Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
Researchers at Cornell University developed a new method that uses machine learning to visualize nanotextures in thin-film materials. This technique overcomes the challenge of preserving the sample, allowing for dynamic study of thin films and discovery of new morphologies.
Developing a technique to create conductive polymer wire connections between electrodes enables artificial neural networks that overcome the limits of traditional computer hardware. The approach allows researchers to control and train the network using small voltage pulses.
Researchers at Nagoya University developed an AI-based technique to predict crystal orientation in polycrystalline materials, revolutionizing the industry. The method uses optical photographs and reduces measurement time from 14 hours to 1.5 hours, enabling large-area materials analysis.
Researchers created a model to forecast eucalyptus leaf health based on temperature data, explaining over 80% of observed damage. This approach has the potential to identify suitable planting regions worldwide.
A team of researchers from Kyoto University and international institutions has developed a mathematical solution to the temporal asymmetry of nonequilibrium disordered Ising networks. This breakthrough offers insights into the behavior of biological systems, machine learning, and AI tools.
Researchers develop predictive model to control gene expression with high accuracy, enabling precise modulation of gene dosage. The AI-powered tool uses deep learning to predict on- and off-target activity of RNA-targeting CRISPRs, holding promise for treating viral infections and developing new therapies.
A new study by researchers at NYU and the New York Genome Center combines deep learning with CRISPR screens to control human gene expression. The model predicts on- and off-target activity of RNA-targeting CRISPRs, enabling precise gene controls for developing new therapies.
A new study from Aarhus University has found that applying AI predictions of protein structures enhances the CRISPR technology, making the cuts in a patient's DNA more precise. This discovery may lead to better treatments for patients with genetic disorders and potentially develop cures for various genetic diseases.
The IceCube Neutrino Observatory has produced an image of the Milky Way using neutrinos, revealing it is a neutrino desert. The observation suggests the galaxy produces significantly fewer high-energy neutrinos than distant galaxies.
A soft robotics glove with integrated sensors and AI can aid patients in relearning daily tasks after neurotrauma, including playing music. The glove provides hand guidance, amplifying dexterity and motor skills.
A new machine learning model developed by University of Pittsburgh researchers uses electrocardiogram (ECG) readings to diagnose and classify heart attacks faster and more accurately than current approaches. The model improves risk assessment, helping patients receive appropriate care without delay.
A team of scientists identified the dorsal medial prefrontal cortex as a key region in predicting rumination, which is linked to depression. The study's findings suggest that dynamic connectivity between brain regions can be used to decode rumination patterns.
The IceCube Neutrino Observatory has produced an image of the Milky Way using neutrinos for the first time. The high-energy neutrinos were detected from the galactic plane, confirming what is known about our galaxy and cosmic ray sources.
Researchers developed AEGIS, the first technique to detect backdoor attacks in robust machine learning models. It improves the trustworthiness of artificial intelligence by analyzing robust models and detecting mixed input distributions for poisoned classes.
New smart pants based on fiber optic sensors can track various types of physical activities in the clinic or at home, detecting signs of distress. The sensing approach achieved 100% accuracy in classifying activities and has several advantages, including low-cost and reliability.
A new theoretical proof shows that overparametrization enhances performance in quantum machine learning, allowing for enhanced learning and classification tasks. The Los Alamos team developed a framework to predict the critical number of parameters at which a quantum machine learning model becomes overparametrized.
A research team at the University of Tsukuba has developed a secure AI technology that allows for the shared analysis of personal and identifiable data from multiple organizations. This will enhance the accuracy of AI analysis, particularly in disease prediction and educational effectiveness.
Research highlights how generative AI models encode biases and negative stereotypes in their users, particularly marginalised groups. This can lead to the mass generation and spread of nonsensical information, influencing human beliefs and perpetuating existing inequalities.
BioAutoMATED is an all-in-one AutoML platform designed for biologists, enabling easy analysis and interpretation of biological sequences. The platform uses three existing AutoML tools to generate models that can predict biological functions from sequence information.
Researchers at the University of Washington created an app called FeverPhone that uses existing phone sensors and screens to estimate whether people have fevers. The app was tested on 37 patients in an emergency department and showed accuracy comparable to consumer thermometers, with potential for early intervention in viral outbreaks.
Researchers applied machine learning to brain responses and achieved near-perfect classification accuracy for songs that may become hits. The approach, called 'neuroforecasting,' uses data from a small group of people to predict population-level effects without needing to measure the brain activity of hundreds of people.
