Generative language models have witnessed substantial traction, notably with the introduction of refined models aimed at more coherent user-AI interactions—principally conversational models. The epitome of this public attention has arguably been the refinement of the GPT-3 model into ChatGPT and its subsequent integration with auxiliary capabilities such as search features in Microsoft Bing. Despite voluminous prior research devoted to its developmental trajectory, the model’s performance, and applicability to a myriad of quotidian tasks remained nebulous and task specific. In terms of technological implementation, the advent of models such as LLMv2 and ChatGPT-4 has elevated the discourse beyond mere textual coherence to nuanced contextual understanding and real-world task completion. Concurrently, emerging architectures that focus on interpreting latent spaces have offered more granular control over text generation, thereby amplifying the model’s applicability across various verticals. Within the purview of cyber defense, especially in the Swiss operational ecosystem, these models pose both unprecedented opportunities and challenges. Their capabilities in data analytics, intrusion detection, and even misinformation combatting is laudable; yet the ethical and security implications concerning data privacy, surveillance, and potential misuse warrant judicious scrutiny.

Convolutional Neural Networks (CNNs) have proven effective in automated plant disease diagnosis, significantly contributing to crop health monitoring. However, their limited interpretability hinders practical deployment in real-world agricultural settings. To address this, we explore advanced CNN architectures, namely ResNet-50 and EfficientNet, augmented with attention mechanisms. These models enhance accuracy by optimizing depth, width, and resolution, while attention layers improve transparency by focusing on disease-relevant regions. Experiments using the PlantVillage dataset show that basic CNNs achieve 46.69% accuracy, while ResNet-50 and EfficientNet attain 63.79% and 98.27%, respectively. On a 39-class extended dataset, our proposed EfficientNet-B0 with attention (EfficientNetB0-Attn), integrating an attention module at layer 262, achieves 99.39% accuracy. This approach significantly enhances interpretability without compromising performance. The attention module generates weights via backpropagation, allowing the model to emphasize disease-relevant image regions, thereby enhancing both accuracy and interpretability.

Three-dimensional physical systems play a pivotal role in the development of cyber-physical infrastructures, particularly in the implementation of digital twins that enable the evaluation of hypothetical and adverse scenarios through high-fidelity simulation. This work presents a real-time 3D monitoring and feedback system designed for a custom transverse-axis wind turbine, integrating physical modeling principles with simulation engines developed initially for game environments. This hybrid architecture facilitates the virtual prototyping, testing, and validation of wind energy systems under near-operational conditions. The proposed framework combines two key components: (i) a physics-based model grounded in the mechanical and electromagnetic dynamics of wind turbine operation, and (ii) a data-driven architecture composed of multiple layers. The physical layer interfaces directly with the sensors and actuators of the turbine, ensuring real-time synchronization between the physical and virtual systems. Data acquisition and communication are managed through the MQTT protocol, enabling low-latency streaming and robust interoperability. A long short-term memory neural network is integrated into the architecture to enhance predictive capabilities and trained to forecast wind energy production. An intelligent battery management system subsequently utilizes the output of the model to optimize charging strategies.

The revolution generated by the Internet of Things (IoT) has radically changed the world; countless objects with remote sensing, actuation, analysis and sharing capabilities are interconnected over heterogeneous communication networks. Consequently, all of today’s devices can connect to the internet and can provide valuable information for decision making. However, the data collected by different devices are in different formats, which makes it necessary to develop a solution that integrates comprehensive semantic tools to represent, integrate and acquire knowledge, which is a major challenge for IoT environments. The proposed solution addresses this challenge by using IoT semantic data to reason about actionable knowledge, combining next-generation semantic technologies and artificial intelligence through a set of cognitive components that enables easy interoperability and integration for both legacy systems and emerging technologies, such as IoT, to generate business value in terms of faster analytics and improved decision making. Thus, combining IoT environments with cognitive artificial intelligence services, COSIBAS builds an abstraction layer between existing platforms for IoT and AI technologies to enable cognitive solutions and increase interoperability across multiple domains. The resulting low-cost cross platform supports scalability and the evolution of large-scale heterogeneous systems and allows the modernization of legacy infrastructures with cognitive tools and communication mechanisms while reusing assets.

Comprehensive healthcare seeks to uphold the right to health by providing patient-centred care in both personal and work environments. However, the unequal distribution of healthcare services significantly restricts access in remote or underserved areas—a challenge that is particularly critical in mental health care within low-income countries. On average, there is only one psychiatrist for every 200,000 people, which severely limits early diagnosis and continuous monitoring in patients’ daily environments. In response to these challenges, this research explores the feasibility of implementing an information system that integrates cloud computing with an intelligent Facial Expression Recognition (FER) module to enable psychologists to remotely and periodically monitor patients’ emotional states. This approach enhances comprehensive clinical assessments, supporting early detection, ongoing management, and personalised treatment in mental health care. This applied research follows a descriptive and developmental approach, aiming to design, implement, and evaluate an intelligent cloud-based solution that enables remote monitoring of patients’ emotional states through Facial Expression Recognition (FER). The methodology integrates principles of user-centred design, software engineering best practices, and machine learning model development, ensuring a robust and scalable solution aligned with clinical and technological requirements. The development process followed the Software Development Life Cycle (SDLC) and included functional, performance, and integration testing. To assess overall system quality, we defined an evaluation framework based on ISO/IEC 25010 quality characteristics: functional suitability, performance efficiency, usability, and security. The intelligent FER model achieved strong validation results, with a loss of 0.1378 and an accuracy of 96%, as confirmed by the confusion matrix and associated performance metrics.