AI Research

Twitter

Facebook

Copy Link

Key Facts

FocusScaling, Reasoning, Safety, Multimodality

Key OrgsGoogle DeepMind, Stanford HAI, MILA, MIT CSAIL, Allen AI

Key EventsNeurIPS, ICML, ICLR

Top News

MIT researchers use AI and ultrasound wristband to turn hand gestures into robot training data

MIT researchers developed an ultrasound wristband that captures muscle and tendon movements beneath the skin to train robots in dexterous tasks. The device tracks all 22 degrees of freedom of the human hand and mirrors gestures within 120 milliseconds, potentially enabling humanoid robots to master complex tasks like housework and surgery without human guidance.

Technology5 sources

MIT researchers use AI and ultrasound wristband to turn hand gestures into robot training data

Technology5 sources

Google NotebookLM upgrades to Gemini 3.5 and Antigravity with new agentic research capabilities

Google is rolling out major updates to NotebookLM, its AI-powered note-taking app that launched in 2023. The research tool now runs on Gemini 3.5 and the Antigravity platform, gaining agentic capabilities that enable it to write and run code, find sources through Google Search, and generate outputs in multiple file formats including PDF, Excel, and PowerPoint.

Technology17 sources

Google NotebookLM upgrades to Gemini 3.5 and Antigravity with new agentic research capabilities

Technology17 sources

AI Wildlife Detection Tool Catches Smuggled Marine Species Hidden in Airport Luggage

Australian researchers have developed an AI-powered tool that identifies smuggled marine wildlife in airport luggage with 92% accuracy. The system uses 3D X-ray scanners to detect shark fins, seahorses, and sea cucumbers—species targeted in the $20 billion illegal wildlife trade. While promising, the technology is designed to complement human inspection and biosecurity dogs, not replace them.

Science3 sources

AI Wildlife Detection Tool Catches Smuggled Marine Species Hidden in Airport Luggage

Science3 sources

New AI technology speeds eye disease diagnosis and predicts heart attack, stroke risks from retinal scans

Researchers at Washington University School of Medicine developed OCTCube-M, an AI system that analyzes 3D eye scans to detect retinal diseases more accurately than older models. The technology identified 43 to 60 additional cases per 1,000 individuals and can predict broader health risks including heart attack, stroke, and kidney failure based solely on retinal imaging.

Health2 sources

New AI technology speeds eye disease diagnosis and predicts heart attack, stroke risks from retinal scans

Health2 sources

New AI tool achieves near-perfect accuracy in distinguishing Alzheimer's from Lewy body dementia

University of Florida researchers developed AIDD, an AI-powered tool that combines brain scans with machine learning to distinguish between Alzheimer's disease and dementia with Lewy bodies with near-perfect accuracy. The breakthrough addresses a critical problem: up to 50% of Lewy body dementia patients are misdiagnosed as having Alzheimer's, leading to treatments that can worsen symptoms.

Health2 sources

New AI tool achieves near-perfect accuracy in distinguishing Alzheimer's from Lewy body dementia

Health2 sources

Sanofi and Owkin deepen AI partnership with five-year K Pro license for drug development

Sanofi and Owkin have expanded their multi-year partnership to co-develop AI-driven biopharma agents through a five-year license for K Pro, Owkin's AI Scientist platform. The collaboration builds on their 90 million euro strategic partnership that began in 2021, focusing on oncology and immunology research. The new initiative aims to automate complex pharmaceutical research tasks across the drug development pipeline.

Technology2 sources

Sanofi and Owkin deepen AI partnership with five-year K Pro license for drug development

Technology2 sources

MIT uses Battleship game to teach AI models smarter problem solving and strategic thinking

Researchers at MIT and Harvard turned the classic Battleship game into an AI training ground, revealing a critical weakness in today's systems: they excel at answering questions but struggle to ask them. By teaching AI models to plan better questions, smaller models like Llama 4 Scout achieved an 82% win rate against humans—up from just 8%—while operating at a fraction of the cost of larger frontier models.

