AI News

RSS

Shaping the Future of Motion Detection: Rippled-Assisted Optoelectronic Arrays

Researchers propose a macroscale integration of rippled-assisted optoelectronic arrays (RAO) for motion detection and recognition, addressing limitations of existing technologies. The RAO array, fabricated with monolayer MoS2 and featuring a specially treated dielectric layer, demonstrates remarkable memory window uniformity, non-volatile photoconductance, and extensive optical storage dynamic range.

27 Feb 2024

Societal Impacts of AI: Toward Ethical Governance and Social Justice

Researchers from the University of Ostrava delve into the intricate landscape of AI's societal implications, emphasizing the need for ethical regulations and democratic values alignment. Through interdisciplinary analysis and policy evaluation, they advocate for transparent, participatory AI deployment, fostering societal welfare while addressing inequalities and safeguarding human rights.

27 Feb 2024

Computational Sensors for Advancing Soft Robotics

Researchers introduce programmable crack array within micro-crumples (PCAM) sensors, leveraging computational design for soft robots. These sensors exhibit robustness and tunability, enabling accurate trajectory prediction and terrain awareness in real-world scenarios, thus addressing key challenges in soft robotics automation. Integrated into an origami robot with machine intelligence, PCAM sensors signify a milestone in bridging predictive design complexities and practical implementation for enhanced soft robot performance in dynamic environments.

27 Feb 2024

A Web Mining and ML Approach to Enhance Learning Object Reusability

Researchers propose a novel approach combining web mining and machine learning (ML) techniques to classify learning objects effectively in e-learning systems, aiming to maximize their reusability. By employing advanced ML algorithms and web mining methods, the study demonstrates significant improvements in resource discovery and knowledge dissemination, ultimately enhancing the efficiency of e-learning environments.

27 Feb 2024

Dynamic Educational Recommendation System Using Deep Learning

Researchers present a hybrid recommendation system for virtual learning environments, employing bi-directional long short-term memory (BiLSTM) networks to capture users' evolving interests. Achieving remarkable accuracy and low loss, the system outperforms existing methods by integrating attention mechanisms and compression algorithms, offering personalized resource suggestions based on both short-term and long-term user behaviors.

25 Feb 2024

Pedipulate: Quadruped Robot Manipulation Using Legs

Swiss researchers introduce Pedipulate, a novel controller trained with deep reinforcement learning, enabling quadruped robots to manipulate objects using their legs. Demonstrating robustness and versatility, Pedipulate tracks foot target points, adapts stance, and handles external disturbances, showcasing potential applications in maintenance, home support, and exploration tasks.

25 Feb 2024

Understanding Patient Attitudes Toward Smart Digital Health Devices

A study in Digital Medicine explores ownership, usage, and willingness to share data from smart devices among Duke University Health System (DUHS) patients. Findings reveal widespread smartphone and wearable ownership, with usage focused on health tracking, while demographic variations influence data sharing preferences, highlighting the need for inclusive digital health strategies to address barriers like cost and privacy concerns.

25 Feb 2024

Optimizing Solar-Powered Hybrid Systems with Grasshopper Optimization Algorithm

Researchers propose an energy management scheme (EMS) using a rule-based grasshopper optimization algorithm (RB-GOA) to efficiently manage solar-powered battery-ultracapacitor (UC) hybrid systems, addressing output fluctuations and maximizing energy extraction. The RB-GOA EMS outperforms other swarm intelligence techniques in minimizing oscillations, reducing power surges, and ensuring stable output across varying load demands, demonstrating its effectiveness in optimizing solar energy utilization.

25 Feb 2024

A CatBoost Approach for Superconducting Material Prediction with AI

This study harnesses the CatBoost algorithm to predict transition temperatures (Tc) of superconducting materials, addressing challenges in dataset refinement and feature selection. Leveraging the Jabir and Soraya packages for generating atomic descriptors and selecting crucial features, the model achieved high accuracy with an R-squared (R2) of 0.952 and root mean square error (RMSE) of 6.45 K. Additionally, a novel web application for Tc prediction underscores the impactful synergy between AI and materials science.

23 Feb 2024

Preference Alignment Framework for Enhancing Multi-Modal Large Language Models

This research introduces a novel preference alignment framework to address performance degradation in multi-modal large language models (MLLMs) caused by visual instruction tuning. By leveraging preference data collected from a visual question answering dataset, the proposed method significantly improves the MLLM's instruction-following capabilities, surpassing the performance of the original language model on various benchmarks.

