Articles tagged with Artificial Neural Networks
A team of computer scientists created 2,300 original sudoku puzzles and asked AI tools like OpenAI's ChatGPT to solve them. The results showed that while some AI models could solve easy sudokus, most struggled to provide accurate explanations, raising questions about the trustworthiness of AI-generated information.
Researchers at the University of Rochester are developing biologically inspired predictive coding networks for digital image recognition using analog circuits, which could lead to more efficient drones. The team aims to approach the performance of existing digital approaches and translate it to complex perception tasks needed by self-d...
Researchers highlight the potential of solvent-based recycling and AI-assisted sorting to recycle complex plastics. However, the study emphasizes that replacing fossil-based plastics with biobased alternatives poses significant challenges, requiring comprehensive approaches and life cycle assessments.
A new study by Cornell University reveals Amazon's AI shopping assistant Rufus gives vague or incorrect responses to users writing in some English dialects like AAE. The researchers propose a framework for evaluating chatbots that can better serve users from diverse linguistic backgrounds.
Researchers found that LLMs are better at representing non-sensorimotor concepts but struggle with sensory and motor concepts. Incorporating sensory input improves LLM performance and leads to more human-like representation, consistent with previous human studies on multimodal learning.
A new study reveals that AI systems transition from relying on word positions to meaning-based understanding as they receive enough data for training. The transition occurs abruptly, similar to a phase transition in physical systems, and is driven by the amount of data available.
A novel method predicts the working status of high-formwork support systems using a combination of finite element model simulations, deep learning, and large language models. The framework achieves superior performance over existing methods and demonstrates potential applications in complex structures.
Researchers developed an AI-powered microscope system to measure soil fungi presence and quantity, providing insights into soil health and fertility. The low-cost optical microscopy with machine learning technology can be used by farmers and land managers worldwide.
Researchers developed a simplified model to explain visual processing in the primary visual cortex, achieving 75% accuracy with fewer layers. The 'minimodels' for individual neurons are just as powerful as large models, providing an accurate and interpretable way to study visual computation.
A new study in Nature Communications found that AI models exhibit a geometric property called convexity, which helps humans form and share concepts. Convexity is also linked to the performance of AI models on specific tasks.
Researchers pioneer a new AI method to uncover cognitive strategies in decision-making, revealing suboptimal behavioral patterns. The study uses small neural networks to predict animal choices with accuracy comparable to larger neural networks.
Researchers used machine learning to simulate galaxy evolution and supernova explosions, achieving speeds four times faster than supercomputers. This breakthrough enables the study of galaxy origins, including the creation of the Milky Way's elements essential for life.
The AI for Good Global Summit 2025 will showcase AI innovations delivering better healthcare and education, reducing disaster risks, ensuring water and food security, and bolstering economic resilience. The event, organized by the International Telecommunication Union (ITU), features talks from AI leaders and 100+ demos.
An interdisciplinary team at TU Wien has developed a method that allows for the exact calculation of how reliably a neural network operates within a defined input domain. This enables mathematical guarantees for the safe use of AI in sensitive applications.
A new diagnostic tool uses AI to analyze handwriting signals, detecting subtle motor symptoms associated with Parkinson's disease. The device has shown an average accuracy of 96.22% in distinguishing patients from healthy individuals.
Computer scientist Zeynep Akata has developed innovative methods to combine visual, linguistic, and conceptual elements in AI to increase user trust. Her research on explainable AI aims to make image classification decisions more transparent.
Researchers developed an AI tool called AAnet to characterize cancer cell diversity, identifying five distinct cell groups with different gene expression profiles. This could lead to more targeted therapies and improved patient outcomes.
An AI model developed by Ehsan Ghane at the University of Gothenburg can predict the durability and strength of woven composite materials, reducing development time. The model integrates material laws to make extrapolations outside training data, enabling better understanding of material behavior.
Researchers found that brain's dopamine neurons encode a map of possible future rewards across time and magnitude, guiding adaptive behavior in uncertain environments. This biological insight aligns with recent advances in AI, particularly distributional RL algorithms, which learn from reward distributions rather than averages.
Dr. Deanna Kaplan's innovative voice-capture app, Fabla, captures unstructured voice narratives to study how clinical interventions influence daily life. The platform has found applications across diverse health domains, including veteran experiences and healthcare provider burnout.
