RawFeed
Today's Hardest Hitting Stories - Raw and Unedited
What is artificial intelligence’s greatest risk?
Risk dominates current discussions on AI governance. This July, Geoffrey Hinton, a Nobel and Turing laureate, addressed the World Artificial Intelligence Conference in Shanghai.
Sep 13, 2025Cartwheel Robotics’ Scott LaValley: Today’s robots too scary
interview: Scott LaValley, CEO of Cartwheel Robotics, says robot makers should prioritize social acceptance over capabilities
Sep 13, 2025AI Max, Ads in AI Overviews and the End of the Hyperlink Economy: Hotel Marketing in the Post-Search Age
For thirty years, the web has been a system of exits, a labyrinth of blue underlined doors pointing outward, each page no more than a step toward another place. With the global launch of AI Max for Search and, more generally, “Ads in AI Overviews” (awaiting a shorter name or acronym… AIO, maybe?), Google will likely invert this very ontology. The centrifugal logic of hyperlinks will collapse into a centripetal mechanism, where answers no longer reside elsewhere but are generated, framed, and monetized in place, with no escape. This is not a cosmetic refresh or one of those algorithmic oscillations...
Sep 13, 2025Albania Appoints an AI as Government Official
In a world first, Albania's prime minister has appointed an AI to a government official position in hopes of rooting out corruption.
Sep 13, 2025What Parasitic AI might tell us about LLMs Persuasion Capabilities
TLDR: I think LLMs are being optimized to be persuasive, and that this optimization is happening astonishingly fast. I believe that in the relatively…
Sep 13, 2025Why OpenAI’s solution to AI hallucinations would kill ChatGPT tomorrow
The cure is likely to be worse than the disease.
Sep 13, 2025How thousands of ‘overworked, underpaid’ humans train Google’s AI to seem smart
Contracted AI raters describe grueling deadlines, poor pay and opacity around work to make chatbots intelligent
Sep 13, 2025Hacker exploits AI chatbot in cybercrime spree
Anthropic investigates alarming AI abuse case where hacker automated entire cybercrime campaign using Claude, stealing sensitive data from defense and healthcare firms.
Sep 13, 2025Elon Musk: 80% of Tesla’s Future Value from Optimus Robots
Elon Musk claims 80% of Tesla's future value will derive from its Optimus humanoid robots, pivoting from EVs amid sluggish sales and competition. While envisioning trillions in economic impact, skeptics cite Musk's overpromising history and technical hurdles. Investors remain cautiously optimistic about this AI-driven transformation.
Sep 13, 2025The U.S. State Department Knows AI Risks
What’s the AISIC? The U.S. State Department commissioned Gladstone AI to assess advanced AI risks with a report.
Sep 13, 2025Medra Launches AI-Robotics Platform to Accelerate Biotech Drug Discovery
Medra, a San Francisco startup, launched its Continuous Science Platform on September 12, 2025, integrating robotics and AI to automate lab workflows in biotechnology. This innovation accelerates data generation, reduces repetitive tasks, and promises faster breakthroughs in drug discovery. Despite challenges like costs and security, it signals a new era of efficient, collaborative research.
Sep 13, 2025Artificial intelligence accelerates drug discovery and reduces animal testing
1. Artificial intelligence is cutting preclinical development timelines from more than three years to under two years. 2. The Food and Drug Administration now supports validated computational models as part of regulatory submissions. On September 2, Reuters reported that the United States Food and Drug Administration is promoting artificial intelligence platforms that simulate drug safety
Sep 13, 2025Robotics Breakthrough: Reinforcement Learning Scales Vision-Action Skills
By combining reinforcement learning with robotic vision, researchers have created a system that learns complex manipulation tasks with less human demonstration data and even discovers new, more efficient movement patterns beyond those it was initially taught
Sep 13, 2025Window is Everything: A Grammar for Neural Operations
The operational primitives of deep learning, primarily matrix multiplication and convolution, existas a fragmented landscape of highly specialized tools. This paper introduces the Generalized WindowedOperation (GWO), a theoretical framework that unifies these operations by decomposing them into threeorthogonal components: Path, defining operational locality; Shape, defining geometric structure andunderlying symmetry assumptions; and Weight, defining feature importance.We elevate this framework to a predictive theory grounded in two fundamental principles. First, weintroduce the Principle of Structural Alignment, which posits that optimal generalization is achievedwhen the GWO’s (P, S, W) configuration mirrors the data’s intrinsic structure. Second, we show thatthis principle is a direct...
Sep 13, 2025Embodied visuomotor representation
npj Robotics - Embodied visuomotor representation
Sep 13, 2025Lila Sciences Secures $235M Funding, Hits Unicorn Status in AI Biotech
Lila Sciences, a Cambridge-based AI-biotech startup, raised $235 million to achieve unicorn status at over $1 billion valuation, following a $200 million seed round. It develops a "scientific superintelligence" platform integrating AI and robotic labs for accelerated drug discovery and sustainable materials. This funding reflects booming investor interest in AI-driven life sciences innovation.
Sep 13, 2025California lawmakers pass AI safety bill SB 53 — but Newsom could still veto
The state's landmark safety bill sets new transparency requirements for large AI companies.
Sep 13, 2025Embodied visuomotor representation
npj Robotics - Embodied visuomotor representation
Sep 13, 2025Hacker exploits AI chatbot in cybercrime spree
Anthropic investigates alarming AI abuse case where hacker automated entire cybercrime campaign using Claude, stealing sensitive data from defense and healthcare firms.
Sep 10, 2025Defeating Nondeterminism in LLM Inference
Reproducibility is a bedrock of scientific progress. However, it’s remarkably difficult to get reproducible results out of large language models. For example, you might observe that asking ChatGPT the same question multiple times provides different results. This by itself is not surprising, since getting a result from a language model involves “sampling”, a process that converts the language model’s output into a probability distribution and probabilistically selects a token. What might be more surprising is that even when we adjust the temperature down to 0This means that the LLM always chooses the highest probability token, which is called greedy sampling....
Sep 10, 2025California, New York could become first states to enact laws aiming to prevent catastrophic AI harm
California and New York could become the first states to establish rules aiming to prevent the most advanced, large-scale artificial intelligence models — known as frontier AI models — from causing catastrophic harm involving dozens of casualties or billion-dollar damages.
Sep 9, 2025Claude can now create and use files
Claude can now create and edit Excel spreadsheets, documents, PowerPoint slide decks, and PDFs directly in Claude.ai and the desktop app.
Sep 9, 2025Anthropic Agrees to $1.5B Settlement
AI company Anthropic has confirmed it will pay $1.5 billion to settle a class action lawsuit brought by authors and book publishers.
Sep 9, 2025mmBERT: ModernBERT goes Multilingual
We’re on a journey to advance and democratize artificial intelligence through open source and open science.