A new study reveals that 81% of O’ahu's coastline could experience erosion by 2100, with a further 40% loss happening by 2030. The research used computer models incorporating satellite imagery to predict the seasonal movement of sand, resulting in more severe erosion projections than previous studies.
A Lehigh University researcher is developing a new platform to enhance flu forecasting through a human judgment temporal forecast system, which aims to improve evidence-based public health decision making. The system will collect unbiased opinions and forecasts from public health experts and the general public.
Dr. Wei Li is creating a virtual lunar welding platform to simulate welding in the moon's harsh environment, addressing temperature fluctuations and extreme vacuum conditions. The project aims to enable reliable large structure assembly on the moon, a crucial step for human colonization.
Researchers created a promising approach that augments LLMs with graph-based models to generate molecules with desired properties. The method generates molecules with better matching user specifications and valid synthesis plans, outperforming existing LLM-based approaches.
Researchers at Florida Atlantic University developed an innovative real-time ASL interpretation system using AI and deep learning. The system achieved a 98.2% accuracy rate with minimal latency, enabling fast and reliable performance for applications such as live video processing and interactive technologies.
Researchers at Concordia University have developed a new approach to identifying fake news on social media using the SmoothDetector model. The model integrates probabilistic algorithms with deep neural networks to capture uncertainties and patterns in multimodal data, providing more nuanced judgments of authenticity.
A recent study from UTSA researchers reveals that large language models (LLMs) can pose a serious threat to programmers who use them to help write code. The study found that up to 97% of software developers incorporate generative AI into their workflow, and 30% of code written today is AI-generated.
A new artificial intelligence-based method detects genetic markers of antibiotic resistance in bacteria, potentially leading to faster and more effective treatments. The method, called Group Association Model, uses machine learning to identify key mutations linked to drug resistance, reducing false positives and misdiagnoses.
A new framework helps clinicians choose phrases that more accurately reflect the likelihood of certain pathologies in X-rays, improving the reliability of critical clinical information. The approach treats certainty phrases as probability distributions, capturing nuances of what each word means.
Researchers developed a machine learning-powered fluid simulation model that significantly reduces computation time without compromising accuracy. The new surrogate model maintains the same level of accuracy as traditional particle-based simulations while reducing computation time from approximately 45 minutes to just three minutes.
A Cornell University-led study predicts long-term increases in precipitation over East Asia and the Western U.S. as the Southern Ocean warms, regardless of climate mitigation efforts. The research suggests that accounting for cloud feedbacks in climate models can help explain uncertainties and improve predictions.
A team of NYU computer scientists has developed a new blockchain system called Bounce that leverages satellites to ensure security, reduce energy consumption, and achieve low response times. Bounce processes over five million transactions per second, outperforming its nearest competitor by 30-100 times.
A new computer modeling tool suggests that Bronze Age people may have traveled directly over the open ocean between Denmark and Norway. The simulations indicate that such trips were possible, but required a boat with specific capabilities and good weather forecasting.
SourcePLOS·JournalPLOS One·TypeComputational simulation/modeling·DateApr 2, 2025
Cleveland Clinic researchers successfully tested quantum computing's ability to simulate proton affinity, a fundamental chemical process critical to life. The study used machine learning applications on quantum hardware, achieving higher accuracy than classical computing in predicting proton affinity.
Researchers have discovered that the underside of the North American continent is experiencing 'cratonic thinning', a phenomenon where the continent is slowly losing its stability and rock layers. This process, driven by the subduction of the Farallon Plate, may eventually stop as the plate sinks deeper into the mantle.
The Flatiron Institute's Center for Computational Astrophysics is now supporting MESA's ongoing maintenance and development after its creator Bill Paxton steps down. The center has hired Philip Mocz as a full-time software engineer to ensure MESA's continued growth and impact in stellar physics.
Researchers develop AI model to predict novel mutations in protein sequences, combining grammatical and semantic changes. The method uses all available information about the sequence and mutations to create a more accurate prediction model.
Crop mapping uses satellite imagery to create accurate crop type maps in various regions. The research team trained machines to recognize crops from satellite images, achieving high accuracy rates. However, models pre-trained on general image datasets performed better than those pre-trained on satellite images.
University of Michigan researchers developed a statistical method to give a more complete sense of human ancestry. The Gaia algorithm estimates an individual's genetic ancestors, identifies their average location based on movement assumptions, and tracks it back over centuries.
