Previous Issue
Volume 6, July
 
 

AI, Volume 6, Issue 8 (August 2025) – 24 articles

  • Issues are regarded as officially published after their release is announced to the table of contents alert mailing list.
  • You may sign up for e-mail alerts to receive table of contents of newly released issues.
  • PDF is the official format for papers published in both, html and pdf forms. To view the papers in pdf format, click on the "PDF Full-text" link, and use the free Adobe Reader to open them.
Order results
Result details
Section
Select all
Export citation of selected articles as:
31 pages, 18843 KiB  
Article
Liquid Adaptive AI: A Theoretical Framework for Continuously Self-Improving Artificial Intelligence
by Thomas R. Caulfield, Naeyma N. Islam and Rohit Chitale
AI 2025, 6(8), 186; https://doi.org/10.3390/ai6080186 - 14 Aug 2025
Abstract
We present Liquid Adaptive AI as a theoretical framework and mathematical basis for artificial intelligence systems capable of continuous structural adaptation and autonomous capability development. This work explores the conceptual boundaries of adaptive AI by formalizing three interconnected mechanisms: (1) entropy-guided hyperdimensional knowledge [...] Read more.
We present Liquid Adaptive AI as a theoretical framework and mathematical basis for artificial intelligence systems capable of continuous structural adaptation and autonomous capability development. This work explores the conceptual boundaries of adaptive AI by formalizing three interconnected mechanisms: (1) entropy-guided hyperdimensional knowledge graphs that could autonomously restructure based on information-theoretic criteria; (2) a self-development engine using hierarchical Bayesian optimization for runtime architecture modification; and (3) a federated multi-agent framework with emergent specialization through distributed reinforcement learning. We address fundamental limitations in current AI systems through mathematically formalized processes of dynamic parameter adjustment, structural self-modification, and cross-domain knowledge synthesis, while immediate implementation faces substantial computational challenges requiring infrastructure on the scale of current large language model training facilities, we provide architectural specifications, theoretical convergence bounds, and evaluation criteria as a foundation for future research. This theoretical exploration establishes mathematical foundations for a potential new paradigm in artificial intelligence that would transition from episodic training to persistent autonomous development, offering a long-term research direction for the field. A comprehensive Supplementary Materials document provides detailed technical analysis, computational requirements, and an incremental development roadmap spanning approximately a decade. Full article
Show Figures

Figure 1

20 pages, 1527 KiB  
Article
Trends in Patent Applications for Technologies in the Automotive Industry: Applications of Deep Learning and Machine Learning
by ChoongChae Woo and Junbum Park
AI 2025, 6(8), 185; https://doi.org/10.3390/ai6080185 - 13 Aug 2025
Abstract
This study investigates global innovation trends in machine learning (ML) and deep learning (DL) technologies within the automotive sector through a patent analysis of 5314 applications filed between 2005 and 2022 across the five major patent offices (IP5). Using Cooperative Patent Classification (CPC) [...] Read more.
This study investigates global innovation trends in machine learning (ML) and deep learning (DL) technologies within the automotive sector through a patent analysis of 5314 applications filed between 2005 and 2022 across the five major patent offices (IP5). Using Cooperative Patent Classification (CPC) codes and keyword analysis, we identify seven sub-technology domains and examine both geographical and corporate patenting strategies. Our findings show that the United States dominates in overall filings, while Japan demonstrates a notably high share of triadic patents, which reflects a strong global-reach strategy. Patent activity is heavily concentrated in vehicle control and infrastructure traffic control, with emerging growth observed in battery management and occupant analytics. In contrast, security-related technologies remain underrepresented, indicating a potential blind spot in current innovation efforts. Corporate strategies diverge markedly; for example, some firms, such as Toyota and Bosch, pursue balanced tri-regional protection, whereas others, including Ford and GM, focus on dual-market coverage in the United States and China. These patterns illustrate how market priorities, regulatory environments, and technological objectives influence patenting behavior. By mapping the technological and strategic landscape of ML/DL innovation in the automotive industry, this study provides actionable insights for industry practitioners seeking to optimize intellectual property portfolios and for policymakers aiming to address gaps such as automotive cybersecurity in future R&D agendas. Full article
Show Figures

