News/Research
Scientists map quantum landscapes that could unlock room-temperature superconductors
Researchers are now mapping the hidden quantum landscape inside materials, a breakthrough that could lead to revolutionary discoveries comparable to the transistor's development in the late 1920s. This quantum topography may determine material properties and could unlock room-temperature superconductors or entirely new classes of materials with transformative applications. The big picture: Despite our increasingly digital world, materials science remains fundamental to modern life, from steel in construction to lithium in batteries, and this new quantum understanding could dramatically expand what's possible. Why this matters: The last major breakthrough in materials science—understanding electron energy bands in the 1920s—led directly to the...
read Sep 9, 2025AI reveals African Great Migration has only 533K wildebeest, not 1.3M
A new AI-powered analysis of satellite imagery has revealed that East Africa's Great Migration involves far fewer wildebeest than previously estimated, with researchers counting only 324,000-533,000 animals instead of the widely cited 1.3 million. This finding challenges decades-old population estimates and demonstrates how artificial intelligence can provide more accurate wildlife census data crucial for conservation efforts. What you should know: Princeton University researchers used deep-learning models to analyze thousands of square kilometers of satellite imagery, producing the first comprehensive AI-based count of wildebeest in the Serengeti-Mara ecosystem. Two AI models—U-Net and YOLOv8—were trained on a dataset containing 70,417 manually labeled...
read Sep 9, 2025Heads up! NASA’s AI gives satellite operators 24-hour warning for solar storms
NASA's Frontier Development Lab has successfully used artificial intelligence to predict disruptive solar events up to 24 hours in advance, working with commercial partner KX Systems to adapt financial market analytics software for space weather forecasting. This breakthrough could help satellite operators prepare for solar activity that causes both stunning auroras and potentially damaging disruptions to GPS and other space-based infrastructure. How it works: The collaboration leveraged KX Systems' kdb+, a data analytics software typically used to track rapid financial market shifts, to analyze space weather patterns. Researchers imported multiple datasets monitoring the ionosphere (the upper layer of Earth's atmosphere),...
read Sep 8, 2025Is an assistant making you insecure? Study finds AI coding creates 10x more security vulnerabilities
Programmers using AI-powered coding assistants create 10 times more security vulnerabilities than developers who code without AI assistance, according to new research from Apiiro, a security firm. The findings reveal a critical trade-off: while AI helps developers produce code faster and with fewer syntax errors, it's simultaneously introducing far more dangerous security flaws that could expose systems to cyberattacks. What you should know: The research analyzed code from thousands of developers across tens of thousands of repositories, revealing that AI-assisted programmers produce three to four times more code overall. Syntax errors dropped 76% and logic bugs decreased 60% when developers...
read Sep 8, 2025RoboBallet AI makes manufacturing elegant, coordinating 8 robotic arms in seconds
UCL, Google DeepMind, and Intrinsic have developed RoboBallet, an AI system that coordinates multiple robotic arms in manufacturing environments, solving up to 40 tasks with eight robotic arms in seconds compared to traditional programming methods that require hours of human work. The system uses reinforcement learning and graph neural networks to choreograph robot movements with unprecedented precision, potentially transforming factory efficiency and scalability across industries from automotive to electronics assembly. How it works: RoboBallet employs AI to learn optimal coordination patterns through trial and error, replacing labor-intensive manual programming. The system uses reinforcement learning (AI that learns through trial and...
read Sep 8, 2025AI erupts with revelation of 54,000 hidden earthquakes in Italy’s volcanic region
Researchers at Stanford University have used artificial intelligence to uncover four times as many earthquakes in Italy's Campi Flegrei volcanic region than previously detected, revealing critical seismic data for an area home to 500,000 people. The AI-powered analysis identified over 54,000 earthquakes from January 2022 to March 2024—compared to the 12,000 previously tracked—and revealed two major fault lines converging beneath the town of Pozzuoli, suggesting the potential for magnitude 5 earthquakes. Why this matters: The enhanced earthquake detection capabilities could fundamentally change how officials prepare for and respond to seismic threats in one of Europe's most densely populated volcanic regions....