23 Feb 2024

VATr++: Advanced Few-Shot Styled Handwritten Text Generation

This study investigates the impact of visual and textual input on styled handwritten text generation (HTG) models, proposing strategies for input preparation and training regularization. The researchers extend the VATr architecture to VATr++, enhancing rare character generation and handwriting style capture. Additionally, they introduce a standardized evaluation protocol to facilitate fair comparisons and foster progress in the field of HTG.

23 Feb 2024

ML-based Prediction of Punching Shear Capacity in Ultra-High-Performance Concrete

Researchers investigated the feasibility of using machine learning (ML) models to predict the punching shear capacity of post-tensioned ultra-high-performance concrete (UHPC) flat slabs. By proposing correction factors based on finite element method-artificial intelligence (FEM-AI/ML) techniques, they extended the validity of punching shear capacity provisions in design codes like EC2 and ACI-318 to include PT-UHPC flat slabs.

23 Feb 2024

UK’s First Master’s Degree in Applying AI Tools to Engineering and Design is Open for Applications

The UK’s first master’s degree course focused on applying skills in AI to engineering and design is to begin this year at the University of Bath.

23 Feb 2024

Enhancing Few-Shot Semantic Segmentation with Transductive Meta-Learning

Researchers propose a novel approach for few-shot semantic segmentation, leveraging an ensemble of visual features learned from pre-trained classification and semantic segmentation networks. Their method utilizes a two-pass strategy, employing transductive meta-learning to improve prediction accuracy and mitigate false positives. Experimental results demonstrate significant performance improvements, achieving state-of-the-art results on benchmark datasets with minimal trainable parameters.

22 Feb 2024

ToolSword: Addressing Safety Concerns in Large Language Models for Tool Learning

Researchers introduced ToolSword, a framework designed to address safety challenges encountered by large language models (LLMs) during tool learning processes. By identifying and examining six safety scenarios spanning input, execution, and output stages, ToolSword comprehensively evaluated LLMs' ability to handle harmful queries, risky tools, and detrimental feedback. Experiments on 11 LLMs revealed persistent safety risks, emphasizing the urgent need for robust safety measures in practical applications of tool learning.

22 Feb 2024

Harnessing Machine Learning for Predicting School Dropout

Researchers utilized various machine learning algorithms to develop predictive models for identifying students at risk of dropping out of secondary and higher education in Mexico. Leveraging demographic, socioeconomic, and educational data, the study demonstrated the effectiveness of artificial neural networks (ANN) in achieving high reliability (99%) in predicting school dropout, highlighting key variables such as school attendance, type, location, occupation, income, and marital status.

22 Feb 2024

AutoGPT+P: Integrating Affordance-Based Scene Representation with LLMs for Robotic Task Planning

Researchers proposed AutoGPT+P, a novel system for robotic task planning that merges an affordance-based scene representation with large language models (LLMs). By combining visual data-based scene affordance extraction with LLMs' capabilities, AutoGPT+P enables robots to understand natural language commands, perceive their environment, and generate plans to fulfill user requests, even in scenarios where required objects are absent.

22 Feb 2024

LONGHEADS: Enhancing Large Language Models' Capability for Processing Long Contexts

Researchers introduced LONGHEADS, a training-free framework aimed at improving the effectiveness of large language models (LLMs) in handling long contexts. By strategically dividing input texts into chunks and allowing each attention head to focus on important segments, LONGHEADS addressed limitations in attention windows and computational demands, showcasing superior performance on various natural language processing tasks without additional training. The framework's query-aware chunk selection strategy and efficient utilization of attention heads demonstrated promise in enhancing LLMs' abilities for processing lengthy inputs.

21 Feb 2024

Ensemble Learning for Botnet Detection to Enhance IoT Security

Researchers proposed a novel intrusion detection system (IDS) leveraging ensemble learning and deep neural networks (DNNs) to combat botnet attacks on Internet of Things (IoT) devices. By training device-specific DNN models on heterogeneous IoT data and aggregating predictions through ensemble averaging, the system achieved remarkable accuracy and effectively detected botnet activities. The study's structured methodology, comprehensive evaluation metrics, and ensemble approach offer promise in bolstering IoT security against evolving cyber threats.

21 Feb 2024

A Novel Approach for Edge Unlearning in Graph Neural Networks

Researchers introduced Unlink to Unlearn (UtU), a novel method designed to simplify edge unlearning in graph neural networks (GNNs) and address the issue of over-forgetting. UtU facilitates unlearning by unlinking forget edges from the original graph structure, effectively preserving privacy protection capabilities and link prediction accuracy while minimizing computational overhead. Experimental evaluation across real-world datasets demonstrated UtU's efficacy in unlearning, maintaining high accuracy in downstream tasks, and effectively preventing over-forgetting, making it a practical and lightweight solution for edge unlearning in GNNs.

21 Feb 2024