Tag: #ai

Preprint finds probabilistic AI models outperform ECMWF ensemble on record-breaking heat extremes

An arXiv preprint shows probabilistic AI forecasting systems outperformed ECMWF's ensemble — including on record-breaking heat — when evaluated probabilistically.

#ai, #weather, #heatwaves, #forecasting

OpenAI spotlights Amp as example of agents using GPT-6 Astra, GPT-Live and Sign in with ChatGPT

OpenAI profiles Amp, which uses GPT-6 Astra, GPT-Live-1, Ultrafast and Sign in with ChatGPT to run persistent agents for writing, design and voice.

#openai, #ai, #agents, #amp

GAO: Pentagon’s Responsible AI Effort Hobbled by Inactive Council and Staffing Shortfalls

GAO finds the Pentagon's Responsible AI effort hindered by a months-long council hiatus and major staffing gaps as DOD scales AI use.

#defense, #ai, #pentagon, #gao

ArXiv preprint: Black-box system REPRO finds single-turn jailbreaks across 36 LLMs, but success depends on judge and stopping rules

An arXiv preprint says REPRO found single-turn jailbreaks across 36 LLMs with a median 24 queries — but reported success hinges on the authors’ judge and stopping rules.

#ai, #jailbreak, #llm, #arxiv, #modelsafety

Meta unveils Muse Wonder and Muse Wonder Full, agentic AI models with 1M-token context

Meta releases Muse Wonder and Muse Wonder Full — agentic AI models with Active Memory, Active Focus and a 1 million-token context window.

$META, #ai, #meta, #muse, #agents

Preprint: Frontier LLMs Can Propose High‑Risk Single‑Site Protein Mutations; Tool Access Raises Risk

An Oct. 9 arXiv preprint finds leading LLMs can propose high‑risk single‑site protein mutations in a benchmark, and tools raise that risk.

#ai, #biosecurity, #mutations, #arxiv

Lancet commission proposes unified framework to compare 21st-century global health threats

A Lancet Commission will unveil a DALY-based framework to compare long-range health threats — from pandemics and climate to AI misuse and antimicrobial resistance.

#globalhealth, #pandemics, #climatechange, #ai, #antimicrobialresistance

Preprint says 'BRANCH' can bypass multi-scanner AI guardrails, including some commercial products

An arXiv preprint from Mindgard and Lancaster University describes BRANCH, which its authors say can evade multi-scanner AI guardrails and transfer to some commercial products.

#ai, #cybersecurity, #moderation, #guardrails

White House packages $6 billion-plus science agenda around AI, virtual biology and space reactors

White House bundled $6B+ in federal, private and philanthropic commitments around AI-for-science — virtual biology, quantum computing and space reactors.

#ai, #virtualbiology, #quantum, #space, #government

NVIDIA and Microsoft Push Windows PCs as Local AI-Agent Platforms; RTX Spark Laptops Due Oct. 16

NVIDIA and Microsoft teamed up to position Windows PCs as secure hosts for local AI agents; RTX Spark laptops opened preorders with first units due Oct. 16.

$NVDA, $MSFT, #nvidia, #microsoft, #ai, #windows

Google Launches Public SynthID Detector to Identify AI Watermarks

Google launched a public SynthID Detector website to check whether images, video and audio carry AI watermark signals from Google and partner systems.

#ai, #watermarking, #google, #synthid

Report: Nemotron fine-tuned to gold level at IMO; IOI run was an unofficial benchmark

NVIDIA and Hugging Face released papers and models claiming fine-tuned Nemotron hit gold-medal scores at IMO (official) and at IOI (unsupervised, unofficial).

$NVDA, #ai, #nvidia, #huggingface, #olympiad, #nemotron

Google DeepMind Releases EmbeddingGemma 2, a Local Multimodal Embedding for Text, Image, Audio and Video

Google DeepMind released EmbeddingGemma 2, an Apache-2.0 multimodal embedding model developers can run locally for privacy-first search and retrieval.

$GOOGL, $GOOG, #ai, #embeddings, #google, #multimodal

Abu Dhabi's TII Releases Falcon-Emirati-7B, a Language Model Tuned to Emirati Arabic

Abu Dhabi’s Technology Innovation Institute released Falcon-Emirati-7B, a 7B-parameter model tuned to Emirati Arabic, claiming top scores on local benchmarks.

#ai, #arabic, #uae, #language-models

Open-weight AI structures 2.18 million German radiology reports on hospital hardware

German researchers used an open-weight LLM on local hospital servers to convert 2.18M radiology reports into structured data, avoiding external AI services.

#ai, #radiology, #healthit, #medical-records