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
When it comes to identifying and dating intracranial haemorrhages in case reports, the distillation of actual reports works well; summarised reports do not even outperform the untrained model.
A review of the literature examines why we attribute human traits to AI, what the costs of this are, and how the way we talk about it shapes the role we allow it to play.
Out of 2,000 cases across the cantons between 2020 and 2023, the recommendation optimised using machine learning increased the three-year employment rate by 5.2 percentage points.
The Argus tool analyses the behaviour and style of code across six years of graduate intake; its alerts show a negative correlation with marks in invigilated examinations, but require human review.
Of the 95 adverts funded by Leading the Future and Public First Action, only seven mentioned AI and none mentioned data centres; the bulk of the funding went to seats that had already been secured.
Released under the MIT licence, the library implements the DeepGEMM API for Ascend 950 NPUs in BF16, FP8 and FP4 — another step towards phasing out CUDA in China.
Cohere’s two new embedding models share the same vector space, process 128,000 tokens, support over 100 languages and images, and cost $0.12 and $0.08 per million tokens respectively.
Its Financial Stability Committee considers it more likely than before that the system’s vulnerabilities will materialise, with AI-related debt having doubled in a year to around $450 billion.
Tasked with responding to buyers of a keyboard, the Meta agent accepted a discounted offer, provided the seller’s address and arranged a meeting, without the seller’s knowledge.
▶
Deux minutes de démonstration du premier modèle d'action de Runway, qui transforme son savoir-faire en génération vidéo en politique de contrôle pour robots.
Chinese researchers have presented an agent framework that coordinates a humanoid robot with the surgical team over long sequences, incorporating three overlapping layers of safety.
Using CASPER, researchers have shown that the error in a downscaling model depends on the climatic discrepancy with its training data far more than on the volume of that data.
Six equity-based adjustments were tested using data from a public secondary school in Ontario. The indicators show a consistent trend, but small marginalised groups remain underserved.
OverdoseMoE combines several models adapted to ICD codes to estimate the risk of overdose over a six-month period based on one year’s worth of medical records.
Rather than comparing the data with theoretical models, this pipeline detects anomalies. It is all the more effective when the colliding black holes are massive.
Topos Bio has released two models that generate the full range of conformations of a protein, including its regions without a fixed structure, and significantly outperforms BioEmu on PeptoneBench.
The Palo Alto-based start-up, backed for the first time by Nvidia and Intel, aims to prevent a faulty laser diode from bringing thousands of accelerators to a standstill.
Launched in beta in July, Anthropic’s offering for federal and state government agencies is now available in a FedRAMP High-certified environment.
Anthropic has evaluated Z.ai’s open-weight model: 50 successful exploits out of 410 attempts, and defences that are reduced from 95 per cent to 6 per cent simply by removing the rejections.
▶
Tour d'horizon en treize minutes du nouveau modèle de Google : usage interne, benchmarks, et pourquoi seule une poignée de défenseurs en cybersécurité peuvent s'en servir pour l'instant.