Figure 1

22 pages, 3593 KiB  
Article
Exploring Artificial Personality Grouping Through Decision Making in Feature Spaces
by Yuan Zhou and Siamak Khatibi
AI 2025, 6(8), 184; https://doi.org/10.3390/ai6080184 - 11 Aug 2025
Viewed by 177
Abstract
Human personality (HP) is seen as an individual’s consistent patterns of feeling, thinking, and behaving by today’s psychological studies, in which HPs are characterized in terms of traits—in particular, as relatively enduring characteristics that influence human behavior across many situations. In this sense, [...] Read more.
Human personality (HP) is seen as an individual’s consistent patterns of feeling, thinking, and behaving by today’s psychological studies, in which HPs are characterized in terms of traits—in particular, as relatively enduring characteristics that influence human behavior across many situations. In this sense, more generally, artificial personality (AP) is studied in computer science to develop AI agents who should behave more like humans. However, in this paper, we suggest another approach by which the APs of individual agents are distinguishable based on their behavioral characteristics in achieving tasks and not necessarily in their human-like performance. As an initial step toward AP, we propose an approach to extract human decision-making characteristics as a generative resource for encoding the variability in agent personality. Using an application example, we demonstrate the feasibility of grouping APs, divided into several steps consisting of (1) defining a feature space to measure the commonality of decision making between individual and a group of people; (2) grouping APs by using multidimensional orthogonal features in the feature space to guarantee inter-individual differences between APs in achieving for the same task; and (3) evaluating the consistency of grouping APs by performing a cluster-stability analysis. Finally, our thoughts for the future implementation of APs are discussed and presented. Full article
28 pages, 4548 KiB  
Article
A Deep Reinforcement Learning Framework for Strategic Indian NIFTY 50 Index Trading
by Raj Gaurav Mishra, Dharmendra Sharma, Mahipal Gadhavi, Sangeeta Pant and Anuj Kumar
AI 2025, 6(8), 183; https://doi.org/10.3390/ai6080183 - 11 Aug 2025
Viewed by 288
Abstract
This paper presents a comprehensive deep reinforcement learning (DRL) framework for developing strategic trading models tailored to the Indian NIFTY 50 index, leveraging the temporal and nonlinear nature of financial markets. Three advanced DRL architectures deep Q-network (DQN), double deep Q-network (DDQN), and [...] Read more.
This paper presents a comprehensive deep reinforcement learning (DRL) framework for developing strategic trading models tailored to the Indian NIFTY 50 index, leveraging the temporal and nonlinear nature of financial markets. Three advanced DRL architectures deep Q-network (DQN), double deep Q-network (DDQN), and dueling double deep Q-network (Dueling DDQN) were implemented and empirically evaluated. Using a decade-long dataset of 15-min interval OHLC data enriched with technical indicators such as the exponential moving average (EMA), pivot points, and multiple supertrend configurations, the models were trained using prioritized experience replay, epsilon-greedy exploration strategies, and softmax sampling mechanisms. A test set comprising one year of unseen data (May 2024–April 2025) was used to assess generalization performance across key financial metrics, including Sharpe ratio, profit factor, win rate, and trade frequency. Each architecture was analyzed in three progressively sophisticated variants incorporating enhancements in reward shaping, exploration–exploitation balancing, and penalty-based trade constraints. DDQN V3 achieved a Sharpe ratio of 0.7394, a 73.33% win rate, and a 16.58 profit factor across 15 trades, indicating strong volatility-adjusted suitability for real-world deployment. In contrast, the Dueling DDQN V3 achieved a high Sharpe ratio of 1.2278 and a 100% win rate but with only three trades, indicating an excessive conservatism. The DQN V1 model served as a strong baseline, outperforming passive strategies but exhibiting limitations due to Q-value overestimation. The novelty of this work lies in its systematic exploration of DRL variants integrated with enhanced exploration mechanisms and reward–penalty structures, rigorously applied to high-frequency trading on the NIFTY 50 index within an emerging market context. Our findings underscore the critical importance of architectural refinements, dynamic exploration strategies, and trade regularization in stabilizing learning and enhancing profitability in DRL-based intelligent trading systems. Full article
(This article belongs to the Special Issue AI in Finance: Leveraging AI to Transform Financial Services)
Show Figures

Figure 1

30 pages, 2591 KiB  
Article
Prompt Optimization with Two Gradients for Classification in Large Language Models
by Anthony Jethro Lieander, Hui Wang and Karen Rafferty
AI 2025, 6(8), 182; https://doi.org/10.3390/ai6080182 - 8 Aug 2025
Viewed by 518
Abstract
Large language models (LLMs) generally perform well in common tasks, yet are often susceptible to errors in sophisticated natural language processing (NLP) on classification applications. Prompt engineering has emerged as a strategy to enhance their performance. Despite the effort required for manual prompt [...] Read more.
Large language models (LLMs) generally perform well in common tasks, yet are often susceptible to errors in sophisticated natural language processing (NLP) on classification applications. Prompt engineering has emerged as a strategy to enhance their performance. Despite the effort required for manual prompt optimization, recent advancements highlight the need for automation to reduce human involvement. We introduced the PO2G (prompt optimization with two gradients) framework to improve the efficiency of optimizing prompts for classification tasks. PO2G demonstrates improvement in efficiency, reaching almost 89% accuracy after just three iterations, whereas ProTeGi requires six iterations to achieve a comparable level. We evaluated PO2G and ProTeGi on a benchmark of nine NLP tasks, three tasks from the original ProTeGi study, and six non-domain-specific tasks. We also evaluated both frameworks on seven legal-domain classification tasks. These results provide broader insights into the efficiency and effectiveness of prompt optimization frameworks for classification across diverse NLP scenarios. Full article
Show Figures

Figure 1

26 pages, 3980 KiB  
Article
Optimization and Performance Comparison of AOD-Net and DehazeFormer Dehazing Algorithms
by Futing Liu, Jingtao Wang and Yun Pan
AI 2025, 6(8), 181; https://doi.org/10.3390/ai6080181 - 7 Aug 2025
Viewed by 266
Abstract
Image dehazing is an effective approach for enhancing the quality of images captured under foggy or hazy conditions. Although existing methods have achieved certain success in dehazing performance, many rely on deep network architectures, leading to high model complexity and computational costs. To [...] Read more.
Image dehazing is an effective approach for enhancing the quality of images captured under foggy or hazy conditions. Although existing methods have achieved certain success in dehazing performance, many rely on deep network architectures, leading to high model complexity and computational costs. To address this issue, this study aims to compare and optimize existing algorithms to improve dehazing performance. For this purpose, we innovatively propose a multi-scale feature-coordinated composite loss mechanism, integrating perceptual loss, Mean Squared Error, and L1 regularization to optimize two dehazing methods: AOD-Net and DehazeFormer. Extensive experiments demonstrate significant performance improvements under the multi-objective loss mechanism. For AOD-Net, the PSNR increased by 22.40% (+4.17 dB), SSIM by 3.62% (+0.0318), VSNR by 43% (+1.54 dB), and LPIPS decreased by 56.30% (−0.1161). Similarly, DehazeFormer showed notable enhancements: the PSNR improved by 11.43% (+2.45 dB), SSIM by 0.8% (+0.008), VSNR by 2.6% (+0.23 dB), and LPIPS decreased by 5.5% (−0.0104). These results fully validate the effectiveness of the composite loss mechanism in enhancing the feature representation capability of the models. Full article
Show Figures

