Why AI Agents Still Won't Pay for Anything — Jan Curn, Apify
x402, AP2, ACP… les standards de paiement pour agents se multiplient, mais le volume réel reste quasi nul. Le fondateur d'Apify décortique où le parcours d'achat d'un agent se casse.
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
x402, AP2, ACP… les standards de paiement pour agents se multiplient, mais le volume réel reste quasi nul. Le fondateur d'Apify décortique où le parcours d'achat d'un agent se casse.
Two researchers argue in favour of ‘injecting’ one’s knowledge into the prompt rather than treating the model as an authority, and propose five principles for responsible prompting.
Rather than fact-checking every piece of fake news, researchers are teaching strategies that recur across different pieces of fake news. When used as language models, humans are better at distinguishing truth from falsehood on unfamiliar topics.
Trained on 130 teleoperated demonstrations, an ACT policy runs at 15 Hz on a Jetson AGX Orin to detach the milling cutters without damaging them. The data, weights and code are published as open source.
Trained on haemodynamic simulations, a model identifies the narrowed artery and the severity of the stenosis by comparing the predicted and observed velocity curves in eight vessels.
Tested in a simulation using Swift’s Burst Alert Telescope, an AI system directly deduces the positions of hard X-ray sources from detector counts, without knowing the pattern of the code-sliced image.
Researchers have used large language models (LLMs) to map the literature on water and energy use in AI infrastructure, from mining to electronic waste. The conclusion: we mainly measure the electricity used by data centres, but the rest is poorly measured.
Researchers have shown that an attacker can automatically hide a malicious payload within a code repository in a location where an AI agent is least likely to detect it. Even commercial agents fall into the trap.
The Californian insurer simulates each house, tree by tree, to assess the fire risk. Its ‘Stand World Model’ is said to have predicted the fate of 310 houses affected by fires with 85 per cent accuracy.
The former Crescent Island has become the Intel GPU D-series: 1.3 petaflops in FP8, 1.5 TB/s of bandwidth and a massive amount of memory capable of handling models with 300 billion parameters on a single card.
Gemini 4 Argon, réservé il y a peu à quelques défenseurs en cybersécurité, s'ouvre aux développeurs, aux entreprises et aux abonnés Pro/Ultra : contexte d'un million de tokens, benchmarks contestés et prix.
Out of 2,000 candidate excerpts, the decoder ranks the correct one in its top 10 for more than 42 per cent of queries, by combining representations of text, audio and speech.
Parabolic jumps between waypoints, image segmentation to detect obstacles, and automatic retreat to a safe location: the framework has been tested in a high-fidelity lunar simulation.
Questions asked in Mandarin, answers in both languages and a technical glossary generated on the fly: a multi-source RAG pipeline tested on 91 Piazza messages.
The RT-DETR detector achieves 98.25 per cent accuracy per image on Gram-stained smears photographed using the Galaxy S23 Ultra. The authors call for further multicentre validation.
Launched by the Korean Publishers Society, the appeal is directed at the government, the National Assembly and AI research laboratories. The backdrop to this is the purchase of Korean books by Anthropic.
Palisades, Three Mile Island and Duane Arnold are set to supply 2.25 GW to the grid, with Microsoft and Google as customers. Critics have raised concerns about the safety of reactors designed in the 1970s.
Bypassing access controls, unauthorised payments, 147,451 lines of private data copied without notification: OpenAI has brought the number of published incidents to 19.
OpenAI will “definitely” build a humanoid, according to its CEO, who prioritises the “brain” over the body. There are 27 vacancies in robotics, including roles involving in-house actuators.
An LLC owned by Anthropic will receive 5 per cent of the shares, plus a further 2 per cent for every 50 per cent increase in the company’s valuation. The first area of focus will be the impact of AI on employment.