Articles tagged with Artificial Intelligence
Lehigh University researchers used machine learning to compare bone marrow extracted from the hip and shoulder, finding six proteins that distinguish between the two extraction sites. This study may lead to standardized BMAC extraction protocols and personalized treatments based on protein concentrations.
Researchers propose using the Agent Deed Consequence (ADC) model to program ethical values into smart city technologies. The model captures human moral judgments by considering agent intent, deed, and consequence, enabling AI systems to distinguish between legitimate and illegitimate orders.
A new quantum-secured data transmission architecture has been proposed to address the challenges of AI-driven data centers. The system achieves terabit-per-second capacity while defending against future quantum threats through self-homodyne coherent transmission and integrated quantum key distribution.
The ESMO Guidance on the Use of Large Language Models in Clinical Practice (ELCAP) provides a structured set of recommendations for safely adopting AI language models in oncology. The guidance emphasizes the importance of human oversight, transparent tools, and continuously monitored systems to ensure trust in AI-driven cancer care.
Researchers used five advanced LLMs to assess their logical reasoning capabilities and found that they consistently complied with requests for false medical information. Targeted training and fine-tuning improved LLMs' abilities to respond to illogical prompts accurately, but challenges remain in aligning models to every type of user.
A recent study found that most users cannot identify AI bias in training data, unless it affects their own group. Researchers created 12 versions of a prototype AI system to detect facial expressions and tested how users might detect bias in different scenarios.
The European Society of Endocrinology has released the EndoCompass Research Roadmap, a major new initiative to align research efforts and improve funding strategies for hormone-related health challenges. The roadmap identifies specific research needs across eight endocrine specialties and five cross-cutting areas.
Recent research found that large language models are not yet able to consistently fool humans in conversations. They struggle with using discourse markers, opening and closing features, and subtle imitations. Despite rapid development, key differences between human and artificial conversations will likely remain.
Researchers highlight the importance of designing 'human-in-the-loop' systems with expert input for AI model training and validation. Machine learning models offer more interpretable results, but generative AI requires robust safeguards to protect patient data.
A new large language model, LassoESM, has been developed to predict lasso peptide properties, enabling the acceleration of rational design for biomedical applications. The model was trained on thousands of lasso peptide sequences and demonstrated accurate prediction of various properties.
Researchers built a digital archaeology framework to learn about ancient humans creating finger flutings, which are marks drawn on cave walls. The study used tactile and VR setups to explore AI image recognition methods, finding promising insights but also limitations.
Researchers from MIT and the MIT-IBM Watson AI Lab have introduced a new training method that enables vision-language models to localize personalized objects in a scene. By using carefully prepared video-tracking data with contextual clues, the model is better able to identify the location of a specific object in a new image.
VFF-Net applies label-wise noise labelling, cosine similarity-based contrastive loss, and layer grouping to improve image classification performance compared to conventional forward-forward networks. The algorithm reduces test errors on various datasets, enabling lighter and more brain-like training methods that make AI more sustainable.
A new computer model called CogLinks simulates brain circuit decisions and adaptability, helping researchers understand how the brain misfires in psychiatric disorders. The model's results suggest that a specific pathway is essential for adaptability, and its predictions are confirmed by an fMRI study
Dr. Benjamin P. Brown proposes a targeted approach to improve the accuracy and speed of machine learning in drug discovery. His work focuses on creating a more generalizable deep learning framework for structure-based protein-ligand affinity ranking.
A landmark study analyzed health data from over 600,000 patients across 10 countries to assess patient risk for non-ST-elevation acute coronary syndrome (NSTE-ACS). The AI-powered model GRACE 3.0 predicts risk more accurately and guides personalized treatment decisions.
A new AI-based method optimizes the operation of solar power generation and battery storage systems, reducing imbalance penalties by approximately 47% compared to conventional control methods. The method maintains stable profits throughout the four seasons and can handle real-world uncertainties such as sudden weather changes and compl...
Researchers discovered a common relational structure across all participants when observing relationships between general activity patterns in neurons. This structure may be the brain's way of organizing information to ensure similar perceptions despite underlying neural coding differences.
A recent study reveals that the response to witnessing robot mistreatment depends on factors such as the robot's humanlike design and the observer's moral identity. Anthropomorphism influences empathy, encouraging customers to treat robots with dignity, while moral identity plays a crucial role in shaping behavioral contagion.
CompositesAI helps users create and analyze composite products without requiring in-depth technical knowledge. The platform is initially focused on rotor blades for air mobility, helicopters, and wind turbines, but its uses will expand to handle other composite structures.