Figure 1

18 pages, 4529 KiB  
Article
LGSIK-Poser: Skeleton-Aware Full-Body Motion Reconstruction from Sparse Inputs
by Linhai Li, Jiayi Lin and Wenhui Zhang
AI 2025, 6(8), 180; https://doi.org/10.3390/ai6080180 - 7 Aug 2025
Viewed by 222
Abstract
Accurate full-body motion reconstruction from sparse sensors is crucial for VR/AR applications but remains challenging due to the under-constrained nature of limited observations and the computational constraints of mobile platforms. This paper presents LGSIK-Poser, a unified and lightweight framework that supports real-time motion [...] Read more.
Accurate full-body motion reconstruction from sparse sensors is crucial for VR/AR applications but remains challenging due to the under-constrained nature of limited observations and the computational constraints of mobile platforms. This paper presents LGSIK-Poser, a unified and lightweight framework that supports real-time motion reconstruction from heterogeneous sensor configurations, including head-mounted displays, handheld controllers, and up to three optional inertial measurement units, without requiring reconfiguration across scenarios. The model integrates temporally grouped LSTM modeling, anatomically structured graph-based reasoning, and region-specific inverse kinematics refinement to enhance end-effector accuracy and structural consistency. Personalized body shape is estimated using user-specific anthropometric priors within the SMPL model, a widely adopted parametric representation of human shape and pose. Experiments on the AMASS benchmark demonstrate that LGSIK-Poser achieves state-of-the-art accuracy with up to 48% improvement in hand localization, while reducing model size by 60% and latency by 22% compared to HMD-Poser. The system runs at 63.65 FPS with only 3.74 M parameters, highlighting its suitability for real-time immersive applications. Full article
Show Figures

Figure 1

22 pages, 7990 KiB  
Article
Detection of Cracks in Low-Power Wind Turbines Using Vibration Signal Analysis with Empirical Mode Decomposition and Convolutional Neural Networks
by Angel H. Rangel-Rodriguez, Jose M. Machorro-Lopez, David Granados-Lieberman, J. Jesus de Santiago-Perez, Juan P. Amezquita-Sanchez and Martin Valtierra-Rodriguez
AI 2025, 6(8), 179; https://doi.org/10.3390/ai6080179 - 6 Aug 2025
Viewed by 400
Abstract
Condition monitoring and fault detection in wind turbines are essential for reducing repair and maintenance costs. Early detection of faults enables timely interventions before the damage worsens. However, existing methods often rely on costly scheduled inspections or lack the ability to effectively detect [...] Read more.
Condition monitoring and fault detection in wind turbines are essential for reducing repair and maintenance costs. Early detection of faults enables timely interventions before the damage worsens. However, existing methods often rely on costly scheduled inspections or lack the ability to effectively detect early stage damage, particularly under different operational speeds. This article presents a methodology based on convolutional neural networks (CNNs) and empirical mode decomposition (EMD) of vibration signals for the detection of blade crack damage. The proposed approach involves acquiring vibration signals under four conditions: healthy, light, intermediate, and severe damage. EMD is then applied to extract time–frequency representations of the signals, which are subsequently converted into images. These images are analyzed by a CNN to classify the condition of the wind turbine blades. To enhance the final CNN architecture, various image sizes and configuration parameters are evaluated to balance computational load and classification accuracy. The results demonstrate that combining vibration signal images, generated using the EMD method, with CNN models enables accurate classification of blade conditions, achieving 99.5% accuracy while maintaining a favorable trade-off between performance and complexity. Full article
Show Figures

Figure 1

25 pages, 5488 KiB  
Article
Biased by Design? Evaluating Bias and Behavioral Diversity in LLM Annotation of Real-World and Synthetic Hotel Reviews
by Maria C. Voutsa, Nicolas Tsapatsoulis and Constantinos Djouvas
AI 2025, 6(8), 178; https://doi.org/10.3390/ai6080178 - 4 Aug 2025
Viewed by 514
Abstract
As large language models (LLMs) gain traction among researchers and practitioners, particularly in digital marketing for tasks such as customer feedback analysis and automated communication, concerns remain about the reliability and consistency of their outputs. This study investigates annotation bias in LLMs by [...] Read more.
As large language models (LLMs) gain traction among researchers and practitioners, particularly in digital marketing for tasks such as customer feedback analysis and automated communication, concerns remain about the reliability and consistency of their outputs. This study investigates annotation bias in LLMs by comparing human and AI-generated annotation labels across sentiment, topic, and aspect dimensions in hotel booking reviews. Using the HRAST dataset, which includes 23,114 real user-generated review sentences and a synthetically generated corpus of 2000 LLM-authored sentences, we evaluate inter-annotator agreement between a human expert and three LLMs (ChatGPT-3.5, ChatGPT-4, and ChatGPT-4-mini) as a proxy for assessing annotation bias. Our findings show high agreement among LLMs, especially on synthetic data, but only moderate to fair alignment with human annotations, particularly in sentiment and aspect-based sentiment analysis. LLMs display a pronounced neutrality bias, often defaulting to neutral sentiment in ambiguous cases. Moreover, annotation behavior varies notably with task design, as manual, one-to-one prompting produces higher agreement with human labels than automated batch processing. The study identifies three distinct AI biases—repetition bias, behavioral bias, and neutrality bias—that shape annotation outcomes. These findings highlight how dataset complexity and annotation mode influence LLM behavior, offering important theoretical, methodological, and practical implications for AI-assisted annotation and synthetic content generation. Full article
(This article belongs to the Special Issue AI Bias in the Media and Beyond)
Show Figures