Researchers developed Precious1GPT, a multimodal transformer-based approach for aging clock development and feature importance analysis. The model utilizes methylation and transcriptomic data to predict biological age and identify disease-related genes, providing a pathway for therapeutic drug discovery.
A new machine learning algorithm analyzed high-resolution digital images of herbarium specimens, revealing that factors other than climate have a strong effect on leaf size within a plant species. The study also demonstrates how AI can be used to transform static specimen collections and quickly document climate change effects.
Researchers developed a new method for controlling lower limb exoskeletons using deep reinforcement learning, enabling more robust and natural walking control. The system has the potential to benefit users with spinal cord injuries, multiple sclerosis, stroke, and other neurological conditions.
Generative AI raises fundamental questions about the creative process and human's role in it. Researchers highlight gaps in understanding perceptions of AI-generated content, ownership, credit, labor economics, and media ecosystem impact.
Research by Francis de Véricourt and Huseyin Gurkan found that trust in machines' decision-making ability is key to effective learning. They discovered that biased learning occurs when humans override algorithmic decisions without observing the machine's correctness, leading to incorrect usage of machines in decision making.
A new research paper challenges the idea that unlimited trials are needed to learn safe actions in unfamiliar environments. The team presents a fresh approach that ensures learning safe actions with complete confidence while managing tradeoffs between optimality and exposure to unsafe events.
A new approach to enhance artificial intelligence-powered computer vision technologies has been developed by UCLA researchers, adding physics-based awareness to data-driven techniques. This hybrid methodology aims to improve how AI-based machinery sense, interact, and respond to their environment in real time.
A new research project at University of Leicester will develop a digital 'twin' of the UK using artificial intelligence and big data to inform environmentally-friendly land use. The project aims to reduce emissions from cattle and sheep farming, which contribute around 10% of the UK's emissions.
A machine-learning study has found that individual characteristics, including age and weight, determine which drug combinations most effectively reduce COVID-19 recurrence rates. The study used real-world data from a hospital in China and identified unique treatment combinations for different demographic groups.
Researchers have developed a new AI model that can quickly screen large libraries of potential drug compounds against target proteins. The ConPLex model uses language analysis to match potential drugs with proteins without needing to calculate molecular structures, enabling fast screening of over 100 million compounds per day.
A chemist at the University of Kansas has developed a digital tool that can spot scientific text generated by ChatGPT with 99% accuracy. The tool, which was published in Cell Reports Physical Science, uses human insight and intuition to identify key differences between human-written and AI-generated texts.
MethaneMapper is an artificial intelligence-powered hyperspectral imaging tool that can detect real-time methane emissions and trace them to their sources. With a performance accuracy of 91%, it has the potential to revolutionize the way we monitor oil and gas operations and curb climate change.
A group of educators is urging caution when using artificial intelligence (AI) in special education, highlighting its potential to benefit students with disabilities. The authors emphasize the need for careful consideration of AI's uses and limitations, including information literacy, consent, and critical thinking.
A team of researchers developed a tool to identify AI-generated academic science writing with high accuracy, using characteristics such as predictability, paragraph structure and vocabulary. The model outperformed existing AI text detectors and has potential applications in assessing student essays.
Researchers at NYU Grossman School of Medicine have developed an AI tool called NYUTron that can accurately estimate patients' risk of death, length of hospital stay, and other factors important to care. The tool achieved impressive results in predicting readmission rates, improving upon standard methods by up to 7%.
Researchers developed a new AI framework that significantly improves the ability to analyze team communication, enabling adaptive training technologies to facilitate effective team collaboration. The framework performed substantially better than previous AI technologies in classifying dialogue and following information flow.
Researchers at Mount Sinai have developed an AI model called HeartBEiT that can analyze electrocardiograms as language, enabling more accurate diagnoses. The model outperformed established methods in comparison tests and demonstrated improved performance with lower sample sizes.
Researchers developed a method using artificial intelligence to track changes in brain synapses, enabling better understanding of how connections change with learning, aging, injury and disease. Machine learning was leveraged to enhance image quality, allowing for detection and tracking of individual synapses.
A new framework developed by Carnegie Mellon University researchers significantly reduces consumers' privacy risk while preserving advertisers' utility in mobile location data analysis. The framework uses machine learning to quantify personalized privacy risks and performs personalized data obfuscation.