Researchers developed a cyberattack detection system that uses federated learning and cloud coordination to detect DDoS attacks in 6G-ready smart grids without exposing user energy-use data. The system achieved high accuracy and precision, but trade-offs were observed in terms of resources consumption.
A new AI-powered tool, EZSpecificity, can predict the best enzyme-substrate combination for various applications. The tool outperformed existing models in accuracy, especially for halogenase enzymes.
Researchers at the University of Córdoba have developed two Artificial Intelligence methodologies capable of predicting extreme wind speeds with greater accuracy than traditional methods. The systems, trained on over 13 years of data, excel in forecasting severe events, allowing turbines to be shut down and preventing damage or collapse.
A novel AI optimization model called GAN-Solar has been developed to address the technical bottleneck of accurate short-term solar forecasting. The model achieves significant improvements on key metrics compared to existing advanced models, producing high-definition forecasts that capture crucial details.
Scientists from Japan developed a theoretical framework that explains how collective cells can perform complex tasks. The key is distributed information processing and reinforcement learning in the environment.
Researchers found that facial self-touching, particularly around the nose, chin, and cheeks, strongly correlates with stress levels during cognitive work. The study also suggests that this behavior may be an evolutionary, self-soothing mechanism to regulate stress.
A new project aims to develop a computationally efficient model that accurately predicts how additive manufacturing process parameters influence the solidification microstructure of binary alloy solidification. This will enable optimization of additively manufactured parts with confidence in critical industries.
A recent study used AI algorithms to analyze medical imaging and predict treatment response in patients with locally advanced nasopharyngeal carcinoma. The AI-based radiomics model achieved high accuracy in predicting treatment response and prognosis, revealing a link between imaging-derived features and the tumor microenvironment.
MetaSeg achieves the same segmentation performance as U-Nets but requires 90% fewer parameters, making medical image segmentation more cost-effective. The new approach leverages implicit neural representations to quickly adjust to new images and decode accurate labels.
A new AI tool, SpectroGen, uses generative AI to quickly assess material quality by generating spectra in less than one minute. It can replace traditional methods that take several hours or days, improving productivity and efficiency in industries such as manufacturing and pharmaceuticals.
A new AI system can generate realistic future X-rays alongside risk scores for osteoarthritis progression, giving doctors and patients a clear visual forecast of how the condition may develop over time. The system outperforms comparable tools in predicting osteoarthritis progression with nine times faster speed and accuracy.
The project aims to improve undergraduate education by introducing situated case studies to help students understand AI complexity, critical reasoning, and applications. The initiative serves the national interest by preparing future professionals to use and develop AI, advancing their understanding of workforce innovation.
CMU researchers use AI and robotic mobility to give everyday objects foresight, predicting interventions and moving across horizontal surfaces to help humans. The system uses computer vision and LLMs to reason about a person's goals, translating what the camera sees into a text-based description of the scene.
The report presents expert views on AI's potential benefits and risks in healthcare, including its application in clinical care, biomedical research, and business domains. The authors discuss the need for regulation, evaluation, and implementation of AI in healthcare to ensure its safe and effective use.
The platform allows users to create personalized LLM-based chatbots, serving as a 24/7 teaching assistant or scraping campus websites to find needed resources. Illinois Chat has already been utilized in various departments and research groups, providing valuable feedback to make it more practical.
Researchers developed AI models to identify high-risk children for sepsis within 48 hours, enabling early preemptive care. The study used electronic health record data from the Emergency Department and showed robust balance in identifying at-risk children without overidentifying those who are not at risk.
The ESMO Congress 2025 will cover key topics including precision oncology, antibody-drug conjugates in breast cancer, dose optimisation strategies, new treatment modalities for melanoma, and immunotherapy-based approaches across various cancer settings. Promising Phase-3 studies on ADCs in bladder and lung cancer are also expected.
Researchers developed a more accurate cough-detection model using wearable health monitors' audio and movement data. The new model can distinguish between coughs and nonverbal sounds, improving the accuracy of respiratory disease tracking.
A new smartphone-based AI system accurately predicts avocado firmness and internal quality with high accuracy, enabling consumers to avoid overripe avocados. The technology has the potential to assess the ripeness and quality of other foods, reducing global food waste by 50% by 2030.
Researchers at the University of Missouri have developed an AI-powered method to detect hidden hardware trojans in chip designs, offering a 97% accurate solution. The approach leverages large language models to scan for suspicious code and provides explanations for detected threats.