Figure 1

24 pages, 1855 KiB  
Article
AI-Driven Panel Assignment Optimization via Document Similarity and Natural Language Processing
by Rohit Ramachandran, Urjit Patil, Srinivasaraghavan Sundar, Prem Shah and Preethi Ramesh
AI 2025, 6(8), 177; https://doi.org/10.3390/ai6080177 - 1 Aug 2025
Viewed by 438
Abstract
Efficient and accurate panel assignment is critical in expert and peer review processes. Traditional methods—based on manual preferences or Heuristic rules—often introduce bias, inconsistency, and scalability challenges. We present an automated framework that combines transformer-based document similarity modeling with optimization-based reviewer assignment. Using [...] Read more.
Efficient and accurate panel assignment is critical in expert and peer review processes. Traditional methods—based on manual preferences or Heuristic rules—often introduce bias, inconsistency, and scalability challenges. We present an automated framework that combines transformer-based document similarity modeling with optimization-based reviewer assignment. Using the all-mpnet-base-v2 from model (version 3.4.1), our system computes semantic similarity between proposal texts and reviewer documents, including CVs and Google Scholar profiles, without requiring manual input from reviewers. These similarity scores are then converted into rankings and integrated into an Integer Linear Programming (ILP) formulation that accounts for workload balance, conflicts of interest, and role-specific reviewer assignments (lead, scribe, reviewer). The method was tested across 40 researchers in two distinct disciplines (Chemical Engineering and Philosophy), each with 10 proposal documents. Results showed high self-similarity scores (0.65–0.89), strong differentiation between unrelated fields (−0.21 to 0.08), and comparable performance between reviewer document types. The optimization consistently prioritized top matches while maintaining feasibility under assignment constraints. By eliminating the need for subjective preferences and leveraging deep semantic analysis, our framework offers a scalable, fair, and efficient alternative to manual or Heuristic assignment processes. This approach can support large-scale review workflows while enhancing transparency and alignment with reviewer expertise. Full article
(This article belongs to the Section AI Systems: Theory and Applications)
Show Figures

Figure 1

25 pages, 1518 KiB  
Article
Command Redefined: Neural-Adaptive Leadership in the Age of Autonomous Intelligence
by Raul Ionuț Riti, Claudiu Ioan Abrudan, Laura Bacali and Nicolae Bâlc
AI 2025, 6(8), 176; https://doi.org/10.3390/ai6080176 - 1 Aug 2025
Viewed by 336
Abstract
Artificial intelligence has taken a seat at the executive table and is threatening the fact that human beings are the only ones who should be in a position of power. This article gives conjectures on the future of leadership in which managers will [...] Read more.
Artificial intelligence has taken a seat at the executive table and is threatening the fact that human beings are the only ones who should be in a position of power. This article gives conjectures on the future of leadership in which managers will collaborate with learning algorithms in the Neural Adaptive Artificial Intelligence Leadership Model, which is informed by the transformational literature on leadership and socio-technical systems, as well as the literature on algorithmic governance. We assessed the model with thirty in-depth interviews, system-level traces of behavior, and a verified survey, and we explored six hypotheses that relate to algorithmic delegation and ethical oversight, as well as human judgment versus machine insight in terms of agility and performance. We discovered that decisions are made quicker, change is more effective, and interaction is more vivid where agile practices and good digital understanding exist, and statistical tests propose that human flexibility and definite governance augment those benefits as well. It is single-industry research that contains self-reported measures, which causes research to be limited to other industries that contain more objective measures. Practitioners are provided with a practical playbook on how to make algorithmic jobs meaningful, introduce moral fail-safes, and build learning feedback to ensure people and machines are kept in line. Socially, the practice is capable of minimizing bias and establishing inclusion by visualizing accountability in the code and practice. Filling the gap between the theory of leadership and the reality of algorithms, the study provides a model of intelligent systems leading in organizations that can be reproduced. Full article
(This article belongs to the Section AI Systems: Theory and Applications)
Show Figures

Figure 1

23 pages, 752 KiB  
Perspective
Quantum Artificial Intelligence: Some Strategies and Perspectives
by Marco Baioletti, Fabrizio Fagiolo, Corrado Loglisci, Vito Nicola Losavio, Angelo Oddi, Riccardo Rasconi and Pier Luigi Gentili
AI 2025, 6(8), 175; https://doi.org/10.3390/ai6080175 - 1 Aug 2025
Viewed by 682
Abstract
In the twenty-first century, humanity is compelled to face global challenges. Such challenges involve complex systems. However, science has some cognitive and predictive limits in dealing with complex systems. Some of these limits are related to computational complexity and the recognition of variable [...] Read more.
In the twenty-first century, humanity is compelled to face global challenges. Such challenges involve complex systems. However, science has some cognitive and predictive limits in dealing with complex systems. Some of these limits are related to computational complexity and the recognition of variable patterns. To overcome these limits, artificial intelligence (AI) and quantum computing (QC) appear to be helpful. Even more promising is quantum AI (QAI), which emerged from the combination of AI and QC. The combination of AI and QC produces reciprocal, synergistic effects. This work describes some of these effects. It shows that QC offers new materials for implementing AI and innovative algorithms for solving optimisation problems and enhancing machine learning algorithms. Additionally, it demonstrates how AI algorithms can help overcome many of the experimental challenges associated with implementing QC. It also outlines several perspectives for the future development of quantum artificial intelligence. Full article
(This article belongs to the Topic Recent Advances in Chemical Artificial Intelligence)
Show Figures

