Articles tagged with Artificial Intelligence
Researchers aim to develop a voice assistant delivering customized cognitive stimulation therapy to improve the quality of life for persons living with dementia and their care partners. The project has the potential to positively impact a large population, including those in remote areas.
Researchers developed a tool to improve data transparency in large language models, enabling practitioners to find suitable datasets for their models. The tool, Data Provenance Explorer, automatically generates summaries of dataset creators, sources, licenses, and allowable uses.
A new automated scoring system using deep learning and pyramid sampling analyzes morphological features at various spatial scales to enhance the accuracy of HER2 assessment. The system achieved a classification accuracy of 84.70% in blind testing, comparable to experienced pathologists.
The Human AugmentatioN via Dexterity (HAND) center aims to develop robots capable of enhancing human labor through engineered systems of dexterous robotic hands, AI-powered fine motor skills, and human interface. The center's goal is to make robotic assistance accessible and applicable to a wide range of physical actions.
A study by Dr. Shunichi Kasahara found that levels of identification with one's face remain consistent regardless of agency or control over facial movements. The results suggest that a sense of agency does not significantly impact our ability to judge our facial identity, even in scenarios like deepfakes.
Researchers found that large language models (LLMs) can improve systems that detect bots by outperforming state-of-the-art systems by 9%. However, LLMs can also reduce the performance of existing detectors by up to 30% when used to evade automated detection.
A new study published in Scientific Reports confirms a linear relationship between blood glucose levels and voice fundamental frequency, suggesting potential for voice-based glucose monitoring. Researchers at Klick Labs used vocal biomarkers and AI to detect Type 2 diabetes with high accuracy.
A new course launched with funding from the UK Space Agency to address the growing skills gap in software, data, and AI for the UK space sector. The Securing the future of space: Space Software and Data/AI CPD programme will equip mid-career professionals with expertise in AI and data-science.
Researchers found that students who struggle with executive functioning (EF) skills perceive generative AI chatbots as significantly more useful for schoolwork. Using these tools critically can support students with EF challenges, but responsible use is crucial to maintain academic integrity.
Researchers at the University of Illinois have developed a method to understand and improve light-harvesting molecules for solar energy applications. By combining AI with automated chemical synthesis and experimental validation, they were able to produce molecules four times more stable than traditional ones.
Researchers have made a significant breakthrough in vasculitis research using AI-powered big data techniques, enabling more precise identification of disease patterns. The study offers new insights into the diagnosis and treatment of systemic vasculitis, a group of rare autoimmune diseases.
Researchers at Boston University created an AI tool that can determine the cause of dementia using commonly collected patient data, boosting doctor accuracy by 26%. The algorithm identifies 10 types of dementia, including vascular and frontotemporal dementia, to help doctors manage patients more effectively.
Researchers developed AI tool AINU to differentiate cancer cells from normal cells and detect early stages of viral infection. The tool scans high-resolution images of cells at nanoscale resolution, enabling it to detect subtle changes in cell structure that are too small for human observers.
Researchers explore AI's role in addressing challenges in immunotherapy, developing new biomarkers for precise disease characterization and predicting treatment response. A comprehensive review applied AI/radiomics to cross-sectional imaging, showcasing the current landscape in IO treatment.
A recent study found that approximately half of FDA-approved AI medical devices are not trained on real patient data, sparking concerns about device accuracy. The researchers analyzed 500+ medical AI devices and discovered that many lacked clinical validation data, which is essential for ensuring the credibility of these technologies.
A new paper in JCOM describes the development of an automatic translation app for sign language, designed through co-creation with deaf communities. The research team used a theatrical performance and AI tools to gather feedback from audiences, informing the app's design and features.
Researchers developed Stain SAN, a novel domain adaptation technique to correct color differences in stained histopathology images. The method improves consistency and comparability of data, leading to better performance in machine-learning-based classifiers.
A new AI-based digital platform has been developed to analyze tissue sections from lung cancer patients, making diagnosis faster and more accurate. The platform uses algorithms that enable fully automated analysis of digitized tissue samples, allowing for personalized therapy based on molecularly specific genetic changes.
Researchers aim to bridge educational divide with a multifaceted approach to equip teachers and students with essential AI skills. The project will foster proficiencies in communication, leadership, teamwork, and critical thinking while addressing real-world challenges through AI.
Researchers at Washington State University developed an AI algorithm that optimizes 3D printing settings, reducing time and cost for engineers. The algorithm improved the accuracy and quality of printed models, particularly for complex biomedical devices like kidneys and prostates.