URNet transforms rapid event signals into accurate depth maps using local-global refinement and uncertainty-aware learning. The system produces confident predictions and adjusts response in low-confidence situations, enhancing robustness and trustworthiness in real-world conditions.
Researchers at Kyushu University have developed a new method to build more energy-efficient magnetic random-access memory (MRAM) using thulium iron garnet. The team successfully produced thin films of platinum on the TmIG material, enabling high-speed and low-power information rewriting at room temperature.
Geoffroy Hautier, a materials scientist at Rice University, has been elected a fellow of the American Physical Society for his groundbreaking research in high-throughput computational materials design and discovery. His work bridges quantum mechanics, computation, and artificial intelligence to accelerate the discovery of new materials...
Research highlights AI scams targeting Latino population with deepfake, voice-cloning and AI voice bots. Technical writing courses can help increase AI literacy and detect deception.
Researchers tested an AI-supported diagnostic method at rural hospitals in Kenya and Tanzania, analyzing cervical cell samples and human papilloma virus (HPV) samples using digital tools. The study found that the AI method worked well technically but faced challenges with inconsistent images and follow-up care.
Researchers from Shanghai Jiao Tong University and Hong Kong universities develop a comprehensive review on advanced AI chip design, focusing on energy efficiency, processing speed, and hardware scalability. The work provides a strategic roadmap for the development of high-performance AI chips through interdisciplinary collaboration.
Researchers at Texas A&M University have developed a new framework to protect industrial processing facilities from cyber threats. The framework identifies vulnerabilities, detects abnormal activity in real-time, and provides safeguards and mitigation strategies to maintain safe operations.
Plant functional diversity varies greatly depending on seasonal cycles and wet-dry periods, according to a new study that analyzed 4,000 satellite images. The researchers used AI algorithms to derive plant traits from the images and calculated quantitative measures of functional diversity.
A council of five AI models trained to deliberate together achieved higher accuracy on medical licensing exams than any single chatbot alone. This collaborative approach showed promise for improving the reliability and trustworthiness of AI tools in healthcare.
A new approach called Streamlined Life Cycle Assessment (SLCA) uses AI, 3D modeling, and secondary databases to assess a product's environmental footprint. SLCA cuts the time required for assessments from months to just one week, while maintaining accuracy.
Researchers developed an AI-based generative approach to discovering technology opportunities from patent maps using machine learning. The system translates patent vacancies into human-readable text, enabling the identification of untapped technologies and facilitating innovation forecasting.
Researchers at the University of Tokyo developed an AI-based framework, scHDeepInsight, to rapidly and consistently identify immune cells from gene data. The system uses hierarchical learning and image-based representation to distinguish broad and fine subtypes with improved accuracy and consistency.
Researchers at Harvard Medical School developed Dr. CaBot, an AI system that can generate a differential diagnosis and explain its nuanced reasoning like an expert diagnostician. The system reached a comparable final diagnosis to a human expert in a published case presentation, demonstrating its potential for medical educators and phys...
Researchers developed DeMemSeg, an AI-driven pipeline that accurately segments overlapping membrane structures with accuracy comparable to expert manual analysis. The approach enables large-scale, objective, and quantitative analysis of morphological data, providing a foundational technology for advancing disease mechanisms.
A new study uses general-purpose AI to classify real cosmic events and explain its reasoning without complex training. The model achieved approximately 93% accuracy and provided plain-English explanations for every classification.
The Indo-Swiss Workshop on 'Research Integrity in the Age of AI' brought together policymakers, scientists, and publishers to explore responsible AI use. Frontiers shared its pioneering approach to detecting research fraud using human-in-the-loop AI systems.
Researchers at NC State University identified a hardware vulnerability that allows attackers to compromise AI privacy by exploiting ML accelerator behavior. The GATEBLEED vulnerability exploits power gating in AI accelerators, creating an observable timing channel for attackers.
Researchers have developed an AI model that creates a detailed map of the mouse brain featuring 1,300 regions/subregions. The CellTransformer model uses spatial transcriptomics data to identify previously uncharted subregions of the brain, opening new avenues for neuroscience exploration.
Researchers use AI, omics, and systems biology to design targeted drugs for cardiovascular disease pathways. The tools could transform heart drug development and save lives, but global health policy leadership is urgently needed.
Scientists at MIT developed a method to predict how plasma in a tokamak will behave during rampdown, achieving high accuracy with limited data. This new model could significantly improve the safety and reliability of future fusion power plants.