Graphical abstract

29 pages, 1119 KiB  
Systematic Review
Phishing Attacks in the Age of Generative Artificial Intelligence: A Systematic Review of Human Factors
by Raja Jabir, John Le and Chau Nguyen
AI 2025, 6(8), 174; https://doi.org/10.3390/ai6080174 - 31 Jul 2025
Viewed by 817
Abstract
Despite the focus on improving cybersecurity awareness, the number of cyberattacks has increased significantly, leading to huge financial losses, with their risks spreading throughout the world. This is due to the techniques deployed in cyberattacks that mainly aim at exploiting humans, the weakest [...] Read more.
Despite the focus on improving cybersecurity awareness, the number of cyberattacks has increased significantly, leading to huge financial losses, with their risks spreading throughout the world. This is due to the techniques deployed in cyberattacks that mainly aim at exploiting humans, the weakest link in any defence system. The existing literature on human factors in phishing attacks is limited and does not live up to the witnessed advances in phishing attacks, which have become exponentially more dangerous with the introduction of generative artificial intelligence (GenAI). This paper studies the implications of AI advancement, specifically the exploitation of GenAI and human factors in phishing attacks. We conduct a systematic literature review to study different human factors exploited in phishing attacks, potential solutions and preventive measures, and the complexity introduced by GenAI-driven phishing attacks. This paper aims to address the gap in the research by providing a deeper understanding of the evolving landscape of phishing attacks with the application of GenAI and associated human implications, thereby contributing to the field of knowledge to defend against phishing attacks by creating secure digital interactions. Full article
Show Figures

Figure 1

21 pages, 4147 KiB  
Article
OLTEM: Lumped Thermal and Deep Neural Model for PMSM Temperature
by Yuzhong Sheng, Xin Liu, Qi Chen, Zhenghao Zhu, Chuangxin Huang and Qiuliang Wang
AI 2025, 6(8), 173; https://doi.org/10.3390/ai6080173 - 31 Jul 2025
Viewed by 379
Abstract
Background and Objective: Temperature management is key for reliable operation of permanent magnet synchronous motors (PMSMs). The lumped-parameter thermal network (LPTN) is fast and interpretable but struggles with nonlinear behavior under high power density. We propose OLTEM, a physics-informed deep model that combines [...] Read more.
Background and Objective: Temperature management is key for reliable operation of permanent magnet synchronous motors (PMSMs). The lumped-parameter thermal network (LPTN) is fast and interpretable but struggles with nonlinear behavior under high power density. We propose OLTEM, a physics-informed deep model that combines LPTN with a thermal neural network (TNN) to improve prediction accuracy while keeping physical meaning. Methods: OLTEM embeds LPTN into a recurrent state-space formulation and learns three parameter sets: thermal conductance, inverse thermal capacitance, and power loss. Two additions are introduced: (i) a state-conditioned squeeze-and-excitation (SC-SE) attention that adapts feature weights using the current temperature state, and (ii) an enhanced power-loss sub-network that uses a deep MLP with SC-SE and non-negativity constraints. The model is trained and evaluated on the public Electric Motor Temperature dataset (Paderborn University/Kaggle). Performance is measured by mean squared error (MSE) and maximum absolute error across permanent-magnet, stator-yoke, stator-tooth, and stator-winding temperatures. Results: OLTEM tracks fast thermal transients and yields lower MSE than both the baseline TNN and a CNN–RNN model for all four components. On a held-out generalization set, MSE remains below 4.0 °C2 and the maximum absolute error is about 4.3–8.2 °C. Ablation shows that removing either SC-SE or the enhanced power-loss module degrades accuracy, confirming their complementary roles. Conclusions: By combining physics with learned attention and loss modeling, OLTEM improves PMSM temperature prediction while preserving interpretability. This approach can support motor thermal design and control; future work will study transfer to other machines and further reduce short-term errors during abrupt operating changes. Full article
Show Figures

Figure 1

33 pages, 14330 KiB  
Article
Noisy Ultrasound Kidney Image Classifications Using Deep Learning Ensembles and Grad-CAM Analysis
by Walid Obaid, Abir Hussain, Tamer Rabie and Wathiq Mansoor
AI 2025, 6(8), 172; https://doi.org/10.3390/ai6080172 - 31 Jul 2025
Viewed by 463
Abstract
Objectives: This study introduces an automated classification system for noisy kidney ultrasound images using an ensemble of deep neural networks (DNNs) with transfer learning. Methods: The method was tested using a dataset with two categories: normal kidney images and kidney images with stones. [...] Read more.
Objectives: This study introduces an automated classification system for noisy kidney ultrasound images using an ensemble of deep neural networks (DNNs) with transfer learning. Methods: The method was tested using a dataset with two categories: normal kidney images and kidney images with stones. The dataset contains 1821 normal kidney images and 2592 kidney images with stones. Noisy images involve various types of noises, including salt and pepper noise, speckle noise, Poisson noise, and Gaussian noise. The ensemble-based method is benchmarked with state-of-the-art techniques and evaluated on ultrasound images with varying quality and noise levels. Results: Our proposed method demonstrated a maximum classification accuracy of 99.43% on high-quality images (the original dataset images) and 99.21% on the dataset images with added noise. Conclusions: The experimental results confirm that the ensemble of DNNs accurately classifies most images, achieving a high classification performance compared to conventional and individual DNN-based methods. Additionally, our method outperforms the highest-achieving method by more than 1% in accuracy. Furthermore, our analysis using Gradient-weighted Class Activation Mapping indicated that our proposed deep learning model is capable of prediction using clinically relevant features. Full article
(This article belongs to the Section Medical & Healthcare AI)
Show Figures