Researchers developed a new method to include uncertainty in predictive algorithms, ensuring accurate and reliable solutions. The approach uses Markov models to explicitly include uncertainty in specific parameters, allowing for faster predictions and more complete analysis.
A team of engineers at the University of Texas at Dallas has developed a new surface design that collects and removes condensates rapidly, challenging conventional theory. The discovery reveals a limitation in existing heat transfer models and inspires a new theory to explain the phenomenon.
Researchers created a memristor-based brain-computer interface decoding system that achieves high accuracy while consuming significantly less energy than traditional systems. The system demonstrated impressive capabilities in real-world testing, including improved stability and accuracy through co-evolution with brain signals.
A team of researchers used computer code M3D-C1 to model different valve configurations and found that six gas valves provide optimal protection for rapidly dispersing cooling gas. The study's findings will help bring fusion power closer to reality by advancing disruption mitigation strategies.
A new approach to AI developed by Texas A&M University engineers mimics the human brain's neural processes, integrating learning and memory in a single system. This 'Super-Turing AI' has the potential to revolutionize the industry by reducing energy consumption and environmental impact.
Researchers found that placing naloxone kits in transit stations could improve availability and save more lives. By optimizing distribution strategies using mathematical models, the team discovered that just 60 kits at 650 locations could cover over half of opioid poisonings in Vancouver.
The study systematically traces Generative AI evolution from deep learning to foundation models, highlighting four distinct stages and successful applications. Key challenges like safety concerns and theoretical breakthroughs require further attention and development in the field of Generative AI.
Researchers have developed a hybrid image-generation tool called HART that combines the strengths of autoregressive and diffusion models. It achieves high reconstruction quality with significantly reduced computational resources, enabling local execution on laptops or smartphones.
Prof. Günter Blöschl's research has made significant contributions to understanding the drivers of increasing flood risks under climate change and reveals that the last two decades have been markedly flood-prone compared to historical records.
A team led by Dr. Marcus Botacin is creating a large language model (LLM) to automatically identify malware and write rules to defend against it. The LLM will use signatures to complement human analysts' skills, identifying malware faster and more accurately.
The Open Brain Institute launches a groundbreaking platform to simulate and study digital brains, empowering researchers to explore brain complexity and diseases. With its virtual neuroscience laboratories, the OBI enables global collaboration and access to cutting-edge virtual labs.
Zhou is conducting GPU modeling research at George Mason University, supported by the US Department of Energy. The project aims to develop smart modeling and simulation for high-performance computing (HPC) applications.
Researchers from Virginia Tech have published a visionary paper on fusing wireless technologies and AI to create human-like common sense. The team aims to develop a network that can think, plan, and imagine like humans, enabling seamless merging of physical, virtual, and digital dimensions.
A new study finds that an open-source AI model, Llama 3.1 405B, has caught up with a leading proprietary model, GPT-4, in solving challenging clinical cases. This development could offer a valuable alternative to proprietary models, benefiting patients and clinicians alike.
Experts at Flinders University are calling for a new approach to water security on Pacific islands, where rising sea levels threaten local water supplies. Infiltration galleries, which extract fresh groundwater from shallow depths, may provide a solution, but more research is needed on their optimal design and performance.
Physicists have created a new code, QUADCOIL, to design stellarators, which could lead to simpler and more affordable fusion facilities. The code helps balance physics and engineering by quickly ruling out unstable plasma shapes and predicting magnet complexities.
A Virginia Tech study reveals that machine learning models are inadequate in detecting critical patient conditions, missing 66% of injuries. The research emphasizes the need for more responsive models incorporating medical knowledge to improve healthcare outcomes.
Anna Balazs, a world-renowned pioneer in materials theory and soft matter science, has been awarded the 2025 Gutenberg Research Award. Her recent research on 'Confine: Sculpting Confined Fluids for Transport using Self-Organization and Information Transfer' has garnered significant attention.
A team of computer engineers developed PhyloFrame, a machine-learning tool that accounts for ancestral diversity in genetic data. This improves precision medicine outcomes by predicting differences between subtypes of diseases and suggesting personalized treatments regardless of patient ancestry.
Researchers have developed a novel technique to overcome the spurious correlations problem in AI by eliminating a small portion of the training data that contains hard-to-understand features. This approach improves performance even when conventional techniques are ineffective.
Scientists from the University of California San Diego have discovered that liquid water separates into two distinct phases under certain conditions, one high-density and one low-density. This finding reveals a unique property of water and has potential applications in fields such as water desalination and pollutant capture.
