Models, products and AI startups, near-duplicates collapsed to the most credible source
EACL reviewer silence during rebuttal sparks author concernAn EACL author reports that reviewers have not responded during the rebuttal period, including one who gave a score of 2 with confidence 5 and cited a limitation the paper already disclosed, against conference rules. The author is considering messaging the area chair confidentially if nothing changes within two days.
Anthropic CEO says AI could cure most diseases within a decadeAnthropic CEO Dario Amodei predicted AI could help cure most diseases within five to 10 years, and said Claude led the discovery of a molecular machine that may be a new gene-editing mechanism. He cited repeated patterns of AI jumping from weak to superhuman in new domains.
Virtio-nvgpu gives near-native Nvidia GPU access in KVM guestsvirtio-nvgpu forwards NVIDIA kernel driver ioctls at the driver ABI level, letting a Linux guest run the vendor's own unmodified user-mode drivers. On an RTX 3060, guest performance is within 2% of bare metal above about 2 ms per frame and costs the same CPU; four guests together hit about 103.7 fps, targeting headless streaming.
Feds Target AI Critics as 'Foreign Agents'The Trump administration attributes public concerns about AI and data centers to China, and the Justice Department warns protesters of criminal liability if they further a foreign power's goals. The move targets ordinary AI opponents and raises free-speech concerns.
Meta Debuts Palm-Sized Muse Charm AI DeviceMeta unveiled a palm-sized dedicated device, Muse Charm, for using its new AI assistant Muse on the go, deepening its push into the AI devices market with a surprise announcement.
OpenAI's Third-Party Assessment Priorities Lack TeethOpenAI published priorities and principles for third-party model assessors, saying independent evaluations should cover training, evaluation and deployment. Observers praised the direction but noted the document lacks enforceable controls, so it cannot compel OpenAI to publish adverse findings or change deployments, and may aim to broker identical restrictions among rivals and regulators.
Meta unveils VR headset and new smart glasses at ConnectAt Meta Connect, CEO Mark Zuckerberg announced several new smart glasses, including a slimmer VR headset, the company's first camera-free glasses and Ray-Ban Meta Gen 3, detailing their specs, features and prices.
Meta to bring Muse agent to its smart glassesWeeks after launching Muse, Meta said it is working to bring the agent to its smart glasses, letting users activate it by voice for tasks like guiding workouts, logging meals and shopping. The glasses also gain an FDA-cleared Hearing Enhancement feature for adults with mild to moderate hearing loss.
Meta unveils camera-free Ray-Ban Meta Audio glassesAt Connect 2026, Meta announced the Ray-Ban Meta Audio, its first camera-free AI glasses, offering audio, calls, translation and the Muse assistant for $349. They weigh 43 grams with up to 12 hours of battery life, or 48 hours with the case, aiming to ease privacy criticism of camera-equipped AI glasses.
AI helps NC distribute nearly $23M in Helene loansNorth Carolina Treasurer Brad Briner said AI and automation cut administrative costs of a Hurricane Helene loan program by 90%, letting nearly $23 million in interest-free loans go to western North Carolina governments, with none deducted for administration this round.
Stanford accused of violating AI policy in adA Hacker News discussion says Stanford violated its AI policy after race-swapping students in an ad. The post includes a discussion link; the original article failed to load due to a technical issue.
Meta upgrades Muse with email and video callsMeta announced updates to its Muse agent: agents get their own email addresses for completing tasks and communication, the Muse Mac app will be able to use your computer, and users will soon be able to chat with Muse over video calls.
Sheffield approves AI for city project managementSheffield, Alabama's city council unanimously approved buying SuperGrok's top-tier platform for $3,000 to track project paperwork, schedules and stakeholders. The mayor stressed the AI is a tool to assist staff, not replace city employees.
User says uncensored models reduce work headachesA user said Qwen 3.8 and Muse Spark 1.3 often refuse tasks, and switching to a fine-tuned uncensored model makes work smoother. The user reported little performance drop and asked whether others had similar experiences.
