Tag: AI News

OpenAI launches GPT-Red – self-play red teaming to harden models OpenAI describes GPT-Red as an automated “self-improvement” system that attacks other models in simulation and uses the results to improve robustness against safety and prompt-injection failures. The key claim – GPT-Red boosted the company’s latest flagship models by giving them…

Siri AI Is Arriving as iOS 27 Enters Public Beta Apple’s first public beta of iOS 27 brings Siri AI to the wider world while keeping the update philosophy focused on speed and fixes rather than flashy new features. The Verge’s hands-on reports tangible improvements in everyday flows like Photos…

A Formalization of the Mean-Field Derivation of the Vlasov Equation – AI-Assisted Lean Formalization as a Strategy Game A mathematician “plays” a strategy game in Lean 4: direct an AI to turn a LaTeX proof into machine-checked Lean code with no sorrys, where the build itself certifies correctness. The case…

Apple’s failed self-driving car program left a legacy of powerful AI chips Apple’s Project Titan never became a driverless product, but it pushed the company hard toward on-device AI processing. That effort helped shape the Neural Engine, which debuted in the iPhone X era and went on to power Apple’s…

Agentic AI for Straight-Through Underwriting An arXiv study proposes an agentic pipeline for straight-through underwriting that compares a single-LLM approach, naive RAG, and a multi-agent “Agentic RAG” system with targeted retrieval, third-party checks, and explicit multi-step rule evaluation. The agentic setup delivers the biggest improvements when information is missing or…

OpenAI shuts down ChatGPT Atlas browser less than a year after launch OpenAI is sunsetting ChatGPT Atlas, the browser feature meant to execute tasks on your behalf, with deprecation targeted for August 9. The move is part of a consolidation push toward ChatGPT Work, folding capabilities from app, coding tools,…

AgentLens: Production-Assessed Trajectory Reviews for Coding Agent Evaluation Instead of judging coding agents with a single pass/fail bit, AgentLens evaluates the full “trajectory” – instruction following, tool use, self-verification, recovery from mistakes, and the explanations given along the way. The benchmark pairs formal checks with LLM-written trajectory reviews, making it…

Narrative World Model: Narratology-Grounded Writer Memory for Long-Form Fiction The Narrative World Model (NWM) targets a specific gap in story-aware AI: general retrieval systems can’t reliably answer multi-hop questions that depend on narratology structure, like who learned what secret and when it was revealed. By combining a narratology-grounded typed temporal-state…

SwarmResearch: Orchestrating Coding Agents for Open-Ended Discovery SwarmResearch tackles a core failure mode of long-running coding agents: they often get stuck on one high-level strategy and then churn through low-level edits. A “Shepherd Agent” steers a population of subagents across separate git branches using global context, leading to better or…

PRX Part 4: Our Data Strategy PhotoRoom’s PRX Part 4 breaks down how it plans, selects, and curates data to improve AI performance while keeping downstream quality in mind. The focus is on turning messy real-world image inputs into training signals that produce steadier results, not just bigger models. Read…

Infuriating Google commercial imagines the founding fathers embracing AI Google’s new Workspace ad turns the drafting of the Declaration of Independence into a timeline of AI-powered collaboration, with Gemini transcribing notes into docs, suggesting edits, and even generating a new national seal. The Verge argues the pitch is deliberately anachronistic…

Fanfiction community turns “AI detection” into a witch hunt A new fanworks crackdown is targeting authors suspected of using generative AI, but the so-called “detectors” being shared are unreliable enough that almost any writer could be wrongly accused. The Verge reports that the effort has quickly escalated beyond fair provenance…

Midjourney Medical’s Ultrasound Scanner: The Promising Tech, the Missing Proof Midjourney Medical shows a behind-the-scenes build of its dunk-tank ultrasound scanner setup, describing a hacked-together array of ultrasound probes plus off-the-shelf compute. The video is a rare look at the hardware pipeline, but it leaves open the big question: where…

Solution space path planning for supporting en-route air traffic control A new conflict-free path-planning approach for en-route ATC is built around “solution-space path planning” so controllers can see all feasible safe actions while shifting operational priorities like separation, maneuver limits, and routing practicality. The method integrates multiple intent-based conflict detection…

ScarfBench: Benchmarking AI Agents for Enterprise Java Framework Migration ScarfBench tests whether AI agents can reliably support enterprise-scale Java framework migrations using realistic tasks and evaluation. The benchmark is designed to stress real-world constraints where agent “it works in demos” isn’t enough, pushing for measurable migration correctness instead of generic…

MER-R1: Multimodal Emotion Reasoning via Slow-Fast Thinking Synergy Researchers show that “slow thinking” – deliberate multimodal reasoning – doesn’t automatically improve multimodal emotion recognition, and that fast, direct answering can outperform after deliberation. MER-R1 uses reinforcement learning to jointly optimize recall and precision with slow-fast confidence calibration, hitting state-of-the-art results…

Prosecutors used ChatGPT logs as evidence in the Palisades fire trial In the Palisades fire arson case, prosecutors went beyond standard digital forensics by introducing ChatGPT conversation logs, including prompts about anger, images of fire, and screen recordings about responsibility for a blaze. The trial also highlights how AI chat…