Figure 1

21 pages, 563 KiB  
Article
Optimized Interdisciplinary Research Team Formation Using a Genetic Algorithm and Publication Metadata Records
by Christian-Daniel Curiac, Mihai Micea, Traian-Radu Plosca, Daniel-Ioan Curiac and Alex Doboli
AI 2025, 6(8), 171; https://doi.org/10.3390/ai6080171 - 30 Jul 2025
Viewed by 482
Abstract
Forming interdisciplinary research teams is challenging, especially when the pool of candidates is large and/or the addressed research projects require multi-disciplinary expertise. Based on their previous research outputs, like published work, a data-driven team formation procedure selects the researchers that are likely to [...] Read more.
Forming interdisciplinary research teams is challenging, especially when the pool of candidates is large and/or the addressed research projects require multi-disciplinary expertise. Based on their previous research outputs, like published work, a data-driven team formation procedure selects the researchers that are likely to work well together while covering all areas and offering all skills required by the multi-disciplinary topic. The description of the research team formation problem proposed in this paper uses novel quantitative metrics about the team candidates computed from bibliographic metadata records. The proposed methodology first analyzes the metadata fields that provide useful information and then computes four synthetic indicators regarding candidates’ skills and their interpersonal traits. Interdisciplinary teams are formed by solving a complex combinatorial multi-objective weighted set cover optimization problem, defined as equations involving the synthetic indicators. Problem solving uses the NSGA-II genetic algorithm. The proposed methodology is validated and compared with other similar approaches using a dataset on researchers from Politehnica University of Timisoara extracted from the IEEE Xplore database. Experimental results show that the method can identify potential research teams in situations for which other related algorithms fail. Full article
Show Figures

Figure 1

16 pages, 2431 KiB  
Article
AppHerb: Language Model for Recommending Traditional Thai Medicine
by Thanawat Piyasawetkul, Suppachai Tiyaworanant and Tarapong Srisongkram
AI 2025, 6(8), 170; https://doi.org/10.3390/ai6080170 - 29 Jul 2025
Viewed by 627
Abstract
Trust in Traditional Thai Medicine (TTM) among Thai people has been reduced due to a lack of objective standards and the susceptibility of the general population to false information. The emergence of generative artificial intelligence (Gen AI) has significantly impacted various industries, including [...] Read more.
Trust in Traditional Thai Medicine (TTM) among Thai people has been reduced due to a lack of objective standards and the susceptibility of the general population to false information. The emergence of generative artificial intelligence (Gen AI) has significantly impacted various industries, including traditional medicine. However, previous Gen AI models have primarily focused on prescription generation based on Traditional Chinese Medicine (TCM), leaving TTM unexplored. To address this gap, we propose a novel fast-learning fine-tuned language model fortified with TTM knowledge. We utilized textual data from two TTM textbooks, Wat Ratcha-orasaram Ratchaworawihan (WRO), and Tamra Osot Phra Narai (NR), to fine-tune Unsloth’s Gemma-2 with 9 billion parameters. We developed two specialized TTM tasks: treatment prediction (TrP) and herbal recipe generation (HRG). The TrP and HRG models achieved precision, recall, and F1 scores of 26.54%, 28.14%, and 24.00%, and 32.51%, 24.42%, and 24.84%, respectively. Performance evaluation against TCM-based generative models showed comparable precision, recall, and F1 results with a smaller knowledge corpus. We further addressed the challenges of utilizing Thai, a low-resource and linguistically complex language. Unlike English or Chinese, Thai lacks explicit sentence boundary markers and employs an abugida writing system without spaces between words, complicating text segmentation and generation. These characteristics pose significant difficulties for machine understanding and limit model accuracy. Despite these obstacles, our work establishes a foundation for further development of AI-assisted TTM applications and highlights both the opportunities and challenges in applying language models to traditional medicine knowledge systems in Thai language contexts. Full article
(This article belongs to the Section Medical & Healthcare AI)
Show Figures

Graphical abstract

33 pages, 906 KiB  
Article
Scratching the Surface of Responsible AI in Financial Services: A Qualitative Study on Non-Technical Challenges and the Role of Corporate Digital Responsibility
by Antonis Skouloudis and Archana Venkatraman
AI 2025, 6(8), 169; https://doi.org/10.3390/ai6080169 - 28 Jul 2025
Viewed by 672
Abstract
Artificial Intelligence (AI) and Generative AI are transformative yet double-edged technologies with evolving risks. While research emphasises trustworthy, fair, and responsible AI by focusing on its “what” and “why,” it overlooks practical “how.” To bridge this gap in financial services, an industry at [...] Read more.
Artificial Intelligence (AI) and Generative AI are transformative yet double-edged technologies with evolving risks. While research emphasises trustworthy, fair, and responsible AI by focusing on its “what” and “why,” it overlooks practical “how.” To bridge this gap in financial services, an industry at the forefront of AI adoption, this study employs a qualitative approach grounded in existing Responsible AI and Corporate Digital Responsibility (CDR) frameworks. Through thematic analysis of 15 semi-structured interviews conducted with professionals working in finance, we illuminate nine non-technical barriers that practitioners face, such as sustainability challenges, trade-off balancing, stakeholder management, and human interaction, noting that GenAI concerns now eclipse general AI issues. CDR practitioners adopt a more human-centric stance, emphasising consensus-building and “no margin for error.” Our findings offer actionable guidance for more responsible AI strategies and enrich academic debates on Responsible AI and AI-CDR symbiosis. Full article
Show Figures

