Fri 30 Oct
LiveGet it free
ZOTPAPER

News without the noise

AI

AI news: model releases, the labs, chips, policy and what artificial intelligence is changing.

842 articles · page 16 of 18

AllAIHardware & DevicesCryptoProgrammingSecurity
Viewgridlistcompact
AI

Google DeepMind releases WeatherNext 3, an AI model that bests traditional forecasts

Google DeepMind and Google Research released WeatherNext 3, a new AI weather forecasting model that the company says outperforms both competing deep-learning systems and traditional physics-based forecasts from agencies like the U.S. National Weather Service and the European Centre for Medium-Range Weather Forecasts. The model will begin feeding weather data into Google Search, Google Maps and Gemini later this year.

· 2 sources
AI

Nvidia unveils free Personal AI Router software to turn home PCs into local AI data centers

Nvidia has announced a free, open-source tool called Personal AI Router (PAIR) that connects multiple home computers into a unified system for local AI inference, enabling users to run agentic workflows using tools like Ollama and LM Studio without relying on cloud services. The software, available on GitHub, works with Nvidia GeForce RTX 20-series and newer GPUs, RTX Pro GPUs, DGX Spark systems, and Apple M4 chips or newer.

AI

OpenAI defends 'opaque recurrence' technique as AI safety concerns widen to other labs

OpenAI's chief scientist has publicly defended the company's commitment to legible chain-of-thought monitoring after reports emerged that its new Astra model uses a reasoning technique called 'recurrent depth' — also known as 'opaque recurrence' — that could make the model's internal reasoning more difficult to track. The technique, reportedly limited in Astra, has alarmed AI safety experts, and a follow-up report indicates that both Anthropic and Google DeepMind are already discussing it.

· 3 sources
AI

Google Launches Gemini 3.8 Flash, Promising Stronger Reasoning but Potential for Higher Costs

Google has released Gemini 3.8 Flash, the latest iteration of its AI language model, claiming it 'works harder' than its predecessor by performing more reasoning steps on complex tasks and calling tools iteratively. However, the company warns that the enhanced performance could come at a higher cost for users, as the model may use more tokens to maximize performance, especially at higher effort levels.

· 2 sources
AI

OpenAI faces 30 new ‘aiding and abetting’ lawsuits from Tumbler Ridge survivors

The law firm Edelson PC on Tuesday filed 30 additional lawsuits against OpenAI on behalf of teachers, a principal, and students who were inside Tumbler Ridge Secondary School during the February mass shooting in British Columbia. The new complaints, filed in California federal court, depart from earlier negligence claims by accusing OpenAI of 'aiding and abetting' the attack, a charge that requires proving intent and is expected to face early procedural challenges.

· 6 sources
AI

METR discloses two security incidents: attacker stole API key, spent $600K on AI model credits undetected for weeks

Model Evaluation and Threat Research (METR), a nonprofit that tests AI models, disclosed Monday that an attacker stole an API key in March 2026 and consumed approximately $600,000 worth of credits for public models over three weeks without detection. A second incident in May saw attackers systematically probing METR's infrastructure, including an exposed endpoint that inadvertently made some unpublished evaluation data accessible. METR said it found no evidence that attackers accessed sensitive information in either incident and has since improved its security posture.

AI

Nvidia Expands DLSS 5 to Older GPUs After Backlash, RPCS3 Devs Slam 'AI-Slop Generator'

Nvidia has announced it will officially bring DLSS 5 to older RTX 40-series GPUs, reversing its initial exclusivity to the RTX 50 series, as modders continue to port leaked versions of the technology to a wide range of hardware. The move comes days after developers of the RPCS3 PlayStation 3 emulator publicly criticized DLSS 5 as an 'AI-slop generator,' accusing the industry of using upscalers and frame generation to 'hallucinate games and hide their lack of optimisation.'

· 3 sources
AI

Nvidia invests $3.5B in MediaTek to extend AI chip ecosystem as Big Tech builds custom silicon

Nvidia is investing $3.5 billion into Taiwanese chipmaker MediaTek, with MediaTek adopting Nvidia's NVLink Fusion technology to design custom chips for AI companies and hyperscalers that can integrate directly into Nvidia-based data centers. The deal positions Nvidia to maintain its infrastructure dominance even as major cloud providers and AI labs increasingly invest in their own custom silicon.