Researchers at Uppsala University used AI to predict the three-dimensional structure of a receptor, identifying molecules that bind to it with higher accuracy than traditional methods. This breakthrough accelerates the development of new drugs for mental health disorders such as schizophrenia and depression.
A commercial AI tool was effective in excluding pathology and had lower rates of critical misses on chest X-rays compared to radiologists. The study found that the AI tool could autonomously report more than half of all normal chest X-rays without decreasing standard of care.
Researchers at USC developed an approach that matches polysomnography using a single-lead echocardiogram, allowing anyone to create their own low-cost, DIY sleep-tracking device. The software significantly outperformed other EEG-less models and assesses sleep stages at the highest level.
Researchers developed Porous-DeepONet, a deep learning framework that efficiently captures complex features of porous media and learns solution operators. It outperforms traditional FEM methods in solving parameterized reaction-transport equations in porous media, handling complex domain geometries and multiphysics coupled equations.
A recent study by Columbia University Mailman School of Public Health found that Generative artificial intelligence (LLM) can't yet reliably read and extract information from clinical notes in medical records. The AI, ChatGPT-4, performed well only when given highly detailed prompts and struggled to replicate its work across trials.
Researchers have identified dozens of potential new antibiotics in the human gut microbiome, with one candidate showing promise against multidrug-resistant infections. The discovery uses artificial intelligence to analyze vast amounts of biological data and mine the world's biological information as a source of antibiotics.
Researchers found large language models are more accurate with concise, textbook-like medical questions than patient-written summaries. The models achieved higher accuracy when using standardized language, but struggled with variable phrasing and format of patient write-ups.
Researchers at Florida State University have developed a statistical analysis method that can detect when AI chatbot ChatGPT is used to cheat on multiple-choice chemistry exams. The study uses Rasch modeling and fit statistics to identify patterns in student responses that are different from those of ChatGPT.
Researchers developed a new brain-computer interface that translates brain signals into speech with up to 97% accuracy, enabling a man with amyotrophic lateral sclerosis (ALS) to communicate with friends and family. The system was tested in real-time conversations with continuous updates, achieving high word accuracy rates.
Researchers developed a framework called SigLLM that uses large language models to detect anomalies in time-series data. The approach converts time-series data into text-based inputs and can be deployed right out of the box, offering an efficient anomaly detection solution for complex systems.
A Chinese Medical Journal study developed an AI-based system to automate embryo selection and eliminate subjectivity in IVF. The system improved human embryo assessment and selection, achieving higher accuracy in embryo aneuploidy screening than experienced embryologists.
The University of Kansas study suggests that educational systems must undergo transformation to fully utilize the potential of artificial intelligence tools. Personalized learning and project-based learning are key strategies for harnessing AI's transformative power, enabling students to focus on their strengths and interests.
The Center will develop a For-Virginia AI Strategy, track business AI adoption rate, and offer programs to SMEs to understand and capitalize on emerging AI technologies. The project aims to bolster economic competitiveness of small and medium enterprises across Virginia.
Researchers at University of Bath and Technical University of Darmstadt found that large language models like ChatGPT cannot learn independently or acquire new skills, making them controllable and predictable. The study concluded that LLMs remain inherently safe, but misuse is still possible.
A Mass General Brigham study highlights inconsistencies in generative AI that can affect patient safety if not addressed. The researchers found 'drift' (model performance changes over time) and 'nondeterminism' (inconsistent results between runs) in their tests, emphasizing the need for repeated testing and monitoring.
Researchers developed a learning-based approach called LVWS to schedule robots and enable voluntary waiting, resulting in faster task completion. This method outperformed other approaches by up to 23% suboptimality, demonstrating its potential to improve manufacturing, agriculture, and warehouse automation.
Researchers at Carnegie Mellon University propose guidelines for using interpretable machine learning methods in computational biology to tackle complex problems. The guidelines address pitfalls such as relying on a single method and cherry-picking results, emphasizing the need for multiple approaches and human-centric considerations.
New research finds that AI explanations can fuel a perception of fairness without being grounded in accuracy or equity. Humans were more likely to override AI recommendations when explanations highlighted gender rather than task-relevance, but this did not improve decision-making accuracy.
A new deep learning-based inverse design method allows for the optimization of complex acoustic metamaterials, reducing noise pollution while maintaining ventilation. The approach enables ultra-broadband sound attenuation across various peak frequencies.
A survey by researchers at the University of Washington found that creating and sharing synthetic media is widely considered unacceptable, but seeking out such content is less so. The study highlights the need to expand norms around consent and privacy in the face of emerging AI technologies.