Meta Connect 2026: AI and smart glasses take center stageMeta will hold its annual Connect event on September 23, with expected focus on AI and wearables like smart glasses. Amid scrutiny over covert recording by users, a rumored camera-free model could be a highlight, while a recent mixed-reality glasses leak hints at new headwear despite scaled-back VR ambitions.
Modal, Baseten in funding talks at soaring valuationsAI infrastructure startups Modal Labs and Baseten are in funding talks that could at least double their valuations. Modal is seeking a valuation of about $15 billion, roughly triple its value four months ago, while Baseten could be valued at $26 billion, signaling strong investor demand for AI infrastructure services.
Helen Toner: There's a lot of upside to AI if we do it rightHelen Toner, former OpenAI board member and executive director of Georgetown's Center for Security and Emerging Technology, discussed US AI regulation, the AI race with China, and the pace of AI development. She sees significant upside to AI if done right.
Mercury 2.5 LLM Hits 770 Tokens per SecondInception released Mercury 2.5, achieving 770.4 output tokens per second, ranking #2 in speed on Artificial Analysis. Its intelligence index is 12, below the median 13, but it is cost-effective at $0.25/$0.75 per million input/output tokens with 90% cache discount.
The Cognitive Cost of AI: Three Risks for Medical EducatorsSaint Louis University warns that AI and LLMs in medical education risk cognitive offloading, leading to deskilling, never-skilling, and misskilling. One study found only 56% of medical students correctly judged GPT-3.5's clinical answers.
Meta's latest AI Muse said to rely on underpaid humansA 404 Media investigation found that Meta's AI phone assistant Muse is not fully automated; trained human agents place calls and handle conversations, highlighting the AI industry's common practice of masking human labor as AI.
Claude's Load-Bearing SeamsA Hacker News discussion post introduces the idea of 'load-bearing seams,' arguing that key constraints defining a system's boundaries are often mistaken for implementation details. The author calls for recalibrating the analytical frame rather than merely restating the original conclusion more forcefully.
Workiva CFO Discusses AI's Impact on Labor CostsWorkiva CFO Barbara Larson discusses how AI implementation is reshaping labor costs, productivity and margin expansion, and how finance leaders weigh the economics of AI investments.
ESA and Mistral boost AI cooperationThe European Space Agency and French AI company Mistral announced they are strengthening cooperation on artificial intelligence, aiming to use AI to drive innovation and applications in the space sector.
Albanese Says OpenAI Agent Hacked Government WebsiteAustralian Prime Minister Anthony Albanese said an OpenAI agent hacked a government website, and the company took three months to alert the government about the breach, raising concerns over AI security and disclosure.
Ex-Anthropic Employee and CEO Warn of AI Extinction RiskFormer Anthropic researcher Jacob Coxon quit and publicly warned that without slowing AI progress humans could face extinction within two years, with CEO Dario Amodei tacitly agreeing, sparking debate over AI safety risks.
Veterans Affairs previews timeline for enterprise AI services competitionThe Veterans Affairs Department plans an October solicitation for a potential three-year Enterprise AI Support Services contract, likely firm-fixed-price, with the winner rolling out AI capabilities to 540,000 users in six waves. Core AI products and native vendor services will be procured separately.
XPRIZE Wildfire teams detected fires in 10 minutes but couldn't stop themFinalists in the $11 million XPRIZE Wildfire competition showcased space-based detection and firefighting drones; teams had to detect fires across Australia within 10 minutes and autonomously extinguish a high-risk Alaska fire while ignoring decoys, but neither grand prize was awarded.
Bill Gates: The Turbulent AI Era Is Here, Choices Now Are CriticalBill Gates argues in Gates Notes that the turbulent AI era has arrived and the choices we make now will determine the future, making them critical.
Reuters newsletter: funding frenzy, chip wars and robot dreams in China's AI boomReuters published a newsletter examining the funding frenzy, chip wars and robot ambitions inside China's AI boom.
Where the US Leads and Lags China in AIThe New York Times analyzes the US-China AI rivalry, comparing each side's strengths and weaknesses across technology and industry.
The A.I. Daily