Research applying artificial intelligence across science and engineering. Covers machine learning, sensing and quality control, immune and health prediction, and 5G networks.
A team of scientists used predictive AI to investigate the mechanism behind 'subjective color' - the phenomenon where people perceive colors in black-and-white patterns. The study suggests that learning the association between motion and color may enable colors to emerge from black-and-white stimuli.
A study by the University of Konstanz found that AI agents can reach consensus with up to 1,000 individual agents, following a mathematical law similar to magnets. However, this can lead to individual values being disregarded as the majority opinion prevails.
Large language models support medical documentation, summarize knowledge, and assist clinical decision-making, but their adoption outpaces oversight and safety systems. Risks include security, model-inherent, and human-AI interaction risks, as well as structural and ethical challenges.
AI agents, such as ERA and The AI Scientist, are increasingly taking autonomous actions, expanding scientific discovery. However, serious limitations, including hallucinated citations and outputs that are difficult to reproduce, highlight the need for human oversight and responsible development.
KAIST researchers develop a new 'oxygen tunnel' structure to stabilize oxygen vacancies in oxide semiconductors, achieving world-class current density and data retention time. The technology is expected to improve next-generation compute-in-memory systems and accelerate AI era advancements.
A recent study found that AI models, including GPT-4o and GPT-5, exhibit face-to-character biases, selecting the more competent-looking individual in most cases. This bias can worsen unfairness in decision-making contexts such as job candidate selection and parole decisions.
The American Heart Association is launching a global health tech competition to accelerate innovation in cardiovascular and brain health. The competition connects market-ready solutions with opportunities to scale into real-world healthcare, emphasizing clinical validation and alignment with evidence-based care.
A new AI framework, SHIC-XE, detects pain in horses from video analysis while providing anatomically consistent explanations for its decisions. The framework achieved strong performance in detecting pain from video, with high levels of accuracy and reliability.
A new study shows that AI can discover novel strategies that humans can adopt and preserve across generations. The study found that AI agents can discover optimal strategies that are difficult for humans to find, and that these strategies can be passed on and maintained over time. This discovery has significant implications for the rol...
Researchers found that hedge funds that effectively bet against public sentiment outperformed those that rode sentiment, earning a 0.4% monthly premium. This result persists even after controlling for fund characteristics and economic risks.
Researchers evaluated 21 language models, finding they alter their discourse to align with users' bias, creating echo chambers that reinforce preexisting beliefs. The models' behavior varies by topic, with greater shifts in stance on public safety and the economy, but consistency on corruption and democratic institutions.
Researchers suggest that biological memory is connected to emotion and thought, and that AI may lack the biological infrastructure to replicate human-like consciousness. The study proposes a biochemical process connecting brain cells and chemistry to memory, potentially offering a new route to understanding mental processes.
A new AI-powered framework, OA-UDNet, enables high-fidelity multispectral optoacoustic tomography with only 32 detectors, reducing hardware cost and complexity. The framework achieves significant improvements in image quality, resolving long-standing issues with sparse-view imaging.
According to Paul Osterman's research, 35% of US workers are marginal employees, freelancers, or gig employees with little career prospects or security. Osterman's book, Disposable Workers, examines the growing trend of firms controlling labor costs, leaving many workers in precarious positions.
SUNY Poly is part of a $19.9M NSF initiative to develop an AI-powered research platform for accelerating materials discovery. The platform will integrate automated synthesis equipment, robotics, and digital twin technology to simulate and remotely conduct experiments.
SUNY Poly will leverage generative AI, advanced modeling, and university collaboration to strengthen U.S. Army strategic operations and readiness. The project aims to provide commanders with a decisive strategic advantage in an increasingly volatile global landscape.
Researchers at Hanbat National University developed a hybrid physics-informed neural network framework for optimization of latent heat thermal energy storage systems. The framework enables rapid, autonomous design optimization by teaching the AI model governing laws of physics.
Researchers at Johns Hopkins University have developed a new system using wearable sensors and AI to generate blood pressure readings, which shows promising results in initial patient tests. The system could potentially replace the need for invasive arterial lines, reducing risks of bleeding, clotting, and infection.
The study suggests that autonomous AI systems can provide better medical care outcomes than those aided by human physicians due to their ability to process large amounts of data quickly and accurately. This is attributed to advancements in machine learning algorithms, allowing for more precise diagnosis and treatment recommendations.
A new study by Bar-Ilan University finds AI-powered tools are changing how students search for knowledge, especially in non-English-speaking countries. The research suggests that generative AI is reducing the 'language tax' faced by millions of students by making complex concepts more accessible in their native language.
