Articles tagged with Machine Learning
A study found that even advanced AI models can default to intuitive answers, overlooking critical details in health care settings. Researchers tested commercially available LLMs using creative lateral thinking puzzles and modified well-known medical ethics cases, revealing blind spots that clinicians can't afford.
Researchers at UMBC developed a new way to predict 2D materials that could transform the electronics industry. Using a mix of data mining, computer modeling, and structural analysis, they predicted 83 possible new materials with desirable properties.
A new UNLV-led study uses AI to detect emerging virus variants in wastewater samples, outperforming existing methods. The algorithm can identify unique signatures for different virus variants with as few as two to five samples, significantly earlier than current methods.
Researchers found that deep neural networks exhibit absorbing phase transitions, a phenomenon observed in physical systems like forest fires. This discovery provides a unified framework describing how the signal propagates between layers of neurons, enabling prediction of trainability and generalizability.
MIT engineers developed a versatile demonstration interface that allows users to teach robots new skills in three intuitive ways: remote control, physical manipulation, or demonstration. This innovation expands the type of users and 'teachers' who interact with robots, enabling robots to learn a wider set of skills.
The review proposes a systematic AI-driven framework for OLED material design, offering theoretical guidance and practical pathways for intelligent materials development. Machine learning models are used to predict core optoelectronic properties, enabling rapid and accurate property prediction while providing structural insights that g...
CodeSteer, a smart coach developed by MIT researchers, guides large language models (LLMs) to switch between text and code generation to solve complex problems. The system boosts the accuracy of LLMs on symbolic tasks like scheduling shipments in supply chains and multiplication problems.
Researchers developed OmicsTweezer, a tool that uses machine learning and single-cell data integration to analyze human tissue. The tool can estimate cell type composition in tumors and surrounding tissues, which could help pinpoint potential therapeutic targets.
Researchers created an 'encyclopedia of sugar utilization pathways' in 263 Bifidobacterium genomes. This resource helps predict which strains thrive in different conditions, enabling personalized probiotics to match children's lifestyles and dietary needs. The study suggests a potential solution to improving outcomes for preterm infants.
Scientists use human-AI collaboration to tackle complex questions in condensed matter physics, leveraging machine learning algorithms to identify patterns in simulation data. This approach successfully models the behavior of frustrated magnets and sheds light on quantum computing and gravity.
Researchers from the Weizmann Institute of Science and Intel Labs have developed new algorithms that allow AI developers to combine the power of different AI models, speeding up large language model performance by 1.5-2.8 times. This enables faster collaboration between models, reducing processing power costs and improving overall effi...
A newly developed AI-based tool uses protein expression to create a stemness index that analyzes tumor similarity to pluripotent stem cells. The model predicts tumor aggressiveness and identifies potential targets for new therapies, offering hope for improved cancer treatment.
Two Bar-Ilan University researchers have won ERC grants to advance cancer treatments using targeted protein degradation and improve regulatory effectiveness through machine learning. The projects aim to pioneer new therapeutics and design more effective interventions based on empirical evidence.
A new study from UNC-Chapel Hill reveals repetitive flooding in North Carolina is far more common and widespread than previously recognized. Over 90,000 buildings flooded in at least one event, with 43% located outside of FEMA's designated high-risk areas.
A new study by Cornell University reveals Amazon's AI shopping assistant Rufus gives vague or incorrect responses to users writing in some English dialects like AAE. The researchers propose a framework for evaluating chatbots that can better serve users from diverse linguistic backgrounds.
Researchers found that friends tend to perceive products similarly and exhibit synchronized brain activity when viewed together, even predicting each other's purchasing intentions. As friendships deepen, this similarity strengthens.
The UK Biobank has completed a groundbreaking imaging project, scanning the brains, hearts, abdomens, blood vessels, bones and joints of 100,000 volunteers to understand how diseases develop. The project is already improving patient care in the NHS and beyond, with over 1,300 peer-reviewed scientific papers published based on the data.
Researchers developed a self-driving lab that collects at least 10 times more data than previous techniques, dramatically expediting materials discovery research while slashing costs and environmental impact. The system uses dynamic flow experiments to continuously characterize samples, capturing data every half second.
A recent study utilizes AI to optimize catalyst design and synthesis, predicting structure-property relationships and minimizing resource-intensive calculations. The research also explores the integration of automated synthesis, characterization, and optimization in closed-loop systems.