A new study by UChicago scientists found that AI-powered weather prediction models are remarkable but not magical, struggling to predict unprecedented weather events. The model can achieve impressive accuracy for short-term forecasts but fails to extrapolate beyond existing training data, leading to false negatives and potential mispre...
A team of researchers developed a machine learning model called Aurora that accurately forecasts various Earth systems, including air quality and tropical cyclone tracks. The model outperforms traditional systems at a fraction of the cost, enabling better preparedness for extreme weather events.
Researchers found that artificial intelligence tools can accurately predict disease for patients with typical symptoms but struggle with those exhibiting atypical symptoms. Human oversight is necessary for high-quality patient-centered care when using AI as an assistive tool.
Training neural networks on simple cognitive tasks improves their ability to handle more complex ones. By adopting the principles of early childhood education, researchers found that recurrent neural networks can be trained faster and with better results.
Researchers at Duke University have developed a new framework called HUMAC that enables robots to collaborate like humans by teaching them Theory of Mind. After just 40 minutes of guidance, robot teams exhibited strong collaborative behaviors and achieved high success rates in simulations and physical tests.
Researchers discovered similarities between AI and human brains with aphasia, offering new insights into diagnosis and improving AI's fluency. The study suggests that understanding internal patterns in AI models may lead to smarter and more trustworthy AI.
A study suggests that groups of artificial intelligence language models can self-organise into societies, reaching consensus on linguistic norms, and are prone to tipping points in social convention. Collective biases emerge between agents through interactions, a blind spot in most current AI safety work.
Researchers propose Input-Driven Plasticity model, which integrates past and new information to guide memory retrieval. The model is robust to noise and uses it as a means to filter out less stable memories.
A University at Buffalo-led study proposes using AI-powered handwriting analysis to identify spelling issues, poor letter formation, and other indicators of dyslexia and dysgraphia. The work aims to augment current screening tools and provide an early detection tool for these neurodevelopmental disorders.
Researchers have designed a headphone system that translates several speakers simultaneously, preserving voice direction and qualities. The Spatial Speech Translation system uses off-the-shelf noise-cancelling headphones fitted with microphones to separate out different speakers in a space and translate their speech.
The Global Confidence Degree-based Graph Neural Network (GCD-GNN) framework improves financial fraud detection by integrating global confidence metrics with advanced graph learning techniques. It achieves record-breaking accuracy on real-world datasets, including a 97.26% AUC on T-Finance.
A new review advocates for building confidence in AI applications by implementing robust data governance frameworks, enhancing transparency, and involving stakeholders. The authors emphasize the importance of addressing ethical implications and ensuring equitable access to AI-driven innovations in clinical oncology.
Researchers at the University of California San Diego have made a groundbreaking discovery about how our brains learn new information. Using sophisticated imaging techniques, they found that individual neurons follow multiple rules during learning, rather than one set of uniform rules as previously thought. This new understanding has s...
Researchers have created a breakthrough photonic chip that can train nonlinear neural networks using light, accelerating AI training while reducing energy use. The chip uses a special semiconductor material to reshape how light behaves, enabling reconfigurable systems with wide mathematical function expression.
Scientists have developed an all-optical activation function based on sound waves for photonic computing, enabling the creation of energy-efficient artificial intelligence systems. This breakthrough could potentially facilitate the scaling up of physical computing systems and pave the way for more efficient optical neural networks.
Scientists have built a digital twin of the mouse brain's visual cortex using AI, predicting neural activity and anatomical features. The model can generalize to new visual inputs and data, speeding up brain research and understanding intelligence.
The conference gathered international researchers to discuss AI's role in drug discovery and development, including generative AI strategies for designing chemical compounds. The speakers emphasized the significance of personalized medicine, where therapies will be tailored to each patient's unique molecular profile.
Researchers developed new AI models, InstaNovo and InstaNovo+, to vastly improve accuracy and discovery in protein science. These models excel in tasks such as de novo peptide sequencing, identifying microorganisms, and discovering novel peptides, with implications for personalized medicine, cancer immunology, and beyond.
Artificial neural networks trained on spontaneous retinal activity patterns show improved motion prediction in natural scenes. The approach also enhances performance when combined with naturalistic movie data.
Göttingen research team develops infomorphic neurons that learn independently and self-organize among neighboring neurons. This allows the smallest unit in the network to control its own learning, enabling novel machine learning approaches and a deeper understanding of brain function.
