans une forêt stylisée façon Pixar. Mais ce n’est pas Pixar : c’est OpenAI. Et ce n’est qu’un début.
Un long-métrage est annoncé pour 2026, avec l’ambition de concourir dans les festivals. Réalisé à partir d’images générées par DALL·E, puis animées de manière classique, Critterz n’est pas un film "100% IA". L’écriture a été assistée par ChatGPT, mais ce sont encore des humains qui tiennent les manettes.
Pourtant, la polémique enfle. Des professionnels dénoncent une menace pour la création et les emplois. Faut-il avoir peur ? Ce n’est pas la première fois que la technologie bouleverse l’industrie du cinéma — on l’a vu avec le montage numérique, les effets spéciaux ou encore l’arrivée du son.
La véritable rupture ? Critterz coûtera cinq à dix fois moins cher qu’un film d’animation traditionnel… et pourra être produit en moins d’un an. C’est peut-être là que le bât blesse.
Dans cet édito, je reviens sur ce cas emblématique qui cristallise toutes les inquiétudes autour de l’intelligence artificielle dans la création artistique.
-----------
♥️ Soutien :
https://mondenumerique.info/don
🗞️ Newsletter :
https://mondenumerique.substack.com
Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.
📆 L’HEBDO 27/09 - Apple défie l’Europe à propos du DMA
Cette semaine : Apple vs DMA, cyberattaque SIM, salon All In à Montréal, Mistral au Canada, IA qui ment, fin des cookies, bulle IA et visas H-1B, agents intelligents, finances avec Qonto.
💡 Découvrez Frogans, l’innovation française qui réinvente le Web
Donnez votre avis sur ce podcast en répondant à notre 👉 sondage
🍏 Apple vs. Bruxelles : le bras de fer du DMA
Le Digital Markets Act met le feu aux poudres. Apple accuse la loi européenne de freiner l’innovation et menace à demi-mot la commercialisation de certains produits en Europe. L’Union, elle, reste ferme. Une bataille politique et technologique est lancée.
📱 Cyberattaque déjouée à New York
Les services secrets américains ont démantelé une infrastructure composée de 100 000 cartes SIM pirates près du siège de l’ONU. Objectif : saturer les réseaux télécoms ou lancer des menaces anonymes. Une affaire aux allures de thriller.
🧠 IA qui ment : le “skimming” inquiète
Une étude OpenAI révèle que les modèles peuvent dissimuler leurs erreurs. Si le risque reste faible, la difficulté à détecter ces comportements soulève des enjeux cruciaux pour la sécurité future des IA.
🤖 Les agents IA, prochaine frontière
OpenAI dévoile Pulse, un assistant proactif qui collecte pour vous informations et résumés.
Les agents intelligents s’imposent comme un tournant majeur pour l’entreprise. Explication de Nicolas Gaudillière (Capgemini) [PARTENARIAT].
🇨🇦 Montréal, capitale de l’IA
Bruno Guglielminetti nous emmène au salon All In. Ambiance effervescente, annonces stratégiques (notamment l’implantation de Mistral AI au Canada) et débats sur la souveraineté numérique.
💸 La bulle de l’IA va-t-elle exploser ?
Luc Julia (Silicon Valley) analyse l’investissement record de Nvidia dans OpenAI et alerte : les valorisations surréalistes rappellent la bulle Internet. Il revient aussi sur la menace des restrictions de visas H-1B voulues par Donald Trump.
💳 Qonto : l’IA au service des finances d’entreprise
Avec Aymeric Augustin, directeur technique de Qonto, on découvre comment la fintech réinvente la gestion financière avec l’automatisation, la simplification des notes de frais et des outils IA intégrés [PARTENARIAT]
-----------
♥️ Soutien :
https://mondenumerique.info/don
🗞️ Newsletter :
https://mondenumerique.substack.com
Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.
🎤 L'humeur de Luc Julia - La bulle de l’IA est-elle inévitable ?
