Articles tagged with Artificial Intelligence
Researchers at University of Waterloo discover workaround for 'no cloning' problem in quantum computing by encrypting quantum information as it's copied. This breakthrough enables redundant and encrypted quantum cloud services, a crucial step towards building quantum computing infrastructure.
Researchers created an AI model to design peptides targeted by proteases, which are overactive in cancer cells. The peptides can be detected in urine, revealing specific types of cancer. This technology could lead to at-home tests for various cancers.
Researchers found that postgraduate students selectively engage with GenAI tools when revising critical reading reports, focusing on key analytical dimensions. This strategic engagement is driven by academic goals, supervisor demands, and career aspirations.
The use of generative AI in software development can lead to increased productivity, but also creates technical debt and hasty coding if not managed properly. Companies must prioritize fixing technical debt, develop clear guidelines for AI-assisted coding, and train developers on the hazards of AI coding.
Generative AI is increasingly used in classrooms, raising concerns about its impact on student learning. Researcher Gerd Kortemeyer explores the benefits and drawbacks of AI in physics education, suggesting that it can be a helpful tool when integrated into teaching methods that focus on critical thinking and collaboration.
The SHASAI project addresses cybersecurity risks in AI systems from design to real-world operation, combining secure hardware and software, risk-driven engineering, and real-world validation. The project will demonstrate its methods and tools in three diverse scenarios, supporting Europe's efforts to promote trustworthy AI.
A new synthesis of seismic research shows that AI combined with physical principles is rapidly transforming how scientists image the Earth's shallow and crustal structures. The study highlights the need for physics-guided AI frameworks that balance computational efficiency with interpretability.
Jundong Li, an associate professor at the University of Virginia, has received the 2025 Tao Li Award for his significant contributions to data mining and machine learning. His research focuses on developing models that can extract actionable insights from structured data, particularly graphs.
The first Editorial of 2026 discusses AI's impact on the scientific community. The journals use select AI tools, with encouraging results in data sharing among published research articles.
Recent progress in designing lighter, faster, and more resource-efficient multimodal AI systems is highlighted in a comprehensive survey. Vision token compression techniques and emerging architectures like mixture-of-experts selectively activate model components to increase capacity without proportional increases in computation.
A new AI model using deep transfer learning accurately predicted spoken language outcomes in deaf children after cochlear implants. The 'predict-to-prescribe' approach allows for earlier intensified therapy for children forecast to struggle with spoken language.
A research team at Huazhong University of Science and Technology developed an Ag/Sb2O3/Au molecular-crystal memristor array for real-time edge computing in smart power grid inspections. The device enables ultra-low-power switching, supporting non-volatile and volatile modes for tasks like image convolution and transient event detection.
A study published in EMBO Molecular Medicine has identified a combination of statins and phenothiazines that shows promise in treating aggressive neuroblastoma. The drug combination was found to impede tumour growth and improve survival rates in laboratory trials with mice.
The new framework reduces processing time by an average of 87.7% compared to the state-of-the-art approach, achieving cache hit rates between 70% and 90%. It demonstrates linear scalability, handling high query loads efficiently.
A team of researchers at Osaka Metropolitan University has created two models to detect and fix common labeling errors in large radiographic collections. The models achieved high accuracy rates for body-part classification and projection/orientation detection, showing promising results for improving deep-learning models in medical AI.
HSE economists found that AI models like ChatGPT and Claude tend to play 'too smart' and lose in strategic thinking games by assuming a higher level of logic in people than is actually present. The study replicated results from previous human participant experiments, showing LLMs adapt to opponents with varying levels of sophistication.
A new study analyzed over 460,000 scientific abstracts to identify key themes, trends, and research gaps in aging research. The study found a growing separation between basic biological studies and clinical research, highlighting the need for integration to translate laboratory discoveries into medical applications.
Researchers developed an AI-based noninvasive tool to predict oropharyngeal cancer spread, identifying patients who should receive aggressive treatment or undergo de-intensification. The tool improved risk stratification, enabling more accurate predictions of survival and cancer spread.
A new study finds that large language models like ChatGPT significantly boost paper production, especially for non-native English speakers. However, this increase in AI-written papers makes it harder to separate valuable contributions from low-quality work.
Researchers at the University of Arizona developed a mesh sleeve that monitors leg movements using AI to detect early warning signs of frailty. The device sends only the analyzed results, reducing transmission and internet requirements.