Researchers developed an AI system that enables a four-legged robot to adapt its gait to different terrain, just like animals. The robot learned to switch gaits on the fly and navigate uneven surfaces without any alterations to the system itself, overcoming previous limitations around adaptability.
The journal is seeking original research, viewpoints, and tutorials exploring the use of AI-driven documentation in clinical settings. Submissions should address practical applications, challenges, and future directions of AI-facilitated clinical documentation.
Researchers developed an AI model that accurately predicts genetic mutations from routine pathology slides, reducing the need for rapid genetic testing by over 40% in lung adenocarcinoma patients. This innovation streamlines clinical decision-making and accelerates access to targeted therapies.
A novel AI-based approach identifies a distinct physicochemical signature near the cleavage site of gamma-secretase substrates, revealing dynamic properties essential for molecular recognition. The study highlights the potential of this methodology to improve understanding of gamma-secretase's role in diseases and aid drug development.
Researchers developed a smarter QKD system using machine learning to synchronize quantum signals, achieving reliable self-correction in unstable timing. This breakthrough enables compact and cost-effective quantum communication systems for secure key sharing over long distances.
Researchers developed a way to boost LLMs' performance on challenging tasks using test-time training. This method involves updating some model parameters using new data, leading to significant improvements in accuracy, especially for tasks requiring logic and reasoning.
A recent study using machine learning and computational modeling reveals that astrocytes play a more active role in brain function than previously thought. Astrocytes subtly modulate communication between neurons during synchronous brain activity, influencing network coordination and stability.
Researchers developed Group Encoding (GE) to forecast electricity demand using On/Off device status, improving prediction accuracy and efficiency in smart energy operation. The method simplifies complex datasets while retaining key information for optimal energy management.
A new technique called WeGeFT improves large language model performance in tasks such as commonsense reasoning and code generation. By fine-tuning key parameters, researchers reduce the need for significant computational power, advancing the field of artificial intelligence.
A new study reveals that AI systems transition from relying on word positions to meaning-based understanding as they receive enough data for training. The transition occurs abruptly, similar to a phase transition in physical systems, and is driven by the amount of data available.
Researchers have launched an open-source platform to benchmark, improve and run competitions for single-cell genomics computational methods. The platform standardizes evaluations, fosters reproducibility and accelerates progress towards open challenges in this fast-moving field.
Researchers developed an AI-powered microscope system to measure soil fungi presence and quantity, providing insights into soil health and fertility. The low-cost optical microscopy with machine learning technology can be used by farmers and land managers worldwide.
A fully autonomous robotic system developed by MIT researchers can measure important material properties like photoconductivity, increasing the speed and precision of research. The system uses machine learning and robotics to analyze new semiconductors and optimize the development of more powerful solar panels.
New research suggests that blurry, color-limited vision in early childhood can contribute to the development of key brain pathways. The study proposes that such limited vision allows the brain to specialize in certain processing units, which later develop into distinct pathways for color and fine spatial detail.
The College of Engineering at Texas A&M is developing a suite of university-wide resources to integrate generative AI into course material, research, and outreach. The initiative aims to make generative AI a core part of the academic toolkit accessible to faculty across disciplines.
A team of researchers at Tohoku University's AIMR used machine learning potential to characterize Sn catalyst activity, identifying the most effective catalysts for CO2 reduction. The study provides novel insights into the behavior of Sn-based catalysts and could lead to more efficient fuel production.
Researchers developed AI-created thermal meta-emitters that can selectively emit heat at various levels, making them ideal for energy efficiency. These materials were found to be up to 20 degrees Celsius cooler than commercial paints in a four-hour exposure to direct sunlight.
A new AI model developed by Johns Hopkins University researchers accurately identifies patients at high risk for sudden cardiac death, surpassing clinical guidelines in accuracy. The model analyzes long-underused heart imaging and medical records to reveal hidden information about a patient's heart health.
The Hebrew University of Jerusalem is part of a €9.3 million, 36-month Horizon Europe project to develop trustworthy GenAI systems. The consortium aims to integrate neural and symbolic methods for human-centric design in complex systems.
Researchers at Helmholtz Munich developed an AI model called Centaur that can simulate human behavior with remarkable accuracy. It was trained on over ten million decisions from psychological experiments and makes decisions that closely resemble those of real people. The model opens new avenues for understanding human cognition and imp...
A new study in Nature Communications found that AI models exhibit a geometric property called convexity, which helps humans form and share concepts. Convexity is also linked to the performance of AI models on specific tasks.