Researchers at National University of Singapore invent new computing cell that can mimic electronic neurons and synapses, reducing size by a factor of 18 and energy consumption. The discovery enables AI systems to process more information while using less energy.
A new brain-like computer uses analog computing to process and store information in the same location as biological neurons, reducing power consumption by 0.25%. The device, called a memristor network, is more efficient than conventional transistor-based computers and has implications for autonomous vehicles and drones.
Researchers developed an AI model that classifies variable stars from light curves with high accuracy, outperforming traditional approaches. The StarWhisper LightCurve series achieves near 90% accuracy with minimal manual intervention, paving the way for parallel data analysis and multi-modal AI applications in astronomy.
A new AI tool, NicheCompass, visualizes a cell's social network to help treat cancer. By analyzing millions of cells from patient samples, the tool predicts molecular changes and identifies potential targets for personalized treatments.
The collaboration aims to accelerate the development and commercialization of inait's innovative AI technology, using its unique digital brain AI platform. It will focus on joint product development, go-to-market strategies, and co-selling initiatives, initially targeting the finance and robotics sectors.
Researchers at Saarland University are developing leaner, customized AI models and techniques like knowledge distillation to reduce energy consumption. These smaller models enable small and medium-sized businesses to access powerful AI technology without a large technical infrastructure.
A new study suggests that artificial intelligence can effectively detect wildfires in the Amazon rainforest, using satellite imaging and deep learning. The technology achieved a 93% success rate in training models via datasets of images with and without wildfires.
Researchers at Technical University of Munich developed a new AI training method that significantly reduces energy consumption. The approach uses probabilities to determine parameters, making the training process 100 times faster while maintaining accuracy comparable to existing procedures.
This study utilized deep learning models to diagnose and predict the likelihood of malignant transformation in oral potentially malignant disorders. AI-driven approaches offer noninvasive, cost-effective, and objective means to enhance early detection and improve patient outcomes.
A new AI model measures how fast the brain ages by analyzing MRI scans, providing a more accurate picture of brain health. The tool closely correlates faster brain aging with increased cognitive decline and dementia risk, offering potential for early biomarkers and personalized treatment.
Research advances higher-order networks to capture multi-agent interactions, enabling accurate modeling of biological, social, and physical systems. The Dirac-Bianconi operator provides a powerful generalization of the graph Laplacian, encoding local and global interactions across different topological dimensions.
A new machine learning model, NAS-WD, has improved the accuracy of detecting 'woody breast' in chicken meat to 95%, allowing for better quality assurance and customer confidence. The model uses hyperspectral imaging to analyze complex data from images, enabling more accurate detection than traditional methods.
The research team successfully integrated miniaturized multilayer optical diffractive neural networks onto the distal end of MMFs, enabling full-optical image transmission. The system achieved exceptional performance in imaging handwritten digits and demonstrated high-quality optical image reconstruction.
A recent study emphasizes the urgent need to address bias in generative AI systems, which can distort outcomes and erode public trust. The research suggests that developing and deploying ethical, explainable AI is crucial to ensure fairness and transparency in critical decision-making areas.
Yann LeCun, NYU's Courant Institute of Mathematical Sciences professor, has been selected as a winner of the 2025 Queen Elizabeth Prize for Engineering for his groundbreaking research on artificial neural networks. His work enabled machines to process and learn from vast amounts of data in ways previously unimaginable.
Researchers developed MUNIS, a deep learning tool that predicts CD8+ T cell epitopes with high accuracy, potentially accelerating vaccine development. The tool was validated using experimental data from influenza, HIV, and EBV, demonstrating its potential to streamline vaccine design.
A recent study reveals that rats' visual recognition abilities are extremely efficient and adaptable, even outperforming advances in artificial intelligence. Rats employ more flexible image processing strategies than CNNs, which could inspire new approaches to AI model development.
Neuromorphic computing is poised to emerge into full-scale commercial use, driven by the need for energy-efficient solutions. The review article proposes strategies for building large-scale neuromorphic systems that can tackle complex real-world challenges.
Researchers propose several key features to optimize sparsity, massive parallelism, and hierarchical structure in neural representation for neuromorphic systems. The goal is to achieve energy efficiency and compactness while retaining information at high fidelity.
A new method has improved AI translation of sign language by adding data on hand and facial expressions, as well as skeletal information. This has led to a significant increase in accuracy, making it easier for people with hearing impairments to communicate.