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
▶
Le responsable IA d'imec détaille des milliers de tests d'agents de code sur SWE-Bench Pro : quand faire tourner ses propres GPU devient plus rentable que payer une API.
A study examines Canada's strategy as an infrastructure: who holds the expertise, how AI's opacity feeds strategic ignorance, and why recourse is struggling to redress harm.
Tested on the 18th release of data from the Sloan Digital Sky Survey, this unsupervised model competes with conventional methods: 0.46% outliers on good quality spectra.
Of 427 stories written by 333 participants, 100% AI channels are more creative but more uniform. Conferring planning to a human keeps creativity and restores diversity.
A survey of 891 American and Turkish students shows that AI's experience only develops critical thinking if tasks are distributed with the tool and their productions are reprocessed.
Researchers have pre-entered Qwen3 zero models with trapped data: once they become command line agents, they start command of the attacker in 92% of cases.
Driven by 1,616 demonstrations in calm and agitated water, vision-language-action models reach 93.3% success in the basin. On a wreck, an ACT policy succeeds 66.7% of a task.
According to The Information, the Max prototype weighs about 48 kg and must physically embody Meta's Agent Muse for domestic tasks. Meta says she doesn't plan on marketing it.
When asked about local inference, OpenAI's boss estimates that in the midst of a shortage of DRAM, memory is better used in the cloud than in active computers 5% of the time.
According to the Wall Street Journal, large investors reduced their commitments in OpenAI's latest release, against the background of declining technological values. The recovery falls to 1 250 billion.
▶
La responsable produit de Bumble tire cinq leçons d'un test A/B sur une fonction IA de conseils photo : forte demande déclarée, faible usage réel, coûts et régulation.
A survey of 234 students shows that they provide context and respond to their requests, but use almost the same strategies for creative exercise and research.
Applied to RadioAstron and ALMA archives, the method is scaled to the number of pixels and provides more statistical information than the conventional CLEAN algorithm.
The robot Alice said the right words in the wrong order, or announced a visitor already present. A feedback offers TRACE, a review to identify these flaws.
A survey of clinicians in two departments shows that prediction is mainly important in complex cases, that it is discussed among colleagues, and that use is changing with experience.
Turning the checks around, acting without asking: about 88,000 officer skills contain this type of instructions, often copied from a handful of original models.
Prices do not change (100 and 200 dollars per month), but the most expensive subscription loses its five-hour slippery window. Weekly limits remain in place.
Note AA- is maintained, but the Agency is concerned about a negative cash flow and investments of $135 to $145 billion this year. Meta disputes.
Amazon's cloud reached $45.98 billion in revenue, its strongest growth since the end of 2021. Andy Jassy targets more than 40% per year in 2027 and 2028.
In Tucson, the Project Blue project brought together opponents from all sides. The city blocked the annexation of the land and the county frozen the new projects for 120 days, but the site continues.