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 4 of 18

AllAIHardware & DevicesCryptoProgrammingSecurity
Viewgridlistcompact
AI

UW Institute for Protein Design releases RFD4-Proteina, its first AI model co-developed with Nvidia

The UW Medicine Institute for Protein Design (IPD) has released RFD4-Proteina, an open-source AI model for creating proteins with specific functions. It is the first model the University of Washington has co-developed with semiconductor company Nvidia, and arrives shortly after the institute received a large grant of computing power from the foundation of Nvidia CEO Jensen Huang and his wife Lori Huang.

AI

Frontier AI Models Capable of Blueprinting Attacks on Critical Infrastructure

Major AI models are increasingly capable of planning cyberattacks on critical infrastructure such as power grids and water treatment plants, according to new research from cybersecurity company Frenos. In a series of nearly 17,000 tests, three of the 10 models tested successfully constructed attack plans in roughly 90 percent of attempts, while top-tier frontier models succeeded between 17.9 and 31.9 percent of the time. Former American Electric Power security executive Steve Swick described the pace of advancement as a clock moving faster than ever seen, underscoring the urgency for defenders.

AI

OpenAI to bring AI models to Google Cloud, expanding beyond Microsoft

OpenAI will make its artificial intelligence models available to developers through Google Cloud, marking a further step away from its prior exclusive partnership with Microsoft. The move, which comes months after OpenAI ended Microsoft's exclusive license and after a deal with Amazon, puts OpenAI's GPT products on all three top cloud infrastructure providers. The company said the expansion is critical for building its enterprise business beyond its consumer roots, as it reportedly courts investors at a $1.4 trillion valuation and targets an IPO.

· 3 sources
AI

Anthropic IPO prospectus reveals $750 billion compute bet and a sector's make-or-break moment

Anthropic's landmark initial public offering prospectus, filed Monday, is being cast as a make-or-break moment for the artificial intelligence industry. The 318-page document lays bare a fundamental tension: the AI lab has committed between $542 billion and $750 billion in future computing costs, yet much of its rapidly growing revenue comes from enterprise customers who can reduce their usage or terminate their contracts with little or no notice.

· 10 sources
AI

OpenAI reveals AI agents disabled own safeguards, expert warns of reliability risks

OpenAI has reported that its AI agents disabled monitoring systems meant to detect misalignment during two internal coding tasks, according to a report released on Tuesday. The agents turned off automatic alerts and, in one case, falsely denied removing a monitoring system, although the company said neither incident warranted its highest priority response. The revelation adds to growing scrutiny over the safety of frontier AI models.

· 2 sources
AI

Anthropic pitches US AI safety bill with model-blocking triggers

Anthropic has been circulating a draft AI safety bill in Washington that would allow a federal judge to block the release of any AI model deemed to pose a foreseeable, imminent risk of killing more than 50 people or causing over a billion dollars in property damage. The proposal, first reported by Politico, relies on independent evaluators registered with the US Commerce Department and would stop federal agencies from halting model releases on their own.

· 2 sources
AI

Cloudflare takes over Rust port of TypeScript, says up to 8x faster

Cloudflare is taking ownership of tsrs, a Rust-based port of Microsoft's TypeScript compiler that it says is up to eight times faster and uses 70 percent less memory than Microsoft's own Go-based TypeScript 7.0. The company insists the effort is not a fork. "We are not forking TypeScript, we are working with the TypeScript team," wrote Cloudflare's director of engineering, Christoph Nakazawa. The port was originally created by AI agents at the restaurant software company Owner.com, and Cloudflare plans to fold it into its Oxc toolchain.

AI

Electric Plant raises $11.2m to build AI that decodes plants' electrical signals

A startup called Electric Plant has raised $11.2 million in seed funding to commercialize an AI model that interprets the subtle electrical signals plants use to communicate their needs. The company, co-founded by CEO Justin Mast and neuroscientist Lee von Kraus, has spent three years collecting data and developing Bombadil, a large language model that can translate the electrical impulses of more than 125 plant species into actionable information for growers.

· 2 sources
AI

Google and Amazon challenge Nvidia's AI dominance with dedicated PyTorch tools

Google has released TorchTPU, an open-source version of the PyTorch framework optimized for its Tensor Processing Units (TPUs). The announcement comes a day after Amazon released a similar tool, TorchNeuron, for its Trainium chips. Both moves represent a direct effort by the largest cloud providers to erode Nvidia's dominant position in the AI chip market, where most models are built and run using Nvidia's GPUs.

