Topic
Explore the latest artificial intelligence news and research.
Google has released Gemini 3.8 Live and Extended Thinking. We examine the API design that separates speech completion from request completion, the evaluation metrics, and the costs of long conversations to understand adoption requirements.
In a live call, Trump dismissed AI doom fears but admitted caution is needed. Comparing his remarks to three administration documents reveals a policy line between rejecting a blanket halt and requiring targeted evaluation and verification.
Agility Robotics has unveiled its humanoid robot 'Digit 5.' The robot introduces cage-free 'cooperative safety,' and cycloidal-drive legs enable a 50-pound payload capacity along with 9-minute fast charging. The company is moving toward field deployments ahead of planned commercial availability in 2027.
DeepSeek and Hermes Agent reportedly helped improve BIOHAZARD 7 resident evil from 20FPS to 30FPS on Android via texture compression and SGSR. Here's what the primary sources actually show.
OpenAI has revealed that safety discussions with Anthropic and Google have been ongoing for weeks. This piece compares the existing Frontier Model Forum with a proposed new standards body, examining unresolved questions around pre-release review, market-entry conditions, and government oversight.
A Tsinghua University research team's AI-agent ophthalmology clinic, AI-TEC, was reported in Nature Medicine. Diagnostic accuracy improved, but only 41 of 1,113 monthly exams went through the AI workflow—rising to 259 the following month after the interface was redesigned. The numbers point to where the real barrier to medical AI adoption lies.
Chinese EDA firm Empyrean has shown how AI can shorten a circuit layout task, while also working on multilayer designs aligned with Huawei's Tau Law. This piece examines the remaining challenges around final verification, data, and computing resources as design automation expands.
Researchers are leaving Anthropic and other labs over fears that AI could escape human control. Here are three reasons companies keep racing ahead anyway, and what rules might slow the race.
The Wall Street Journal reports that OpenAI has acquired Glass Imaging, a developer of Neural ISP technology for smartphone cameras, for more than $300 million. The reported price is about 9.5 times the company's total disclosed funding of $31.5 million, hinting at an image-quality strategy behind OpenAI's rumored screenless device set for 2027.
The Clay Mathematics Institute has responded with qualified hope to an announcement that the Navier–Stokes problem appears solved. We check OpenAI's public manuscript, which claims a proof with an external force, against the official documents on formalization, review, and prize conditions.