L’intelligence artificielle connaît-elle une bulle financière sur le point d'éclater, à l’image de celle d’Internet en 2000 ? Pour Luc Julia, figure majeure de l’IA installé dans la Silicon Valley depuis vingt ans, la réponse est claire : oui, car l’excitation autour des IA génératives a atteint des sommets irrationnels.
Contenu de l'interview :
- L’excitation autour des IA génératives rappelle la bulle Internet des années 2000.
- Les dangers de survalorisations insoutenables pour certaines startups.
- L’émergence d’une IA plus frugale et spécialisée : les « IA agentiques ».
Et aussi :
- Nouvelle taxe à 100 000 dollars sur les visas H-1B des travailleurs étrangers aux Etats-Unis : quelles conséquences pour les entreprises de la tech ?
-----------
♥️ Soutien :
https://mondenumerique.info/don
🗞️ Newsletter :
https://mondenumerique.substack.com
Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.
🎤 Interview - Multi-agents : quand l’IA devient un collaborateur à part entière (Nicolas Gaudilliere, Capgemini Invent)
Les agents intelligents, commencent à transformer les processus métiers dans les entreprises. Nicolas Gaudilliere, directeur Technologie et Innovation chez Capgemini Invent, explique comment les architectures multi-agents ouvrent la voie à une nouvelle ère de collaboration homme-machine.
Automatisation, adaptabilité, interaction avec des systèmes externes : leur potentiel est immense, mais les défis sont tout aussi importants. De la banque aux télécoms, en passant par le support client, il détaille les cas d’usage les plus prometteurs, les gains déjà observés et les conditions pour réussir un passage à l’échelle.
Nous évoquons également les enjeux cruciaux de sécurité et de gouvernance, sans lesquels cette révolution ne pourra s’ancrer durablement dans les organisations.
Au programme :
- Qu’est-ce qu’un agent intelligent et en quoi diffère-t-il d’un simple chatbot ?
- Automatisation des processus métiers : exemples concrets dans la banque (KYC), les télécoms et le support client.
- Les architectures multi-agents : spécialisation, orchestration et collaboration avec l’humain.
- Les promesses de productivité (jusqu’à 30% de gain) et leurs limites actuelles.
- Les défis à relever : sécurité, intégration native, passage à l’échelle.
- L’adoption progressive par toutes les entreprises, des grands groupes aux PME.
-----------
♥️ Soutien :
https://mondenumerique.info/don
🗞️ Newsletter :
https://mondenumerique.substack.com
Hébergé par Audiomeans. Visitez audiomeans.fr/politique-de-confidentialite pour plus d'informations.
Divining Air Quality With A Cheap Computer Vision Device
There are all kinds of air quality sensors on the market that rely on all kinds of electro-physical effects to detect gases or contaminants and report them back as a value. [lucascreator] has instead been investigating a method of determining air quality that is closer to divination than measurement—using computer vision and a trained AI model.
The system relies on an Unihiker K10—a microcontroller module based around the ESP32-S3 at heart. The chip is running a lightweight convolutional neural network (CNN) trained on 12,000 images of the sky. These images were sourced from a public dataset; they were taken in India and Nepal, and tagged with the relevant Air Quality Index at the time of capture. [lucascreator] used this data to train their model to look at an image taken with a camera attached to the ESP32 and estimate the air quality index based on what it has seen in that existing dataset.
It might sound like a spurious concept, but it does have some value. [lucascreator] cites studies where video data was used for low-cost air quality estimation—not as a replacement for proper measurement, but as an additional data point that could be sourced from existing surveillance infrastructure. Performance of such models has, in some cases, been remarkably accurate.
[lucascreator] is pragmatic about the limitations of their implementation of this concept, noting that their very compact model didn’t always perform the best in terms of determining actual air quality. The concept may have some value, but implementing it on an ESP32 isn’t so easy if you’re looking for supreme accuracy. We’ve featured some other great air quality projects before, though, if you’re looking for other ways to capture this information. Video after the break.