Researchers found that when participants were told their decisions would train an AI bot, they became more likely to seek a fair share of the payout. This behavior persisted even after the experiment ended, suggesting a lasting impact on decision-making.
Researchers have developed AI systems that use photos or videos to create simulations for training robots in real settings. The RialTo system generates highly accurate simulations of specific environments, while the URDFormer system creates generic simulations quickly and cheaply. These advancements aim to lower costs and increase acce...
A recent study found that generative AI images of physicians exhibit demographic biases, with a disproportionate representation of white and male physicians. The study highlights the potential for these biases to reinforce stereotypes and undermine diversity initiatives in healthcare.
A recent study published in JAMA Network Open demonstrated the effectiveness of AI in detecting myopia, strabismus, and ptosis using smartphone images. This technology has the potential to facilitate early detection of pediatric eye diseases in a convenient and accessible manner.
A new study led by CU Boulder computer scientist Theodora Chaspari found that AI algorithms can be confused by natural variations in speech patterns between different genders and races. This can lead to underdiagnosis or misdiagnosis of mental health concerns like depression.
Researchers developed a real-time stream-based data compression technology that automatically detects frequently occurring data patterns and compresses them to minimize data volume. This innovation enables high-speed, compact data compression modules in hardware without additional processors or memory.
A new study found that hospital pneumonia diagnoses are often uncertain and revised, with over half of all cases involving a change in diagnosis. This uncertainty can lead to poorer health outcomes for patients who initially lack a pneumonia diagnosis but later receive a diagnosis.
Researchers at FAMU-FSU College of Engineering have developed an AI-powered tool to train surgeons by analyzing video of their surgical technique and providing feedback. The system uses deep learning models to evaluate surgical skills effectively, offering real-time feedback for aspiring surgeons.
A team of scientists from Scripps Research has developed an AI tool that can diagnose heart conditions, including heart attacks, using just three electrodes and a simpler electrocardiogram technology. The tool was tested on 238 ECGs and showed accurate clinical assessment results similar to those from original 12-lead ECGs.
A team from Doshisha University has developed two approaches to generate adversarial examples for image cropping, achieving significant reductions in perturbation sizes. The white-box approach manipulates gaze saliency maps to produce effective images, while the black-box approach uses Bayesian optimization to target specific regions.
A novel AI tool has been shown to accurately estimate gestational age from blind ultrasound sweeps, comparable to expert sonographers. The technology has the potential to democratize prenatal care in resource-limited settings by expanding access to quality diagnostic tools without the need for expensive equipment or specialized training.
The Chameleon testbed has secured $12 million in funding to expand its services in computer science research. With this new funding, the platform will continue to innovate and support its growing community, enabling groundbreaking discoveries in CS systems research.
A study published in PLOS ONE analyzed 24 state-of-the-art Large Language Models (LLMs) and found that most produced responses rated as left-of-center when asked politically charged questions. The researchers used various tests, including the Political Compass Test and Eysenck’s Political Test, to evaluate the LLMs’ political orientation.
A new AI protocol called DeepHRD enables accurate, instantaneous detection of cancer genomic biomarkers directly from tumor biopsy slides. This breakthrough technology saves weeks and thousands of dollars from clinical oncology treatment workflows, providing access and equity in cancer care for resource-constrained settings.
The EMERGE project aims to demonstrate a new framework for coordinating artificial systems and humans. Shared awareness enables simpler AI systems to work together effectively, reducing energy costs and increasing efficiency.
A new study found that an AI program generated explanations of heart test results in most cases, with 73% deemed suitable for patients without changes. The AI explanations were also rated as easy to understand by non-clinical participants, reducing worry in many cases.
Researchers evaluate ChatGPT as a diagnostic tool for medical learners and clinicians, finding it reliable only half the time. The study warns against relying on the AI for medical advice due to its limitations.
Researchers have introduced a new AI calibration method called Thermometer, which enables efficient calibration of large language models for various tasks. This technique leverages temperature scaling to adjust a model's confidence and can generalize to new tasks without requiring additional labeled data.
A new study uses artificial intelligence to predict which breast cancer patients are most at risk for developing chronic pain, with anxiety and depression being leading factors. The AI model was built with detailed data on over 1,000 breast cancer patients and correctly predicted chronic pain more than 80% of the time.
A recent study published in the Journal of Hospitality Marketing & Management found that including artificial intelligence in product descriptions reduces purchase intentions. Emotional trust plays a critical role in how consumers perceive AI-powered products, and mentioning AI can lower trust levels.