Researchers at UMass Amherst have designed an edge AI system that leverages hyperdimensional computing algorithms and analog in-memory computing hardware to achieve high accuracy and efficiency. The system achieved 95.24% accuracy in language identification while reducing computing resources by 90%.
Tianhao Wang receives $677,866 CAREER Award to advance privacy-preserving data synthesis and protect sensitive information from being disclosed. His research aims to create realistic synthetic datasets that preserve useful patterns while protecting individual privacy.
Portland State University is leading a national research team using artificial intelligence to lower the cost of finding geothermal energy. The ARISE project, supported by the US Department of Energy, aims to narrow the range of estimated costs by at least 10% through machine learning and data analysis.
Multi-source data-driven machine learning is transforming lung cancer diagnosis, treatment, and prognosis by analyzing complex medical data. The review highlights the innovative applications of this technology in early screening, personalized treatment optimization, and dynamic prognostic risk stratification.
The AssistiveXR4Work project develops an AI-powered XR glasses system to automatically recognize and process visual information, enhancing daily work tasks for sight-impaired individuals. The system aims to address the shortage of skilled workforce by helping employees with declining vision continue their careers.
Researchers used machine learning to analyze elemental composition of biochar and found hydrogen-to-carbon ratio and oxygen content to be key predictors of persistent free radicals concentration and radical type. The study provides a data-driven framework for linking elemental properties to biochar reactivity and environmental risks.
A new study highlights the growing dependence on 'black-box' tools in science, which can process enormous amounts of information but lack transparency. Researchers recommend prioritizing open-source software, benchmarking proprietary tools, and intensifying efforts towards open science to address this issue.
Researchers developed an AI framework that optimizes ship routes and speed profiles to reduce fuel consumption and peak air-pollutant exposure. The system achieved 20-35% improvement in optimization performance while reducing peak pollutant exposure by 34-78%.
A new AI model, EarlyDetect, can detect precursor signals of active region emergence in the Sun's acoustic activity and magnetic field, forecasting solar eruptions nearly nine hours in advance. This technology has the potential to allow satellite communications companies or power grid companies to prepare for solar storms.
The PolyU developed AI Virtual Patient Simulation System combines multimodal data from genomic, medical imaging and clinical records to create a digital twin model tracking real-time changes in a patient's condition. It predicts the effectiveness of different cancer treatment options and supports personalized medical solutions.
Clinical AI innovations are enhancing healthcare capabilities through competition-driven development, customization, and cost-effectiveness. Modern technologies, such as rehabilitation robotics and AI-powered robots, are personalizing treatment plans to improve recovery rates and reduce healthcare costs.
Professor Fioretto's team develops AI-driven methods for autonomous power grid topology control, improving resilience and efficiency. The Genesis Mission Platform provides access to advanced AI models and high-performance computing resources.
A team of researchers developed a reusable magnetic sensing platform combining surface-enhanced Raman scattering with machine learning to detect trace uranyl ions. The system maintained its detection limit even in complex aquatic environments, with strong selectivity and resistance to interference.
By focusing on the specific regions of antibodies responsible for recognizing disease targets, the AI model improved binding affinity prediction and required fewer computational resources. This approach is similar to human-language AI models, where smaller domain-specific models trained on high-quality data can outperform larger ones.
A longitudinal study of Reddit posts since 2022 reveals that trust in generative AI is generally higher than distrust, with 31% of posts expressing trust and 26% expressing distrust. The study's findings suggest that attitudes toward AI have remained divided over the past four years.
The study proposes a public health agenda for AI to improve delivery of proven health interventions through patient identification, outreach, and care coordination. Government leadership, cross-sector partnerships, and sustained investment are essential for realizing this potential.
Researchers developed an AI model that analyzes routine whole histopathology images to predict cancer subtype, genetic mutations, and survival outcomes across 32 solid cancers. The model achieved a strong predictive accuracy score for TP53 mutation detection and demonstrated the ability to infer RNA expression levels and tumor taxonomy.
A national survey reveals half of sleep-deprived nursing moms rely on AI for breastfeeding advice, often leading to preventable health scares. Pediatricians emphasize the importance of human guidance and expertise over AI-driven solutions.
Researchers developed a novel semiconductor integration platform, BBCube, combining advanced chip packaging, high-density interconnects, and improved thermal management. This enables more precise chip placement, faster communication, and efficient cooling for powerful and energy-efficient AI accelerators.
The review highlights security and ethical risks in AI-powered embodied systems, including hallucinations, synthetic forgeries, and adversarial attacks. It proposes a roadmap toward dependable embodied intelligence through safeguards like contextual checking, forgery detection, and risk-aware reasoning.