read Sep 4, 2025NSF launches national AI research operations center to scale pilot
The National Science Foundation has launched a solicitation for proposals to establish the National Artificial Intelligence Research Resource Operations Center (NAIRR-OC), a community-based initiative designed to create a comprehensive management framework for national AI research infrastructure. This represents a critical step in transitioning the successful NAIRR Pilot program into a sustainable, scalable national resource that could significantly accelerate AI research capabilities across the United States. What you should know: The NAIRR-OC will serve as the organizational backbone for America's national AI research infrastructure, building on the foundation established by the current pilot program.• The center will develop overarching frameworks, management...
read Sep 4, 2025African researchers release 9,000-hour speech dataset for AI in 18 languages
African researchers have released what's believed to be the largest known dataset of African languages for AI development, capturing 9,000 hours of speech across 18 languages from Kenya, Nigeria, and South Africa. This $2.2 million Gates Foundation-funded initiative addresses a critical gap in AI accessibility, as most current AI tools like ChatGPT are trained primarily on English and other European languages, leaving millions of Africans excluded from the AI revolution. Why this matters: With Africa home to over a quarter of the world's languages—more than 2,000 in total—the lack of African language representation in AI creates barriers to essential services...
read Sep 4, 2025Google DeepMind AI boosts gravitational wave detection by 100x
Google DeepMind has developed Deep Loop Shaping, a novel AI method that reduces noise in gravitational wave observatories by 30 to 100 times, significantly improving their ability to detect cosmic events. Published in Science and tested at LIGO's Louisiana facility, this breakthrough could help astronomers detect hundreds more black hole mergers and neutron star collisions annually, potentially revealing intermediate-mass black holes that represent the "missing link" in understanding galaxy evolution. Why this matters: LIGO's extraordinary sensitivity—measuring distances 1/10,000th the size of a proton—makes it vulnerable to any vibration, even ocean waves 100 miles away, limiting its ability to peer deeper...
read Sep 3, 2025MIT’s FlowER AI predicts chemical reactions while conserving physics
MIT researchers have developed FlowER (Flow matching for Electron Redistribution), a new generative AI system that predicts chemical reactions while maintaining fundamental physical constraints like conservation of mass and electrons. The breakthrough addresses a critical limitation in existing AI models that often violate basic scientific principles by creating or destroying atoms during reaction predictions, potentially revolutionizing drug discovery and materials science. The big picture: Previous AI attempts at reaction prediction have struggled because they ignore fundamental physics, leading to scientifically impossible outcomes where atoms appear or disappear during reactions. Large language models like ChatGPT use computational "tokens" to represent atoms,...
read Sep 3, 2025Future sport activity sees 4-legged robot ANYmal play badminton with humans
Scientists at ETH Zürich have successfully trained a four-legged robot called ANYmal to play badminton against human opponents, achieving rallies of up to 10 shots through advanced AI and reinforcement learning. The breakthrough demonstrates how quadruped robots can master complex sports requiring coordinated whole-body movement and real-time visual perception, potentially opening new applications for disaster relief and debris removal operations. What you should know: ANYmal combines sophisticated hardware with AI-powered learning to compete in dynamic sports scenarios. The robot weighs 110 pounds, stands 5 feet 3 inches tall with an attached arm, and operates 18 joints simultaneously across four legs...
read Sep 3, 2025NSF launches national AI research center to scale America’s computing power
The U.S. National Science Foundation announced a new solicitation to establish the National AI Research Resource Operations Center (NAIRR-OC), marking a pivotal transition from the successful NAIRR pilot program to a sustainable national AI research infrastructure. This initiative directly supports the White House's AI Action Plan and aims to expand access to critical AI resources for researchers and educators across the country, addressing a significant gap in computational tools needed for AI innovation and training. What you should know: The NAIRR-OC represents a strategic investment in scaling America's AI research capabilities through a community-based operational framework. The solicitation invites proposals...