Researchers have employed artificial intelligence and data analysis to improve forest protection, enabling real-time monitoring and predictive analysis. The innovative forest regeneration model forecasts how forests will grow and change over time, identifying best-suited tree species for different environments.
Researchers at UC San Francisco have enabled a paralyzed man to control a robotic arm through a device that relays signals from his brain to a computer. The device, known as a brain-computer interface (BCI), worked for a record 7 months without needing to be adjusted.
A new machine learning algorithm can fully characterize systems of merging neutron stars in under a second, compared to traditional methods which take around an hour. This allows for rapid localization of the source and pointing of telescopes towards the merging neutron stars.
The new AI model leverages hypergraphs to quickly and accurately identify therapeutic gene targets for diseases. HIT outperformed existing models in all tested metrics, demonstrating its accuracy in classifying therapeutic gene targets with great precision.
The Morpho software can model a range of soft materials, including biological tissues and shape-shifting fluids, with high accuracy. Researchers and engineers can solve complex optimization problems using the open-source program.
Researchers have developed a new AI tool that uses sensors and real-time data to predict water quality across the US. This tool can be applied nationwide, benefiting communities by providing water quality forecasts, streamlining operations, and informing strategies for managing turbidity in basins worldwide.
Researchers from Carnegie Mellon University and UC-Riverside developed a system to monitor and analyze beehive health using sensors and forecasting models. The Electronic Bee-Veterinarian (EBV) uses real-time temperature data to calculate the hive health factor, helping beekeepers take corrective actions before colony collapse.
Researchers from Cornell and Tel Aviv University developed ProtoSnap to copy cuneiform characters from photos of tablets. The approach allows for accurate reproduction of characters and whole tablets, making it easier for experts to analyze and compare ancient texts.
TaxaBind, a new multimodal AI tool, addresses ecological problems by combining multiple models for species classification, distribution mapping, and other tasks. The tool outperformed state-of-the-art methods in zero-shot classification and cross-modal retrieval tasks.
Dr. Sarah Du, an associate professor at Florida Atlantic University's College of Engineering and Computer Science, has been selected as a Senior Member of the National Academy of Inventors for her significant contributions to advancing medical technology. Her research focuses on developing point-of-care diagnostic tools and monitoring ...
A recent study by Washington University in St. Louis explores the potential of large language models to predict postoperative complications from preoperative assessments and clinical notes. The work shows that specialized LLMs can significantly outperform traditional machine learning methods in forecasting postoperative risks.
A study developed by the US Department of Energy's Thomas Jefferson National Accelerator Facility aims to lower data center costs using machine learning. The Digital Data Center Twin (DIDACT) system detects anomalies and diagnoses their source using AI continual learning, reducing downtime for scientists processing data from experiments.
A new collaboration between ASU, MSU, and LLNL aims to answer the question of how planets form by analyzing the atmospheres of seven young exoplanets. The team will use the James Webb Space Telescope and powerful computers to create atmospheric models that can reveal insights into planetary formation and evolution.
A new study developed a computer model to predict household displacement duration in US communities after disasters, accounting for socioeconomic factors. The model combines physical damage estimates with socioeconomic characteristics to help inform risk mitigation strategies.
Researchers at Mizzou have developed a proactive approach using artificial intelligence to address evolving threats against smart grids. The CIBR-Fort system can predict cyberattacks with 91.88% accuracy and defend against them in real-time, enabling scalable security for power grids of the future.
Researchers developed a new method to search through billions of molecules to identify potential anti-inflammatory drug candidates. The method uses computer algorithms to explore vast chemical space and has the potential to speed up the costly drug development process.
An international research team developed a user-friendly software method called Segment Anything for Microscopy, which can precisely segment images of tissues, cells, and similar structures. The new model improved performance for cell segmentation, enabling researchers to automate tasks that previously took weeks of manual effort.
A Cornell University study used a computer model to predict the transmission of bird flu to backyard poultry when migratory mallards stop to rest. The researchers found that mallard ducks are natural carriers of avian influenza and can spread the disease to backyard poultry if they stop to rest in an area with high risk for introduction.
Researchers developed innovative techniques and optical solutions to overcome long-standing challenges in facial projection mapping. They achieved unprecedented speed in dynamic facial projection mapping while maintaining high accuracy, and proposed methods to simulate video annotations and minimize alignment artifacts.
Researchers found that large language models use a semantic hub mechanism to process diverse data, integrating information from various modalities. They demonstrated that models can change their outputs by intervening with text in the dominant language, even when processing other languages.