Optical memristors have the potential to transform high-bandwidth neuromorphic computing, machine learning hardware, and artificial intelligence. However, scalability is a significant challenge that needs to be addressed to unlock their full potential.
A new study by Carnegie Mellon University researchers finds that crime risk assessment tools, such as machine learning models, are prone to cohort bias when ignoring social change dynamics. This bias can generate inequality in the criminal justice system, distinct from racial bias.
A recent study has found that microbes play a crucial role in storing carbon in the soil, with a four-fold greater importance than other processes. This breakthrough could lead to improved soil health and increased food security through targeted farm management practices.
Researchers trained a robotic chef to watch and learn from cooking videos, enabling it to identify ingredients and actions. The robot recognized 93% of the correct recipe from 16 videos, including variations and new recipes, showcasing its potential for automated food production and cost-effective deployment.
Researchers updated their protein localization prediction model, MULocDeep, to provide more targeted predictions for biological discoveries. The tool helps researchers design more effective experiments and advance scientific discoveries related to drug development and treating diseases like epilepsy.
Researchers at North Carolina State University have developed a new methodology called Patch-to-Cluster attention (PaCa) that addresses the challenges of vision transformers. PaCa improves ViT's ability to identify, classify, and segment objects in images while reducing computational demands and enhancing model interpretability.
Researchers develop AI-based method to quantify cracking patterns in reinforced concrete structures, enabling more accurate and efficient assessments of structural damage. The approach uses graph theory and machine learning algorithms to create a unique 'fingerprint' for each set of cracks, allowing for quick and consistent evaluations.
Researchers developed an AI system, Geneformer, to predict how disruptions in human gene connections cause disease. The model, trained on data from thousands of genes, can identify potential drug targets for diseases like heart disease and cancer.
A new study published in The Neuroradiology Journal introduces an artificial intelligence computer program that can accurately identify changes in brain structure resulting from repeated head injury. This AI tool uses machine learning to process magnetic resonance imaging (MRI) scans and distinguish between the brains of male athletes ...
Researchers are exploring the use of large-scale pre-trained vision-language models (PT-VLM) to develop a new methodological framework for harnessing their power. The project aims to identify basic skills required by VisualQA and design methods to augment pre-trained models with additional skills, such as object recognition and spatial...
Researchers identified five subtypes of heart failure using machine learning, including early onset and atrial fibrillation related. These subtypes have different mortality risks, with some patients at higher risk of dying within a year after diagnosis.
Researchers used AI to identify a compound that kills Acinetobacter baumannii, a bacterium responsible for many drug-resistant infections. The new antibiotic shows promise in combating this growing public health threat.
Scientists at McMaster University and MIT have used AI to discover a new antibiotic targeting Acinetobacter baumannii, a deadly drug-resistant pathogen. The new antibiotic, abaucin, targets only A. baumannii, reducing the risk of rapid resistance development.
Researchers found that formalin fixation does not significantly alter the polarimetric properties of brain tissue, making it suitable for training machine-learning models. The study suggests that formalin-fixed brain tissue specimens can provide high-quality data for rapid and accurate diagnostic imaging in surgery.
Researchers developed a mobile application to detect Alzheimer's and mild cognitive impairment from speech data. The app achieved high accuracy rates, demonstrating its potential as an early detection tool.
Researchers at Carnegie Mellon University argue against granting rights to robots, instead suggesting a Confucian approach of assigning roles to promote teamwork and harmony. This alternative perspective recognizes the moral status of robots as entities capable of participating in rites and contributing to society.
Researchers at the Beckman Institute for Advanced Science and Technology have developed a new framework for super-resolution ultrasound using deep learning, reducing processing speeds from minutes to seconds. The new technology enables real-time blood flow visualization, overcoming challenges faced by conventional methods.
Researchers create an AI-based approach to predict precipitation intensity and variability, addressing the missing piece of cloud organization in traditional climate models. The new algorithm improves precipitation predictions, including extreme events, and enables better projections of future changes in the water cycle.
A study from the University of Chicago uses machine learning to record intricate tongue movements and neural activity, revealing that brain patterns can accurately predict 3D tongue shape. This breakthrough could lead to brain-computer interface-based prosthetics for restoring lost functions of feeding and speech.
Researchers used AI models to predict effective peptide sequences for safe drug delivery in eye cells, promising new treatments for glaucoma and macular degeneration. The model accurately predicted a peptide sequence that bound to melanin, releasing medications over several weeks.