read Sep 3, 2025Fed study links AI adoption to rising unemployment in tech jobs, most notably computing and math
A new Federal Reserve study suggests the U.S. may be witnessing the early stages of AI-driven job displacement, with industries that adopted artificial intelligence most heavily now showing the highest unemployment increases. The research found a 0.57 correlation coefficient between AI adoption rates and rising joblessness, particularly affecting computing and mathematics professionals where AI usage reached nearly 80% while unemployment climbed 1.2% over three years. What you should know: The St. Louis Fed research examined unemployment changes between 2022 and 2025 across different sectors, correlating them with AI exposure rates to identify potential displacement patterns. Computing and math professionals experienced...
read Sep 2, 2025Michigan political science professor Erik Trump to examine AI’s impact on reading, research skills
Erik Trump, a political science professor at Saginaw Valley State University, will deliver a lecture titled "AI Reading, Writing, and Research: What's Left Behind for Us?" on September 24 at Hoyt Public Library in Saginaw, Michigan. The presentation will examine whether artificial intelligence represents a pivotal moment for public knowledge, literacy, and libraries, addressing both contemporary concerns and long-standing cultural anxieties about AI's impact on fundamental learning processes. What you should know: The lecture will explore AI's influence on reading, writing, and research while demonstrating actual AI tools during the presentation. Trump serves as director of the Center for Excellence...
read Aug 29, 2025AI analyzed 630K paintings to decode 600 years of economic history through emotion extraction
A team of economists has used artificial intelligence to analyze over 630,000 European paintings spanning 600 years, discovering that collective shifts in artistic mood often aligned with historical moments of prosperity, hardship, or upheaval. The research demonstrates how AI can extract emotional signals from art to enhance traditional economic data, particularly for periods where standard historical records are scarce. How it works: The researchers trained AI to detect nine emotions—including sadness, fear, anger, awe, contentment, and amusement—across paintings from 1400 to 2000. The dataset was sourced from Google Arts and Culture, WikiData, and WikiArt, featuring predominantly traditional, figurative European painting...
read Aug 29, 2025Scratching failure off the list, Korean researchers develop AI to classify skin irritations with 98% accuracy
South Korean researchers have developed an AI model using YOLO v5 technology that can accurately classify skin irritation from patch tests, achieving a classification accuracy of 0.983. The breakthrough addresses longstanding challenges in dermatological diagnostics by providing consistent, objective assessments that could reduce variability between human evaluators and accelerate clinical decision-making. What you should know: The AI model represents a significant advancement in automated dermatological assessment, moving beyond traditional convolutional neural networks to object detection algorithms. Researchers from Sungkyunkwan University, a South Korean research institution, trained the model on 83,629 images collected from patch test participants between 2020-2023. The system...
read Aug 28, 2025New “R-Zero” framework lets AI models improve by teaching each other
Tencent AI Lab and Washington University researchers have developed R-Zero, a training framework that enables large language models to improve themselves without any human-labeled data. The breakthrough technique uses two AI models that challenge and teach each other through reinforcement learning, potentially eliminating one of the most expensive bottlenecks in AI development while allowing models to surpass human-defined limitations. How it works: R-Zero splits a base model into two independent roles that co-evolve through continuous interaction cycles. The "Challenger" creates new tasks at the threshold of the "Solver's" current abilities—neither too easy nor impossible to complete. The Solver is rewarded...
read Aug 28, 2025MIT’s VaxSeer AI outperformed WHO flu vaccine picks in 9 of 10 seasons
MIT researchers have developed VaxSeer, an AI system that uses machine learning to predict which influenza strains should be included in seasonal vaccines months before flu season begins. The tool aims to reduce the guesswork in vaccine selection by analyzing decades of viral sequences and lab test results to forecast virus evolution and vaccine effectiveness. What you should know: VaxSeer combines two prediction engines to forecast both viral dominance and vaccine effectiveness against future flu strains. The system estimates how likely each viral strain is to spread using a protein language model, then determines dominance by accounting for competition among...
