AM-IA brings together news, research papers and applications relating to artificial intelligence published around the world into a single feed — and makes them available in your European language at the click of a button.
Scroll to browse · Click to filter
▶
Branislav Brzak montre qu'un canal auxiliaire temporel sur un GPU H100 partagé permet de deviner l'activité d'un autre locataire : profondeur du modèle, vitesse de décodage, type de tâche.
Maia 300, Microsoft’s third-generation AI chip, promises 40 per cent more tokens per megawatt than GPU-based systems and will be deployed on a gigawatt scale from 2027.
Founded this year by Hu Han, former head of robotic modelling at Tencent, Winverse is reportedly raising tens of millions of dollars from DeepSeek and Long-Z (Meituan).
The Microsoft chairman is calling for ‘emergency brakes’, mandatory incident reporting and an independent assessment of models, and sees the UK taking the lead on these rules.
In Singapore, Hock Tan and Lip-Bu Tan believe that funding could hold back AI more than chips or energy, despite demand outstripping supply until 2029.
The programme focuses on local projections, hybrid models and their evaluation. The presentation, updated in October, emphasises data and partnerships.
▶
Le directeur scientifique de Modular explique pourquoi la multiplication des puces IA a désorganisé la pile logicielle, et comment Mojo et MAX veulent faire tourner l'inférence sur tous les matériels.
Changing the way in which the distance between brain matrices is measured has little effect on a classifier’s accuracy, but may alter its individual decisions and the explanations it provides for them.
Traditional models of chemistry struggle to account for enzyme-catalysed reactions; VenusRX learns about reactants, products and catalytic function all at once.
In a formal model, ensuring both a maximum level of risk and a development deadline requires ruling out the assumption of inefficient safety research.
By generating counterfactual variants of recorded movements and assessing the physical plausibility of the predicted videos, DreamTrue reduces the interaction failure rate from 48 per cent to 6 per cent.
AI emulators are already rivalling physical models in weather forecasting; when it comes to climate, they will need to explicitly represent forcings and withstand conditions that have never been observed.
Rather than an orchestrator that delegates tasks to specialised models, Reka presents a single network that comprehends, generates and acts, with a distilled version capable of rendering 5 seconds of video in approximately one second.
A side panel allows Claude to view the open file and edit it on the spot, with each change saved by default; connectors also create Google files directly from the chat.
According to the Wall Street Journal, Amazon is in talks to acquire the Israeli start-up Decart, which specialises in world models and video, for nearly 7 billion dollars.
A Ukrainian strike on Sunday brought Yandex’s 40–50 MW facility in Vladimir to a standstill, disrupting more than 80 computing, storage and AI services, including the YandexGPT API.
▶
Atelier pratique : comment encadrer un agent de code par des étapes de contrôle successives — plan, implémentation, vérifications locales et revue de pull request — sur une petite application de paiement.
A filing with the European Commission reveals that Apple is recruiting staff from Huxe and has secured a non-exclusive licence to its intellectual property, without acquiring the start-up.
The assistants who write up medical notes do not pick up on body language or emotions, and prioritise clinical information over the patient’s account. They could also undermine doctors’ memory and reasoning.
From the very first bulletin, the National Hurricane Centre based its forecasts on WeatherNext 3 rather than on physical models, and announced a Category 2 storm when the system did not even have a name.