Researchers combined AI, genetics, and gut microbiome analysis to shed light on intestinal fibrosis in Crohn's disease. They identified a shared set of 43 key genes linked to disease progression and found that bowel fibrosis is driven by ongoing immune activation, damage to the intestinal lining, and changes in gut bacteria.
Hankelformer improves forecasting of extreme weather events by capturing local spatiotemporal dynamics and enhancing feature invariance. It achieves state-of-the-art performance on multiple datasets, including energy, transportation, and extreme weather domains.
A pilot study by Bayes Business School found that shoppers using digital screens on their supermarket trolleys spent up to 32% more than those without. The study recorded significant spikes in spending and consumption among 'superusers' who engaged with the technology most.
The open-source image analysis pipeline accurately measures brain tissue changes in animal models, reducing variability among thousands of scans. The tool has the potential to improve early-stage evaluation of new therapies for acute ischemic stroke.
Large language models exhibit a preference for higher-quality options first, but later options when quality decreases. Models also show unique biases in name preferences, regardless of context, making them harder to predict and prevent.
A new survey finds that opposition to local data centers has risen sharply among US adults, with 61% opposing construction in their areas. Meanwhile, views on artificial intelligence remain stable, with 39% expressing concerns about its impact.
The DOE has committed funding for UNITY-3, a cutting-edge breeding blanket test facility at ORNL, with Kyoto Fusioneering relocating its US headquarters to Tennessee. This partnership establishes a shared commercial-scale testing infrastructure to validate critical systems and train digital twins of breeding blanket concepts.
Arkansas researchers used a machine-learning approach to study the organization of neighboring genes in bacteria. The novel method distinguished disease-causing strains of Enterococcus cecorum from nonpathogenic ones by analyzing how neighboring genes are organized within the bacterial genome. This new approach may provide valuable clu...
Researchers have developed a machine learning model that can help clinicians assess uncertain variants in prenatal genetic testing, providing more accurate diagnoses and clearer information for families. The approach uses tissue-agnostic episignatures to overcome limitations in epigenetic testing.
The KAIST research team developed an explainable AI technology that detects patterns of foreign-linked influence operations in online news comments. The model identifies 23,998 accounts exhibiting patterns consistent with public-opinion manipulation, targeting division and confrontation within Korean society.
Researchers have developed an AI semiconductor device that temporarily remembers recent inputs while autonomously forgetting older information. This technology enables continuous processing of complex time-series signals without a separate reset process, representing a key innovation for low-power edge AI systems.
Researchers at Aalto University developed an AI model that accurately explains how humans read, using reinforcement learning to recreate reader choices. The model can power smarter AR displays and tailor complex texts to different readers and situations.
A team of researchers found that humans and AI process language similarly during the initial stages of reading, relying on next-word predictions. However, as passages become more complex, human processing differs from AI, highlighting areas where human and machine language understanding diverge.
Florida Atlantic University's video compression technology has been acquired by Dolby Laboratories, contributing to the international VVC Video Compression Standard and reducing video data requirements by approximately 50% while maintaining visual quality. The portfolio comprises 683 IP assets, including 250 granted patents worldwide.
Halide cathode materials, long overlooked due to dissolution in liquid electrolytes, now enable dramatically higher energy densities through a fundamental shift in battery chemistry. Key strategies include multi-electron reactions, protective coatings, and nanostructuring to address stability issues.
A new study by Dr. Wu Yuan's team reveals that retinal AI can lose accuracy in high-altitude populations due to altitude-associated domain shift. MIXFound, a lightweight framework, offers a practical solution to improve robustness and correct for these differences.
The SNU team introduces Cluster-aware Upcycling, leveraging semantic structure of pretrained models to promote specialization among expert modules. This approach outperforms conventional Sparse Upcycling on image-text retrieval and various image classification benchmarks.
The new integrations provide enterprise AI agents direct access to live, structured research data from Dimensions' 430M+ interconnected records. This allows for AI-assisted analytics across one of the world's most comprehensive linked views of global research activity.
Assistant Professor Yingxue Zhang's project aims to develop urban AI models that can efficiently process vast amounts of human-generated data to optimize commute times, traffic safety, and more. The model will utilize offline reinforcement learning to tackle spatial-temporal dynamics in urban life.
Prof. Haim Sompolinsky receives the 2026 Dirac Medal for pioneering contributions to equilibrium statistical mechanics and non-equilibrium statistical mechanics. His work helps establish the field of theoretical neuroscience, elucidating how collective activity supports memory, computation, and learning.