Paulina DeVito, a Ph.D. candidate in computer science at Florida Atlantic University, has been awarded the NSF Graduate Research Fellowship to pursue research on developing large language model-based approaches for analyzing public discourse on social media. Her work aims to understand how people discuss emerging technologies and infor...
The University of Illinois team created a user-friendly process to improve enzyme performance using AI and automated robotics. By predicting sequence changes and testing variants, they increased the activity of two key industrial enzymes by up to 26 times and 90 times.
Researchers used machine learning to simulate galaxy evolution and supernova explosions, achieving speeds four times faster than supercomputers. This breakthrough enables the study of galaxy origins, including the creation of the Milky Way's elements essential for life.
The AI for Good Global Summit 2025 will showcase AI innovations delivering better healthcare and education, reducing disaster risks, ensuring water and food security, and bolstering economic resilience. The event, organized by the International Telecommunication Union (ITU), features talks from AI leaders and 100+ demos.
An interdisciplinary team at TU Wien has developed a method that allows for the exact calculation of how reliably a neural network operates within a defined input domain. This enables mathematical guarantees for the safe use of AI in sensitive applications.
Researchers discovered AI art protection tools have critical weaknesses that cannot reliably stop AI models from training on artists' work. LightShed, a new method, can detect and remove distortions, stripping away protections and rendering images usable again for generative AI model training.
Researchers found several volatile phases in the polar jet stream over the past 125 years that predate significant climate change effects. The study suggests that natural fluctuations may be driving recent erratic behavior of the jet stream, rather than climate change.
Computer scientist Zeynep Akata has developed innovative methods to combine visual, linguistic, and conceptual elements in AI to increase user trust. Her research on explainable AI aims to make image classification decisions more transparent.
A research team at Washington University in St. Louis used machine learning to analyze sleep data from pregnant participants and found that variability in sleep patterns can predict preterm birth. The study aims to provide accurate prediction of preterm birth, which is the primary cause of death among children under age 5.
A new study introduces a machine learning-based approach to improve GNSS ambiguity resolution, achieving an 83% success rate in independent testing. The method leverages multiple diagnostic metrics into a Support Vector Machine model, enhancing reliability and reducing convergence time prediction errors.
The Association for Computing Machinery (ACM) has announced a new open-access journal, TAISAP, focusing on AI security and privacy. The journal aims to develop methods for assessing the security of AI models, systems, and environments, including adversarial attacks, privacy concerns, and cybersecurity applications.
Researchers found that LLMs factor in unrelated information like typos, extra space, and colorful language when making treatment recommendations. This nonclinical information can lead to inaccurate advice for female patients, who are more likely to be erroneously advised not to seek medical care.
Scientists developed an algorithm that can accurately simulate atomic interactions on material surfaces, reducing the need for massive computing power. This breakthrough enables the analysis of complex chemical processes in just two percent of unique configurations, paving the way for improved battery performance.
A machine learning model developed by researchers at Washington University in St. Louis can predict the risk of persistent post-operative pain and provide uncertainty estimates for each prediction. The model achieves better performance than other prediction algorithms and offers a valuable tool for doctors to guide their decisions.
A research team from Tampere University and Université Marie et Louis Pasteur has demonstrated a novel way to process information using light and optical fibers. The study used femtosecond laser pulses and an optical fiber to mimic the processing of artificial intelligence, achieving accuracy of over 91% in under one picosecond.
A new study reveals that large language models exhibit 'position bias', favoring information at the beginning and end of documents or conversations. Researchers identified design choices and training data as contributing factors to this phenomenon, which can be mitigated through adjustments in model architecture and fine-tuning.
A review of AI eye imaging devices approved for patient care found significant gaps in evidence, including lack of transparency on training data and limited diversity in clinical evaluations. The study highlights the importance of rigorous, transparent evidence and data to ensure equitable and effective AI-based solutions.
Researchers Mostafa Bedewy and Ahmed Aziz Ezzat are advancing nanomanufacturing by using machine learning to control the formation of nanoparticles and grow carbon nanotubes. The team aims to reveal which nanoparticles act as seeds for growing nanotubes, a key step towards creating ideal high-density structures.
A research team developed AI technology that analyzes individual personality traits and values to generate personalized analogies, allowing people to understand others' feelings through familiar experiences. This approach significantly improved emotional understanding and empathy in participants compared to traditional methods.
Researchers have developed a new method to physically restore original paintings using digitally constructed films that can be removed if desired. The process uses a polymer film mask printed on a very thin film and aligned to an original painting, which takes around 3.5 hours from start to finish.