Graphical abstract

42 pages, 2224 KiB  
Article
Combined Dataset System Based on a Hybrid PCA–Transformer Model for Effective Intrusion Detection Systems
by Hesham Kamal and Maggie Mashaly
AI 2025, 6(8), 168; https://doi.org/10.3390/ai6080168 - 24 Jul 2025
Cited by 1 | Viewed by 666
Abstract
With the growing number and diversity of network attacks, traditional security measures such as firewalls and data encryption are no longer sufficient to ensure robust network protection. As a result, intrusion detection systems (IDSs) have become a vital component in defending against evolving [...] Read more.
With the growing number and diversity of network attacks, traditional security measures such as firewalls and data encryption are no longer sufficient to ensure robust network protection. As a result, intrusion detection systems (IDSs) have become a vital component in defending against evolving cyber threats. Although many modern IDS solutions employ machine learning techniques, they often suffer from low detection rates and depend heavily on manual feature engineering. Furthermore, most IDS models are designed to identify only a limited set of attack types, which restricts their effectiveness in practical scenarios where a network may be exposed to a wide array of threats. To overcome these limitations, we propose a novel approach to IDSs by implementing a combined dataset framework based on an enhanced hybrid principal component analysis–Transformer (PCA–Transformer) model, capable of detecting 21 unique classes, comprising 1 benign class and 20 distinct attack types across multiple datasets. The proposed architecture incorporates enhanced preprocessing and feature engineering, followed by the vertical concatenation of the CSE-CIC-IDS2018 and CICIDS2017 datasets. In this design, the PCA component is responsible for feature extraction and dimensionality reduction, while the Transformer component handles the classification task. Class imbalance was addressed using class weights, adaptive synthetic sampling (ADASYN), and edited nearest neighbors (ENN). Experimental results show that the model achieves 99.80% accuracy for binary classification and 99.28% for multi-class classification on the combined dataset (CSE-CIC-IDS2018 and CICIDS2017), 99.66% accuracy for binary classification and 99.59% for multi-class classification on the CSE-CIC-IDS2018 dataset, 99.75% accuracy for binary classification and 99.51% for multi-class classification on the CICIDS2017 dataset, and 99.98% accuracy for binary classification and 98.01% for multi-class classification on the NF-BoT-IoT-v2 dataset, significantly outperforming existing approaches by distinguishing a wide range of classes, including benign and various attack types, within a unified detection framework. Full article
Show Figures

Figure 1

21 pages, 4369 KiB  
Article
Breast Cancer Classification via a High-Precision Hybrid IGWO–SOA Optimized Deep Learning Framework
by Aniruddha Deka, Debashis Dev Misra, Anindita Das and Manob Jyoti Saikia
AI 2025, 6(8), 167; https://doi.org/10.3390/ai6080167 - 24 Jul 2025
Viewed by 578
Abstract
Breast cancer (BRCA) remains a significant cause of mortality among women, particularly in developing and underdeveloped regions, where early detection is crucial for effective treatment. This research introduces an innovative hybrid model that combines Improved Grey Wolf Optimizer (IGWO) with the Seagull Optimization [...] Read more.
Breast cancer (BRCA) remains a significant cause of mortality among women, particularly in developing and underdeveloped regions, where early detection is crucial for effective treatment. This research introduces an innovative hybrid model that combines Improved Grey Wolf Optimizer (IGWO) with the Seagull Optimization Algorithm (SOA), forming the IGWO–SOA technique to enhance BRCA detection accuracy. The hybrid model draws inspiration from the adaptive and strategic behaviors of seagulls, especially their ability to dynamically change attack angles in order to effectively tackle complex global optimization challenges. A deep neural network (DNN) is fine-tuned using this hybrid optimization method to address the challenges of hyperparameter selection and overfitting, which are common in DL approaches for BRCA classification. The proposed IGWO–SOA model demonstrates optimal performance in identifying key attributes that contribute to accurate cancer detection using the CBIS-DDSM dataset. Its effectiveness is validated using performance metrics such as loss, F1-score, precision, accuracy, and recall. Notably, the model achieved an impressive accuracy of 99.4%, outperforming existing methods in the domain. By optimizing both the learning parameters and model structure, this research establishes an advanced deep learning framework built upon the IGWO–SOA approach, presenting a robust and reliable method for early BRCA detection with significant potential to improve diagnostic precision. Full article
(This article belongs to the Section Medical & Healthcare AI)
Show Figures

Figure 1

22 pages, 1346 KiB  
Article
Understanding Video Narratives Through Dense Captioning with Linguistic Modules, Contextual Semantics, and Caption Selection
by Dvijesh Bhatt and Priyank Thakkar
AI 2025, 6(8), 166; https://doi.org/10.3390/ai6080166 - 23 Jul 2025
Viewed by 531
Abstract
Dense video captioning involves identifying, localizing, and describing multiple events within a video. Capturing temporal and contextual dependencies between events is essential for generating coherent and accurate captions. To effectively capture temporal and contextual dependencies between events, we propose Dense Video Captioning with [...] Read more.
Dense video captioning involves identifying, localizing, and describing multiple events within a video. Capturing temporal and contextual dependencies between events is essential for generating coherent and accurate captions. To effectively capture temporal and contextual dependencies between events, we propose Dense Video Captioning with Dual Contextual, Semantical, and Linguistic Modules (DVC-DCSL), a novel dense video captioning model that integrates contextual, semantic, and linguistic modules. The proposed approach employs two uni-directional LSTMs (forward and backward) to generate distinct captions for each event. A caption selection mechanism then processes these outputs to determine the final caption. In addition, contextual alignment is improved by incorporating visual and textual features from previous video segments into the captioning module, ensuring smoother narrative transitions. Comprehensive experiments conducted using the ActivityNet dataset demonstrate that DVC-DCSL increases the Meteor score from 11.28 to 12.71, representing a 12% improvement over state-of-the-art models in the field of dense video captioning. These results highlight the effectiveness of the proposed approach in improving dense video captioning quality through contextual and linguistic integration. Full article
Show Figures