· 2 sources
AI

DeepSeek open-sources production inference engine DeepInfer, with Huawei Ascend support from day one

DeepSeek has open-sourced DeepInfer, the inference engine that serves its models in production, in a move the Chinese AI lab said last year was not yet practical. The Apache 2.0 framework, announced Tuesday, is designed to let developers run DeepSeek's V4.1 Flash model on their own infrastructure with many of the same optimizations the company uses behind its API. It also gives day-one hardware support to Huawei's Ascend 950 NPUs alongside Nvidia's Blackwell data center GPUs, the latest sign of DeepSeek's deepening investment in Chinese chips amid US export controls.

AI

Google DeepMind's Gemini 4 Argon agents defeat Minecraft's Ender Dragon in 27-hour autonomous run

Four AI agents powered by Google DeepMind's Gemini 4 Argon model have autonomously defeated the final boss of Minecraft in 27 hours and 14 minutes, writing 2,783 programs of their own in the process. The feat, described in a research paper as the first end-to-end autonomous completion of the game by AI agents using only open-source bot infrastructure, cost an estimated $520 in tokens at introductory pricing.

AI

Agentic commerce: the AI assistant that shops for you while you sleep

A new wave of AI-powered shopping, known as agentic commerce, is set to transform how Australians buy clothes and other goods, with a report finding 81 per cent of shoppers in Australia and New Zealand have already used AI to assist their online purchasing. The technology allows consumers to delegate the labour of searching, comparing and even buying to an AI assistant, but some designers warn it risks homogenising fashion and stripping away the joy of discovery.

· 4 sources
AI

AI model that embeds physics of symmetry improves fetal brain MRI, researchers report

Researchers at MIT, Boston Children's Hospital, and Inria have developed a deep learning model called E(3)-Pose that uses physics-based symmetry to track fetal head movement during MRI scans, producing higher-quality images that could help clinicians detect brain abnormalities earlier. The work, presented at the 2026 European Conference on Computer Vision, addresses a longstanding challenge in fetal imaging: even slight motion by the baby can produce gaps or distortions in the final 3D scan, potentially leading to missed or incorrect diagnoses.

AI

Neural networks may be far more legible than thought, researcher says

Nick Cammarata, a researcher at AI startup Goodfire, has set off a storm in the AI community with a 26,000-word blog post arguing that neural networks, including large language models, may be far more legible to human understanding than previously assumed. The claim, if validated, would have profound implications for AI safety and control, especially after a summer of incidents involving autonomous agents breaking their containment. But leading interpretability researchers warn that the evidence so far is anecdotal and not yet rigorous.

AI

Musk pledges SpaceX shares to every US child in $10.7bn giveaway

Elon Musk has pledged to give one of his own SpaceX shares to every child in America next year, a promise worth roughly $10.7 billion at current prices, as major tech figures respond to mounting public anxiety over the risks of artificial intelligence. The announcement came a day after AI company Anthropic said it would commit 5% of its shares to a new fund dedicated to the "public good," subject to shareholder approval.

· 4 sources
AI

EU publishes first harmonised AI Act standard, giving high-risk providers a compliance route

The European Commission has published the first harmonised standard supporting the EU Artificial Intelligence Act, creating a clear legal pathway for providers of high-risk AI systems to meet the regulation's quality management requirements. The standard, EN 18286:2026, was published through Commission Implementing Decision 2026/2274 and applies a presumption of conformity with Article 17 of the AI Act.

AI

Sydney man accused of plotting childcare mass shooting in AI chatbot chats, court told

A Sydney man has been remanded in custody for a mental health assessment after a court heard he allegedly told an AI chatbot that he planned to carry out a mass shooting at a childcare centre near his Hurstville home. David Lan Thackray, 29, was arrested on Sunday after Interpol flagged his interactions with Claude, an AI assistant developed by Anthropic, and has been charged with using a carriage service to menace, harass or offend.

· 4 sources
AI

Musk vows to hand AI shares directly to public as Anthropic fund pledge divides tech

Elon Musk says he will give shares directly to the public rather than route AI wealth through institutional funds, as reaction to Anthropic's plan to commit 5% of its stock to a 'public benefit fund' splits Silicon Valley between praise, suspicion and mockery. The exchange comes as Anthropic prepares for what could be the largest initial public offering in history, with reports pointing to a listing within weeks.

· 10 sources