read Aug 28, 2025HEALTH AI Act targets healthcare’s administrative burden with federal funding
Representative Ted Lieu has introduced groundbreaking federal legislation that could reshape how artificial intelligence transforms American healthcare. The Healthcare Enhancement And Learning Through Harnessing Artificial Intelligence Act, known as the HEALTH AI Act, establishes the first dedicated federal grant program specifically designed to fund research into generative AI applications across the healthcare sector. This legislative initiative arrives at a critical moment when healthcare providers face mounting administrative burdens, widespread clinician burnout, and persistent disparities in patient care quality. The bill, designated as H.R. 5045 and introduced on August 26, 2025, currently has one cosponsor and awaits consideration by the House...
read Aug 27, 2025Chatbots are training us, too: Study finds ChatGPT’s AI buzzwords doubled in spoken English
Florida State University researchers have discovered that AI buzzwords commonly overused by ChatGPT are now appearing in everyday spoken English, marking the first peer-reviewed evidence that large language models may be directly influencing human speech patterns. The study, which analyzed 22.1 million words from unscripted conversations, found that nearly three-quarters of AI-associated words showed increased usage after ChatGPT's 2022 release, with some more than doubling in frequency. The research breakthrough: FSU's interdisciplinary team conducted the first academic study to examine whether chat-based AI is changing how humans naturally speak, not just write. The study will be published in AIES Proceedings...
read Aug 26, 2025Better old than young? In tech?! AI automation hits 22–25 year-old workers 12x harder
New research from Stanford economists reveals that AI automation is disproportionately impacting younger workers, with employment among 22-25 year-olds declining by 12 percentage points in AI-exposed fields since 2022. The findings suggest AI's tendency to replace "book-learning" over experience-based knowledge is fundamentally reshaping entry-level job markets, particularly in software development where roles for young workers dropped nearly 20% in 2025 compared to their 2022 peak. What you should know: The research tracked millions of payroll records through July 2025, providing real-time data on how AI deployment affects different age groups in the workforce. Workers ages 22 to 25 experienced the...
read Aug 25, 2025AI turns Rosetta Stone for animal languages as it helps decode whale, dolphin and bird communication
Scientists are on the verge of deciphering animal communication using artificial intelligence, with recent breakthroughs suggesting we may soon be able to "talk" with other species for the first time. The development represents a fundamental shift from decades of scientific reluctance to acknowledge non-human language, driven by AI's ability to detect patterns in massive datasets of animal sounds and behaviors. The big picture: Multiple research teams are racing to crack interspecies communication, spurred by the new Coller Dolittle Challenge offering $100,000 annually and a $10 million grand prize for breakthrough discoveries. The challenge, established by Tel Aviv University and Jeremy...
read Aug 22, 2025Student’s AI model accidentally reconstructs real 1834 London protests through adjacent historical data
A computer science student at Muhlenberg College accidentally discovered his AI model trained on Victorian-era texts could accurately reference real historical events from 1834 London, including protests related to Lord Palmerston's actions. Hayk Grigorian's TimeCapsuleLLM reconstructed these historical connections from scattered references across thousands of documents without being explicitly taught about these specific events, demonstrating how AI models can synthesize factual information from ambient patterns in training data. What you should know: Grigorian has been developing TimeCapsuleLLM over the past month, training it exclusively on texts from 1800-1875 London to capture an authentic Victorian voice. When prompted with "It was...
read Aug 22, 202575% prefer AI chatbots for more open-ended polling, over traditional surveys
OpenResearch, the OpenAI-funded research nonprofit, has successfully tested AI chatbots as polling assistants in its ongoing unconditional cash transfer study, with more than three-quarters of respondents choosing to engage with the bot-assisted survey format. The breakthrough could transform the polling industry by enabling researchers to conduct qualitative research at scale while gathering richer, more nuanced data than traditional multiple-choice surveys allow. What you should know: The AI-assisted polling approach produced significantly more engaged respondents and comprehensive data than traditional survey methods. Participants who chose the chatbot option spent a median of 16 minutes on the survey, offering detailed responses that...
read