Every story we have kept, newest first.
Looking for the daily editions? → Past editions
Page 119 of 1374
Digital Commerce, operator of the adult e-commerce site FANZA, plans to launch FANZA Studio, an AI content platform for adult content. A beta early-access trial is scheduled to begin on August 24.
ITmedia AI+ (日本) · 31d ago
Jeff Dean discusses his departure from Google in his first public interview since leaving the company. He highlights the ability of small teams to maintain extreme focus as one reason for the move.
量子位 · 31d ago
AI-powered sizing tools aim to help shoppers choose the right clothing size online. Brands hope better recommendations will improve customer satisfaction and reduce costly returns.
Bloomberg Technology · 31d ago
This arXiv paper introduces bounded sovereignty, a framework for evaluating AI oversight when organisations use models through managed APIs without owning the weights or infrastructure. Its synthetic experiments show that logs, intervention gateways, trace access, and model-version controls materially affect safety, diagnosis, and post-incident accountability.
ArXiv AI · 31d ago
Researchers evaluated nine open- and closed-source LLMs on generating realistic air traffic control transmissions for a general-aviation flight over San Francisco’s Bay Tour route. Worked examples improved performance, while heavily scripted prompts often amplified errors unless corrected dialogue history was injected.
ArXiv AI · 31d ago
This paper frames context acquisition as active inference, enabling agents to choose among clarifying questions, retrieval, tool calls, task actions, or stopping. It introduces an Optimal Question Asking benchmark and evaluates language models on tasks involving 25 to 300 candidates, as well as clarification before generation and prompt optimization.
ArXiv AI · 31d ago
Symposium is a formal framework and implementation for maintaining immutable, auditable records of AI-agent-driven scientific research. It captures claims, evidence citations, assumptions, and permissible evidence usage so communities can assess trust and build on prior work.
ArXiv AI · 31d ago
R2-OPD improves on-policy distillation by suppressing teacher feedback that conflicts with a model's actual reasoning progress. It compares teacher-derived reward rankings with independently estimated progress rankings to provide more reliable supervision.
ArXiv AI · 31d ago
Outcome Monitors detect tool results that violate expected outcome contracts, preserving the result while issuing a recovery receipt. In evaluations, they more than doubled ToolMaze completion and improved tau-bench retail completion by up to 14 percentage points.
ArXiv AI · 31d ago
Researchers introduce Scientific Data Skill (SciDSK), a reusable representation that gives AI agents dataset context, procedures, quality checks, and provenance. The Scientific Data Skill Bank publishes these resources across six disciplines and improves automated dataset discovery and interpretation.
ArXiv AI · 31d ago
This arXiv review examines robust population-based metaheuristics for berth allocation and quay crane assignment under uncertain vessel arrivals, handling times, and resource availability. It organizes existing methods, introduces an uncertain BACAP benchmark suite, reports baseline comparisons, and highlights open research challenges.
ArXiv AI · 31d ago
This research introduces a black-box method for classifying unsafe LLM outputs using prompt-response embedding dynamics and Koopman-based models. Evaluations across three safety benchmarks show that prompt information improves detection of interaction-dependent violations, while dense semantic embeddings work better for response-only violations.
ArXiv AI · 31d ago
Researchers introduced a pose-based computer-vision framework that separates dairy-cattle interactions into affiliative and agonistic social networks. Analysis of 1,183 interactions among 36 cows showed that pooled interaction counts can hide substantially different community structures and individual roles.
ArXiv AI · 31d ago
This research proposes telemetry-informed runtime policies that dynamically adjust retrieved-context compression for edge-based RAG. Experiments on NVIDIA Jetson AGX Thor show that intermediate compression can reduce GPU energy by up to 53.2% and SoC energy by up to 48.2% with negligible quality loss.
ArXiv AI · 31d ago
This arXiv paper proposes shifting AI consciousness debates toward the more tractable question of AI valence. Developers can assess whether an AI has states that would be positive or negative experiences if it were conscious, enabling more responsible development practices.
ArXiv AI · 31d ago
Apple Music appears to be focusing more closely on AI-generated music and does not want such tracks to escape its oversight. The update could affect how AI-created songs are identified, labeled, or distributed on music platforms.
Ifanr (爱范儿) · 31d ago
The article reports that ChatGPT can access text messages through app-level integrations. It raises concerns about how much personal data AI assistants may obtain from apps and calls on Apple to strengthen its safeguards.
Ifanr (爱范儿) · 31d ago
The article examines how domestic computing infrastructure can meet the rising demands of AI4S, particularly in marine computing scenarios. It highlights that AI significantly increases the complexity beyond existing requirements for stability and real-time performance.
钛媒体 · 31d ago
China continues to lead global patent applications, with AI and other core technology areas highlighted as key strengths. Global patent filings reached 3.72 million in 2024, more than double the 1.7 million recorded in 2005, while Japan’s share declined from 25% to 8%.
cnBeta (Full RSS) · 31d ago
Counterpoint Research predicts Samsung will regain the global smartphone market lead this year. Despite a projected 14.3% decline in global smartphone shipments in 2026, Samsung’s shipments are expected to grow 0.8% due to its integrated supply chain and distribution network.
cnBeta (Full RSS) · 31d ago
FAW Hongqi has launched the refreshed HS5 SUV at a promotional price starting at 134,800 yuan. The vehicle adds Lingxi Smart Cabin 5.0, Hongqi Sinuo intelligent driving, upgraded displays, and a 2.0T powertrain.
IT之家 · 31d ago
Jingxin Microelectronics has successfully powered on its domestically developed RapidIO 3.2 switch chip SR1820 and RapidIO 3.2-to-PCIe 4.0 bridge chip ST0420. The chips completed all primary functional use cases in under one hour after returning from fabrication.
IT之家 · 31d ago
China’s prefabricated computing infrastructure reportedly achieves over 90% factory prefabrication and can enter operation within 24 hours. The approach is said to improve delivery efficiency by 70%.
量子位 · 31d ago
UBTECH’s Walker embodied intelligence foundation model has completed China’s generative AI service filing. The milestone places the model among the first services to complete this regulatory process.
量子位 · 31d ago
Anthropic has launched Claude Academy, a free official learning site covering AI fundamentals and Claude-related products. The site provides service-specific guidance for tools such as Claude Code and Cowork.
ITmedia AI+ (日本) · 31d ago
At WRC2026, UBTECH is showcasing humanoid robot applications across industrial, commercial, and home-consumer scenarios. The presentation emphasizes a broad, multi-market deployment strategy.
量子位 · 31d ago
NetEase Youdao reported Q2 2026 net revenue of RMB1.47 billion on August 20. The company said its AI-native strategy is accelerating the realization of commercial value.
量子位 · 31d ago
Shanghai is advancing a fourth-generation semiconductor ecosystem in Lingang, building on early research efforts launched in March 2024. The initiative focuses on gallium oxide and diamond materials, with incubators such as Sinan helping move laboratory results toward commercialization.
Pandaily · 31d ago
Baidu has upgraded its app with interactive, GUI-based search components and an improved ERNIE Teacher. ERNIE Assistant task engine 2.0 now handles end-to-end office workflows, from information gathering to final delivery, at no cost.
Pandaily · 31d ago
As industrial systems, infrastructure, robots, and AI agents gain the ability to change physical conditions, cybersecurity must protect actions and outcomes—not only data and access. The article argues that future defenses need to evaluate whether an authorized action is appropriate for the current environment, state, and safety boundaries.
虎嗅 · 31d ago