Why is Apple asking me to pay more for Big Tech’s AI obsession? Tim Cook points to AI-driven component stress as a key reason for Apple’s recent price hikes, framing them as “unavoidable” even as critics call them “unsustainable.” With products like the 16-inch MacBook Pro, 11-inch iPad Air, and…

Accelerating Skill Assessment in Chess: A Drift-Diffusion-Enhanced Elo Rating System DD-Elo upgrades traditional Elo by folding move-by-move information into rating updates using a drift-diffusion model-inspired framework, aiming to measure fast skill shifts without being drowned by noisy game-state variation. The authors prove bounded deviation from classic Elo while showing faster…

Ford had to hire back former engineers to fix mistakes made by its automated systems Ford is acknowledging that automated systems in production and design weren’t as robust as expected, forcing the company to bring experienced technicians back in to correct robot-made errors. The reporting points to a hard lesson…

Accelerating Transformers Fine-Tuning with NVIDIA NeMo AutoModel NVIDIA’s NeMo AutoModel aims to speed up transformer fine-tuning by automating key model-selection and optimization steps, reducing the amount of manual trial-and-error teams typically need. The pitch is practical: faster iteration loops for adapting LLMs to new tasks, backed by an end-to-end workflow…

On the Identifiability of User Adaptation in Co-Adaptive Neural Interfaces This paper shows that closed-loop encoder estimates in co-adaptive brain or body interfaces do not uniquely reveal what the user is adapting – the measurements conflate user changes with properties of the joint human-machine system. The result is a warning…

Three things to watch amid Anthropic’s latest feud with the government Anthropic’s dispute with the US government has moved beyond policy arguments into practical questions about model oversight, compliance, and how enforcement will work for frontier labs. The story breaks down what to monitor next – potential regulatory pressure points,…

Samsung Electronics brings ChatGPT and Codex to employees Samsung is rolling out OpenAI’s ChatGPT Enterprise and Codex across its workforce worldwide, positioning the deployment as one of OpenAI’s largest enterprise AI initiatives to date. The move signals a fast pivot from experimental pilots to broad internal adoption, putting generative AI…

REVEAL++: Differentiable Phenotypic Grouping for Vision-Language Retinal Modeling of Alzheimer’s Disease Risk REVEAL++ replaces hard “phenotypic grouping” in retinal vision-language contrastive learning with a continuous, learnable weighting of inter-subject similarity derived from both retinal embeddings and clinical risk profiles. The resulting soft multi-positive objective enables graded supervision across the spectrum…

Diffusion Language Models: An Experimental Analysis A new head-to-head study benchmarks eight diffusion language models across reasoning, coding, translation, knowledge, and structured tasks, explicitly factoring generation quality against compute efficiency. The authors show inference-time knobs—like denoising steps, context length, and parallel unmasking—can dominate outcomes, forcing clear trade-offs between performance and…

Skill-Constrained Model Predictive Control for Resilient Manufacturing Supply Chains This work tackles production planning when human certifications decay over time and training competes with production capacity, framing workforce capability as a real operational bottleneck. The authors benchmark a closed-loop skill-constrained model predictive controller against production-only, maintenance-only, and static “skill insurance”…

SpeechDx: A Multi-Task Benchmark for Clinical Speech AI SpeechDx introduces a large-scale clinical speech benchmark spanning 12 datasets and 27 tasks, organizing evaluations by the specific stage of speech production that each condition disrupts (conceptualization, formulation, articulation). Testing across limited-label settings and cross-dataset transfers, the authors find large-scale speech models…

Apple 2027 rumors: AirPods with cameras for AI and the second folding iPhone Bloomberg’s Mark Gurman shares insights on upcoming Apple hardware, including AirPods with integrated cameras and a second folding iPhone, both slated for a late 2027 release. This development follows AI enhancements introduced at WWDC. Read the full…

Anthropic Faces Shutdown of AI Models Amid Government Orders Anthropic has taken its newly launched AI models, Fable 5 and Mythos 5, offline following a White House order requiring the company to restrict foreign access, even to its own employees. This shutdown underscores the intense regulatory environment surrounding AI technologies…

China May Have Accessed Mythos A report from Semafor reveals that the White House’s decision to enforce export restrictions on Anthropic’s Mythos was partially influenced by concerns that a group linked to China may have gained access to the AI model. If true, this could pose significant national security risks.…

Anthropic Cuts Off Fable 5 and Mythos 5 Access Following Government Order Anthropic has halted access to its AI models, Fable 5 and Mythos 5, for all foreign nations as mandated by a government order citing national security concerns. This decision affects both the company’s customers and employees, signaling increased…

Siri Has Improved Significantly Apple has released a new version of Siri that showcases the virtual assistant’s remarkable enhancements, contrasting its previous ineffectiveness. Users can now expect a more reliable experience, which might finally meet their needs for practical tasks. Read the full article here Elon Musk Becomes the World’s…

Amazon’s Data Centers Used 2.5 Billion Gallons of Water Last Year Amazon disclosed that its global data center operations consumed 2.5 billion gallons of water in the last year. This revelation comes as Seattle’s data center moratorium stirs concerns over water and energy use amidst AI data center construction debates.…