Figure 1

34 pages, 2669 KiB  
Article
A Novel Quantum Epigenetic Algorithm for Adaptive Cybersecurity Threat Detection
by Salam Al-E’mari, Yousef Sanjalawe and Salam Fraihat
AI 2025, 6(8), 165; https://doi.org/10.3390/ai6080165 - 22 Jul 2025
Viewed by 466
Abstract
The escalating sophistication of cyber threats underscores the critical need for intelligent and adaptive intrusion detection systems (IDSs) to identify known and novel attack vectors in real time. Feature selection is a key enabler of performance in machine learning-based IDSs, as it reduces [...] Read more.
The escalating sophistication of cyber threats underscores the critical need for intelligent and adaptive intrusion detection systems (IDSs) to identify known and novel attack vectors in real time. Feature selection is a key enabler of performance in machine learning-based IDSs, as it reduces the input dimensionality, enhances the detection accuracy, and lowers the computational latency. This paper introduces a novel optimization framework called Quantum Epigenetic Algorithm (QEA), which synergistically combines quantum-inspired probabilistic representation with biologically motivated epigenetic gene regulation to perform efficient and adaptive feature selection. The algorithm balances global exploration and local exploitation by leveraging quantum superposition for diverse candidate generation while dynamically adjusting gene expression through an epigenetic activation mechanism. A multi-objective fitness function guides the search process by optimizing the detection accuracy, false positive rate, inference latency, and model compactness. The QEA was evaluated across four benchmark datasets—UNSW-NB15, CIC-IDS2017, CSE-CIC-IDS2018, and TON_IoT—and consistently outperformed baseline methods, including Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Quantum Genetic Algorithm (QGA). Notably, QEA achieved the highest classification accuracy (up to 97.12%), the lowest false positive rates (as low as 1.68%), and selected significantly fewer features (e.g., 18 on TON_IoT) while maintaining near real-time latency. These results demonstrate the robustness, efficiency, and scalability of QEA for real-time intrusion detection in dynamic and resource-constrained cybersecurity environments. Full article
Show Figures

Figure 1

15 pages, 1306 KiB  
Article
Risk Perception in Complex Systems: A Comparative Analysis of Process Control and Autonomous Vehicle Failures
by He Wen, Zaman Sajid and Rajeevan Arunthavanathan
AI 2025, 6(8), 164; https://doi.org/10.3390/ai6080164 - 22 Jul 2025
Viewed by 434
Abstract
Background: As intelligent systems increasingly operate in high-risk environments, understanding how they perceive and respond to hazards is critical for ensuring safety. Methods: In this study, we conduct a comparative analysis of 60 real-world accident reports, 30 from process control systems (PCSs) and [...] Read more.
Background: As intelligent systems increasingly operate in high-risk environments, understanding how they perceive and respond to hazards is critical for ensuring safety. Methods: In this study, we conduct a comparative analysis of 60 real-world accident reports, 30 from process control systems (PCSs) and 30 from autonomous vehicles (AVs), to examine differences in risk triggers, perception paradigms, and interaction failures between humans and artificial intelligence (AI). Results: Our findings reveal that PCS risks are predominantly internal to the system and detectable through deterministic, rule-based mechanisms, whereas AVs’ risks are externally driven and managed via probabilistic, multi-modal sensor fusion. More importantly, despite these architectural differences, both domains exhibit recurring human–AI interaction failures, including over-reliance on automation, mode confusion, and delayed intervention. In the case of PCSs, these failures are historically tied to human–automation interaction; this article extrapolates these patterns to anticipate potential human–AI interaction challenges as AI adaptation increases. Conclusions: This study highlights the need for a hybrid risk perception framework and improved human-centered design to enhance situational awareness and responsiveness. While AI has not yet been implemented in PCS incident studies, this work interprets human–automation failures in these cases as indicative of potential challenges in human–AI interaction that may arise in future AI-integrated process systems. Implications extend to developing safer intelligent systems across industrial and transportation sectors. Full article
Show Figures

Figure 1

18 pages, 1554 KiB  
Article
ChatCVD: A Retrieval-Augmented Chatbot for Personalized Cardiovascular Risk Assessment with a Comparison of Medical-Specific and General-Purpose LLMs
by Wafa Lakhdhar, Maryam Arabi, Ahmed Ibrahim, Abdulrahman Arabi and Ahmed Serag
AI 2025, 6(8), 163; https://doi.org/10.3390/ai6080163 - 22 Jul 2025
Viewed by 515
Abstract
Large language models (LLMs) are increasingly being applied to clinical tasks, but it remains unclear whether medical-specific models consistently outperform smaller, generalpurpose ones. This study investigates that assumption in the context of cardiovascular disease (CVD) risk assessment. We fine-tuned eight LLMs—both general-purpose and [...] Read more.
Large language models (LLMs) are increasingly being applied to clinical tasks, but it remains unclear whether medical-specific models consistently outperform smaller, generalpurpose ones. This study investigates that assumption in the context of cardiovascular disease (CVD) risk assessment. We fine-tuned eight LLMs—both general-purpose and medical-specific—using textualized data from the Behavioral Risk Factor Surveillance System (BRFSS) to classify individuals as “High Risk” or “Low Risk”. To provide actionable insights, we integrated a Retrieval-Augmented Generation (RAG) framework for personalized recommendation generation and deployed the system within an interactive chatbot interface. Notably, Gemma2, a compact 2B-parameter general-purpose model, achieved a high recall (0.907) and F1-score (0.770), performing on par with larger or medical-specialized models such as Med42 and BioBERT. These findings challenge the common assumption that larger or specialized models always yield superior results, and highlight the potential of lightweight, efficiently fine-tuned LLMs for clinical decision support—especially in resource-constrained settings. Overall, our results demonstrate that general-purpose models, when fine-tuned appropriately, can offer interpretable, high-performing, and accessible solutions for CVD risk assessment and personalized healthcare delivery. Full article
Show Figures

Graphical abstract

Previous Issue
Back to TopTop