Topic
Explore the latest artificial intelligence news and research.
A review of how Iris-mini and Iris-pro were trained and evaluated. On BrowseComp, discarding history added about 2.2 times the gain from moving to the larger model, and key training steps remain unreleased.
In the GENESIS-IPF trial for the lung fibrosis drug rentosertib, all six aging clocks built by five different research groups pointed toward a lower biological age. But the clocks differ in what they're designed to predict—chronological age or mortality risk—and in their underlying methods, so they can't simply be reduced to a single number.
Sam Altman said OpenAI will accept independent evaluators, while Elon Musk and Demis Hassabis also voiced support for Dario Amodei's call to pace AI development. Here is how their statements differ, and what would make oversight effective.
Anthropic CEO Dario Amodei has proposed slowing AI development and embedding external evaluators inside the company. We compare the plan with METR's post-incident investigation, looking at access to training, publication rights, the reality of self-improvement, and the difficulty of international coordination.
Google's TimesFM-3 is a time-series forecasting model that handles sales, related series and planned promotions without extra training. We look at how it differs from the XReg-based TimesFM-2.5, how it forecasts, what its benchmarks compare, and its non-commercial weight license.
Toshiba's U.S. subsidiary and DIMAAG are teaming up on SCiB-based storage for AI data centers. We look at the cooling behind continuous 10C charging and discharging, how it fits with 800V power delivery, and the conditions behind the sub-1% claim.
OpenAI has released the Agents API, which exposes the execution infrastructure behind Codex. This article explains how it differs from the Agents SDK, the lifespan of sessions and working files, and the data retention conditions that apply even when running in your own environment.
Graphics card shipments rose in Q2, according to JPR data and TrendForce's memory supply analysis. We look at the rush-buying hypothesis, discrepancies in the published figures, and what the card-to-CPU shipment ratio can and cannot show.
As demand for AI infrastructure surges, Samsung Electro-Mechanics is pushing to sharply expand sales of MLCCs for AI servers, while Murata Manufacturing moves to discontinue nine product lines. Here's a look at the supply crunch and structural shift now reshaping the passive components market for the first time since 2018.
OpenAI has announced Data agent for ChatGPT Work. It supports questions about internal data, dashboard creation, and approved actions. Here is what it can handle in conversation, and where companies still need to set up definitions, permissions, and verification.