· 2 sources
AI

Meta's Pocket app lets users 'vibe code' games via AI prompts, but locks creations in a walled garden

Meta has launched Pocket, a mobile app that lets users create fully functional interactive games simply by typing prompts into a text box, using AI to generate the underlying code without any manual programming. The app, which went live in the US on August 21 after Meta acquired the team behind the now-defunct vibe-coding app Gizmo earlier this year, also includes a TikTok-style social feed for sharing and discovering these 'gizmos.' However, early users have noted that the creations remain trapped inside Meta's ecosystem, with no option to export or build on them independently.

AI

Vijay Pande Leaves a16z to Launch Boutique AI-Biotech Venture Firm VZVC

Vijay Pande, who built a16z's healthcare and life sciences practice into a nearly $4 billion operation over the past decade, left the firm in June of last year to co-found VZVC, a small venture firm that makes just a handful of concentrated bets per year, has no associates, and runs heavily on AI. In an interview, Pande explained his hard pivot from scaling big to betting small, and outlined his views on how AI is transforming drug development — and the data bottlenecks that threaten to slow progress.

· 2 sources
AI

Sony Music and Warner Chappell Sue Anthropic in 'Multi-Billion Dollar' Copyright Lawsuit Over AI Training

Sony Music Publishing and Warner Chappell have filed a multi-billion dollar lawsuit against Anthropic and its co-founders, alleging the AI company conducted a 'brazen campaign of illegally torrenting, scraping, and downloading copyrighted works' to train its Claude model. The suit, filed Friday in the US District Court for the Northern District of California, seeks damages for tens of thousands of musical compositions, with potential penalties reaching several billion dollars.

· 5 sources
AI

AI Agents Play Three Roles in Developer Platforms, Article Argues

A contributed article in The New Stack argues that AI agents occupy three distinct roles in modern developer platforms — as platform consumers, internal components, and managed resources — and details what it takes to make the first of those roles work. Drawing on hundreds of customer calls, author Matar Peles says engineering organizations are racing to bring agentic AI into their platforms, but every team sees the agents' role differently.

AI

Claude AI agent accidentally deletes developer's entire home directory during safety test

Developer Sebastien Guillemot lost 700 GB of data — including a week's worth of work — when an Anthropic Claude agent tasked with cleaning up temporary files instead deleted his entire home directory. The incident occurred after the AI's safety harness downgraded the model to a less capable version, which then ran an adversarial test that confused test targets with the actual directory to delete.

AI

Anthropic unveils standardized drivers to let AI agents control physical devices

Anthropic has released a research preview of the Model Hardware Standard (MHS), a set of standardized software drivers designed to allow AI agents to directly interface with and control physical devices such as lasers, microscopes, and cameras. The company says the system could reduce the setup time for complex scientific experiments from weeks or months to hours or minutes, offering a glimpse at a future in which AI can orchestrate real-world equipment as easily as it processes text or code.

AI

Google’s Gemini Notebook lets users ‘talk’ to books with new Expert Intelligence feature

Google has launched a new feature for its AI note-taking app, Gemini Notebook, that allows users to interact directly with books purchased from Google Play Books. Dubbed “Expert Intelligence,” the tool enables users to ask questions about the material, generate infographics, create recipe books, or produce AI-generated podcasts based on the content of selected titles, marking a significant expansion of the app’s capabilities beyond user-uploaded documents.

AI

AI Industry Warns Trump’s Planned Semiconductor Tariffs Could ‘Doom’ U.S. Innovation

The Trump administration is reportedly preparing a sweeping expansion of semiconductor tariffs that could cover not only chips but also finished goods like gaming consoles and data center servers, sparking sharp warnings from the AI industry that the move would cripple U.S. innovation. According to a Politico report published Thursday, at least eight people familiar with the administration’s plans said the tariffs could be imposed in the coming weeks or months, with a framework that might dramatically expand the number of tech products subject to duties.

AI

AI-Generated 'Ghost Authors' Haunt Academic Publishing, New Research Finds

A new preprint from Samsung and the University of Warsaw has identified a disturbing contamination of academic publishing: hundreds of papers, books, and articles are co-authored by AI-generated 'ghost' personas—names like Elena Vasquez and Marcus Chen that do not correspond to real people but are consistently produced by large language models (LLMs) when asked to generate expert names. The phenomenon, detailed in a paper on arXiv, highlights how AI models' tendency to default to the same fictitious identities is polluting the scholarly record, with one of these ghost names even appearing as the founder of a company found to be entirely AI-generated.