As chatbots become more sophisticated, they are likely to generate and spread gossip, leading to reputational damage, shame, and social unrest. Drs. Joel Krueger and Lucy Osler warn that this 'feral' AI gossip can be particularly damaging due to its ability to operate unconstrained by human norms.
Researchers at Tsinghua University Press propose integrating AI technology into launch vehicles to address challenges of high-frequency launches and reliable flights. The system aims to achieve hour-level test, inspection, and decision-making times for large launch vehicles.
Researchers used causal AI to extract insights from ARPES data of cesium vanadium antimonide, a kagome superconducting material. The technology revealed that the chemical bonding state of cesium atoms strongly influences the electronic state of the V3Sb5 layer, responsible for superconductivity.
Researchers from RIKEN and their international collaborators used deep learning techniques to analyze unexamined nuclear emulsion data and identified a new double-Lambda hypernucleus. This is the world's first AI-assisted observation of such an exotic nucleus, providing new insight into neutron star cores.
The new program focuses on advancing foundational research in AI, including innovation in language models and algorithmic efficiency. Google will support research grants, scholarships for students, and educational initiatives at the TAU Center.
A new study uses AI to regulate mechanical ventilation for patients with ARDS, aiming to provide personalized support. The tool combines live patient monitoring with historical data to modulate ventilation parameters and reduce lung injury.
Researchers found that AIs consistently converged on 12 common themes despite diverse prompts, suggesting biases in training data. The models failed to generate novel or creative outputs, highlighting the need for anti-convergence mechanisms and human input for AI's creative potential.
CORAL accurately predicts genetic subtypes and patient outcomes in multiple myeloma without costly genomic tests. The technology could be applied to other cancers, opening the door to personalized, faster and widely accessible cancer care.
Researchers developed RNACOREX, a new open-source software tool that identifies gene regulation networks in cancer. The tool analyzes thousands of molecules simultaneously to detect key interactions, providing an interpretable molecular map that improves understanding of tumors.
Researchers developed a scientifically validated personality test framework for AI chatbots, showing that larger models accurately emulated human traits and could be reliably shaped through prompts. The study raises concerns about AI safety and ethics, highlighting the need for regulation to prevent misuse.
Researchers have identified DNA switches that control how brain cells called astrocytes work, which are known to play a role in Alzheimer's disease. The study used CRISPRi technology and single-cell RNA sequencing to test nearly 1000 potential switches, finding that about 150 of them controlled genes implicated in Alzheimer's disease.
A new AI method has identified 51 old pea varieties with high starch and protein content, potentially suitable for producing plant-based foods. The study shows a close relationship between seed appearance and chemical composition, enabling partial prediction of properties based on images.
The ACM's Technology Policy Council has issued a report on automated speech recognition (ASR) highlighting the need for rigorous audit and governance practices. The report raises concerns about technological bias, inequity, and lack of transparency in ASR systems.
A novel approach has been proposed to facilitate collaboration between human researchers and AI in self-driving labs. The AI advisor model enables humans to make real-time decisions while leveraging AI's data-processing prowess.
A new AI-assisted device, CanvasDx, has been shown to improve autism care access in rural Missouri communities. By reducing travel distances and processing times, families can receive diagnoses up to 5-7 months earlier.
Researchers will use revolutionary technology to transform understanding of extreme environments, revealing how black holes behave and evolve. The project aims to deliver dynamic gravitational tomography, creating 3D movies showing plasma flows around black holes.
A research team at the University of Ottawa developed an AI system to autonomously defend wireless networks against jamming attacks. The dual-agent framework predicts interference events and makes real-time decisions to maintain essential communications during attacks.
Researchers at IIT identified a candidate molecule called Apt1 that enhances existing anticancer therapies by making tumour cells more vulnerable to chemotherapy drugs. The molecule slows DNA repair and impairs the interaction between RAD51 and BRCA2 proteins, inducing synthetic lethality in cancerous cells.
Mayo Clinic researchers developed an AI-ECG model that accurately diagnoses advanced chronic liver disease, allowing for early intervention and potential lifesaving treatment. The study found the model detected twice as many cases as standard methods, highlighting the promise of AI in preventive care.
A new study found remarkable variation in how populations evolve in variable environments, with some cases benefiting from changes and others being hindered. The research has implications for understanding evolution and adapting to climate change, as well as informing AI and machine learning.
