All AI news
Browse, filter, and search every article in the archive. The homepage shows the last 24 hours; everything older lives here.
The skills that earn top grades are the ones AI can fake best
Researchers find that AI can easily replicate skills that lead to top grades, like writing and problem-solving. This raises concerns about the authenticity of student work and the effectiveness of traditional grading methods.
LAION drops massive open video dataset with 10 million hours of footage for AI research
LAION just released a massive open video dataset containing 10 million hours of footage for AI research. This resource will significantly enhance training and development for video understanding models in the AI community.
AI benchmarks have a trust problem and Google wants to fix it
Google is addressing trust issues in AI benchmarks by developing new evaluation methods. This shift aims to improve the reliability of AI performance assessments for developers and users alike.
The Download: a secretive antiaging drug and joining virtual power plants
Researchers are developing a secretive anti-aging drug that shows promise in extending lifespan. This could revolutionize how we approach aging and longevity treatments.
Aug 28, 2026AlignmentAutomated researchers can reliably mitigate alignment failures
Anthropic just developed automated researchers that can effectively address alignment failures in AI systems. This advancement helps ensure AI behaves in ways that align with human values and intentions.
The Humanoids at China’s Robot Games Were Faster Than Usain Bolt—but I’m More Impressed by Their Tweezer Mastery
Humanoid robots at China's Robot Games outpaced Usain Bolt in speed tests. Their impressive dexterity with tweezers showcases advancements in robotic manipulation skills.
How to evaluate LLMs before production
GitHub just released a guide on evaluating large language models (LLMs) before deploying them. This helps developers ensure their models meet production standards and perform reliably in real-world applications.
Addressing a sticking point in sustainable adhesives
Researchers at MIT are developing a new sustainable adhesive that uses plant-based materials. This innovation could reduce reliance on harmful chemicals in manufacturing processes.
Amping up T cells to target cancer
Researchers are enhancing T cells to better target and destroy cancer cells. This breakthrough could lead to more effective immunotherapies for patients.
Taking your temperature from the inside
MIT researchers developed a new AI system that can measure internal body temperature using a small ingestible sensor. This could lead to more accurate health monitoring without the need for traditional thermometers.
Seeing through murky waters
Researchers developed an AI model that can analyze underwater images to identify and classify marine life. This tool enhances our understanding of ocean ecosystems and aids in conservation efforts.
Your brain on AI
MIT researchers are studying how AI affects human cognition and decision-making. Their findings could reshape how we integrate AI into daily tasks and improve user interactions with technology.
AI could make scientists do more work less well, not less work better, study argues
A new study suggests AI might lead scientists to do more work but with lower quality. This raises concerns about the effectiveness of AI tools in enhancing scientific productivity.
World models that ignore human beliefs predict the wrong actions, new research shows
Researchers found that world models not considering human beliefs lead to incorrect action predictions. This means AI systems might struggle with real-world tasks if they don't account for human perspectives.
Psychological methods reveal major weaknesses in AI security testing
Researchers are uncovering significant flaws in AI security testing using psychological methods. This means current testing approaches may not effectively identify vulnerabilities, raising concerns about AI safety.
Simulation: the new Scaling Law — Joon Sung Park, Simile AI
Simile AI just introduced a new scaling law for simulations. This approach enhances the efficiency of training AI models, potentially reducing costs and time for developers.
From Atari to EVE Online: Building on 15 Years of AI Research in Games
Google DeepMind is showcasing 15 years of AI research applied to gaming, from classic Atari games to complex environments like EVE Online. This evolution enhances how AI can interact and adapt in dynamic game worlds, pushing the boundaries of player experiences.
Mother tongue
Researchers at MIT are developing an AI that can understand and generate text in multiple languages, focusing on low-resource languages. This advancement could enhance communication and accessibility for speakers of less common languages worldwide.
When AI designs a drug, who gets the credit?
Researchers are debating who gets credit when AI designs a new drug. This raises questions about authorship and intellectual property in the rapidly evolving field of AI-assisted drug discovery.
Measuring benchmark optimization in speech recognition
Hugging Face is optimizing benchmarks for speech recognition systems. This improvement aims to enhance the accuracy and efficiency of voice-based AI applications.