Science2 sources

MIT uses Battleship game to teach AI models smarter problem solving and strategic thinking

Science2 sources

AI-designed vaccine passes first human trial, offering broad protection against coronaviruses

Researchers at the University of Cambridge successfully tested the world's first AI-designed vaccine in humans. The trial involved 39 volunteers and targeted multiple coronaviruses, including COVID-19 and SARS. The vaccine uses an AI-designed super-antigen that could provide lasting protection against entire virus families, even as they mutate, potentially transforming how we prepare for future pandemics.

Health11 sources

AI-designed vaccine passes first human trial, offering broad protection against coronaviruses

Health11 sources

US and Japan launch $1 billion AI for science partnership through Genesis Mission

The United States and Japan have announced a five-year, $1 billion partnership to advance scientific technologies using AI. Each country will invest $500 million in the Genesis Mission initiative, focusing on quantum technologies, fusion energy, and biotechnology. The collaboration aims to accelerate research timelines and maintain technological advantage over China.

Policy2 sources

US and Japan launch $1 billion AI for science partnership through Genesis Mission

Policy2 sources

AI epilepsy detection identifies early warning signs in EEG data without visible seizures

University of Delaware researchers developed an AI-powered method to detect early warning signs of epilepsy in baseline brain activity, eliminating the need to capture seizures during routine EEG recordings. The machine learning algorithm successfully identified genetic epilepsy markers in mouse models with high accuracy and is now advancing to clinical trials with children at Nemours Children's Health.

Health2 sources

AI epilepsy detection identifies early warning signs in EEG data without visible seizures

Health2 sources

Airbnb CEO Brian Chesky to launch AI lab focused on design, entering competition with OpenAI

Brian Chesky is backing a new artificial intelligence lab focused on user interaction and design, marking his entry into AI development. The Airbnb CEO, who helped Sam Altman return to OpenAI in 2023, will remain at Airbnb while funding the venture. The move signals frustration with existing LLM products and a belief that travel and e-commerce need richer visual interfaces than text-based chatbots.

Startups5 sources

Airbnb CEO Brian Chesky to launch AI lab focused on design, entering competition with OpenAI

Startups5 sources

Wildfires reverse decades of US progress on ozone pollution, causing 318 deaths annually

A new study reveals that wildfires are undoing decades of air quality improvements in the United States. After national smog levels dropped 11% from 2003 to 2015, wildfire smoke has reversed that trend, increasing ground-level ozone by 4% since 2015. Researchers used artificial intelligence to map ozone pollution nationwide and found that wildfire emissions are now causing approximately 318 additional premature deaths per year.

Science3 sources

Wildfires reverse decades of US progress on ozone pollution, causing 318 deaths annually

Science3 sources

Anthropic warns AI may soon build itself, calls for global pause on frontier development

Anthropic is urging the world's leading AI companies to consider pausing frontier AI development as its Claude chatbot now writes over 80% of the code merged into its systems. The company warns that AI systems may be approaching recursive self-improvement, where they can design their own successors with minimal human input, potentially increasing the risk of humans losing control of the technology.

Policy35 sources

Anthropic warns AI may soon build itself, calls for global pause on frontier development

Policy35 sources

Consistency in training data helps robots master complex skills better than varied examples

Researchers from NYU Tandon and the Robotics and AI Institute discovered that robots learn dexterous manipulation tasks more effectively from consistent, structured demonstrations rather than highly variable training examples. The breakthrough approach achieved 90% success in real-world trials and could reshape how robots acquire complex physical skills.

Science2 sources

Consistency in training data helps robots master complex skills better than varied examples

Science2 sources

BioCoach AI delivers real-time biomechanical feedback to prevent workout injuries at home

Researchers from Drexel University and Michigan State University developed BioCoach AI, an AI-powered tool that analyzes exercise form through smartphone cameras and provides anatomy-specific corrections in real time. The system uses 3D skeletal reconstruction and biomechanical modeling to deliver personalized feedback, addressing the 48% spike in at-home exercise injuries during the pandemic.

Technology2 sources

BioCoach AI delivers real-time biomechanical feedback to prevent workout injuries at home

Technology2 sources

Prev 1 234…31 Next