A new study from the University of East Anglia finds that AI tools excel at speed and clarity, but human translators remain crucial for tone, cultural nuance, and natural sound. Generative AI can handle parts of video translation, but consumers still notice when something feels off.
A small analysis found AI-generated images of doctors exaggerate and reinforce existing stereotypes on sex, gender, race, and ethnicity. This contradicts recent medical workforce statistics, which show a more diverse range of doctors in the UK and US.
Dr. Tom McClelland from the University of Cambridge believes that our current evidence is insufficient to determine whether artificial intelligence has achieved consciousness. He suggests that sentience, which involves positive and negative feelings, is a more critical factor in making AI ethically significant.
A new AI framework uses a pre-trained viral protein language model to predict the adaptation risk of unknown viruses. The framework, called GIVAL, achieves high accuracy and robustness even with incomplete data, providing critical support for early warning and proactive prevention and control of viral infectious diseases.
A new computational framework allows researchers to recreate different evolutionary trees by changing the structure of the world and tasks AI agents complete. The framework showcases how tasks drove eye evolution in agents, with navigation tasks leading to compound eyes and object discrimination driving camera-type eyes.
Researchers propose integrating processing capability within memory units to reduce energy consumption and latency in AI applications. Inspired by the brain's efficient processing mechanisms, spiking neural networks (SNNs) can respond to irregular events and store information in the same place.
Researchers developed an AI model that can detect coronary microvascular dysfunction using a common electrocardiogram, outperforming previous models in diagnostic tasks. The model can accurately identify a condition often missed in emergency department visits, providing a cost-effective and non-invasive way to diagnose serious heart co...
Researchers have launched a new multimodal medical dataset, Observer, capturing anonymized, real-time interactions between patients and clinicians. The dataset links video, audio, transcripts, and electronic health records to study subtleties like body language and environmental factors affecting care.
A new study led by Harvard Medical School researchers found that AI models analyzing pathology samples can infer demographic information, leading to biased diagnoses. The team developed a framework called FAIR-Path to reduce bias in these models, which showed improved performance across demographics.
An AI-driven robotic assembly system allows people to build physical objects by simply describing them in words. The system uses generative AI models to build a 3D representation of an object's geometry based on the user's prompt, and then iterates on the design based on feedback from the user.
A recent study reveals that sampling biases in bacterial population structure can severely undermine the accuracy of AI models predicting antimicrobial resistance. The researchers recommend evaluating future methods to address this issue.
Researchers propose a novel approach to AI hardware design by integrating neuromorphic systems and compute-in-memory techniques to overcome the limitations of modern computing hardware. This could lead to more efficient data center energy use and enable real-time intelligence in compact, power-constrained systems.
A new study warns that AI is racing ahead of safety checks in GP clinics, putting patients at risk. The research found that many GPs use AI tools without thorough evaluation or regulatory oversight, carrying risks like automation bias and loss of important social details.
A team of MIT engineers developed a deep-learning model that predicts how individual cells will fold, divide, and rearrange during a fruit fly's earliest stage of growth. The model achieved 90% accuracy in predicting the movement of 5,000 cells over the first hour of development.
AI is becoming increasingly used in peer review, with 53% of reviewers now utilizing AI tools to enhance research quality and trust. The study highlights significant untapped potential for AI to support rigor, reproducibility, and deeper methodological insight.
Researchers at Mayo Clinic have developed a smartwatch-based alert system that signals parents to intervene in children with emotional and behavioral disorders, reducing tantrum duration by an average of 11 minutes. The system uses physiological stress signals detected by the smartwatch to prompt parental intervention.
The University of Oklahoma researcher is working on a project using inverse design techniques and AI to create advanced materials that can quickly switch between conductive and insulating states. The goal is to reduce uncertainty in the discovery process and create a scalable material-design methodology.
Stanford engineers develop an AI framework called MetaChat, which rapidly accelerates metasurface design using computational tools and self-reflective AI assistants. This enables rapid solving of optics-related problems, such as designing metal lenses that can focus different wavelengths of light.
Researchers at UT Health Sciences Center have developed an AI-enhanced platform called ENRICH, which leverages advanced analytics with community-based human support to reduce treatment interruptions and improve cancer outcomes. The program identifies patients at highest risk for interruption and deploys targeted support resources.
Researchers used AI to study covert attention and found emergent properties in artificial neural networks, including new neuron types with response properties never highlighted before. The findings have implications for our understanding of the human brain and its ability to process information.