The emergence of the transformer architecture has ushered in a new era of possibilities, showcasing remarkable capabilities in generative tasks exemplified by models like GPT4o, Claude 3, and Llama 3. However, these advancements come with a caveat: predominantly trained on data gleaned from social media platforms, these systems inadvertently perpetuate societal biases and toxicity. Recognizing the paramount importance of AI Safety and Alignment, our study embarks on a thorough exploration through a comprehensive literature review focused on toxic language. Delving into various definitions, detection methodologies, and mitigation strategies, we aim to shed light on the complexities of this issue. Our study primarily focuses on transformer-based architectures, with deep learning methods serving as a crucial baseline for comparison in our exploration of toxic language detection and mitigation. Through our investigation, we uncover a multitude of challenges inherent in toxicity mitigation and detection models. These challenges range from inherent biases and generalization issues to the necessity for standardized definitions of toxic language and the quality assurance of dataset annotations. Furthermore, we emphasize the significance of transparent annotation processes, resolution of annotation disagreements, and the enhancement of Large Language Models (LLMs) robustness. Additionally, we advocate for the creation of standardized benchmarks to gauge the effectiveness of toxicity mitigation and detection methods. Addressing these challenges is not just imperative, but pivotal in advancing the development of safer and more ethically aligned AI systems.
The crystallization of soluble salts poses a significant challenge to mural painting conservation. While cellulose poultices are widely used to desalinate mural paintings due to their high absorption and ease of handling, their effectiveness within the porous network of wall paintings remains a complex issue. For the first time, this study explores the potential of micro-structured cellulose-based foams as an alternative to conventional poultices for desalinating fresco wall paintings. A laboratory experiment compared the efficacy of foams and poultices, using fresco wall painting mock-ups (produced with the Roman technique) that were vacuum-impregnated with salt solutions (chlorides, sulfates, and mixtures). Short and long application times were considered, and foam reusability across multiple application cycles was assessed. Micro-energy dispersive X-ray fluorescence (& micro;-EDXRF) imaging was employed to quantitatively evaluate salt content reduction, both superficially and throughout the mock-up stratigraphy. Results show that foams are considerably more effective than poultices, achieving a salt removal efficiency between 6 and 10 times higher. The uniform micro-porous foam network enables faster desalination, reducing treatment risks and minimizing waste while supporting circular economy principles. This study also demonstrates the utility of & micro;-EDXRF imaging in monitoring desalination efficacy for both surface and cross-section analyses when assessing new desalination protocols.
This paper reports a descriptive, process-oriented case study of the application of Z-Inspection®, an ethically aligned co-design methodology, in the early design phase of an artificial intelligence (AI) system for healthcare. The methodology was applied within the Horizon Europe project VALIDATE [1], which develops and validates a prognostic clinical decision support system (CDSS) to support treatment decisions in acute ischaemic stroke. An interdisciplinary team of Z-Inspection® experts, AI developers, and clinical stakeholders jointly identified ethical, legal, and technical issues relevant to the planned system. The assessment resulted in 22 ethical issues, 12 dilemmas, 18 risks, and 48 derived requirements, each mapped to the European Commission’s trustworthy AI principles. The identified issues, dilemmas, risks, and derived requirements constitute the primary findings of this descriptive case study and document how general trustworthy-AI principles were translated into context-specific technical, clinical, governance, and organisational requirements. By documenting the process and publishing the complete requirements, the study provides a detailed case for critical scrutiny and potential uptake in other healthcare AI projects, though its transferability and scalability remain to be established in further studies.
This study advances the scale-up of recovering the refrigerant R-32 (difluoromethane), a hydrofluorocarbon with moderate global warming potential and high thermodynamic efficiency that is a major component in most next-generation low-GWP blends. In particular, R-32 recovery from R-410A (R-32/R-125: 69.7/30.3 vol%) is sought as R-410A is being phased-down and represents a substantial stockpile of recoverable R-32 from end-of-life refrigeration equipment. To that end, asymmetric hollow fiber membranes made of the highly selective 6FDA-TMPD polyimide were extruded via dry-jet wet spinning, and the spinning parameters and dope composition are reported. These membranes were used to assemble membrane modules with a surface area of 31 cm2 and their integrity was verified using CO2 and N2. The performance of the 6FDA-TMPD prototype for separating R-410A was assessed under relevant pressures up to 7 bar, achieving exceptional R-32 product purity (99.5 vol%) together with high R-32 recovery (85.4%). Moreover, a mathematical model was developed incorporating local fugacity gradients along the fiber length and concentration-dependent permeance that captured the strong condensability effects of fluorinated hydrocarbons on membrane plasticization. Finally, this model was applied to optimize the membrane area and compressor duty of a two-stage membrane process designed to maximize R-32 recovery while meeting the product specifications of the virgin refrigerant, R-32 purity > 99.5 wt%. The normalized total energy required was as low as 0.06 kWh kg-1 of R-410A treated, showing that membrane separation is an extremely energy efficient way of reclaiming R-32 from waste R-410A refrigerants. Overall, the results support improved resource efficiency and reduced uncontrolled emissions of potent greenhouse gases, thus contributing to more sustainable practices in the refrigeration sector.
Low-temperature water electrolysis technologies exhibit a significant potential not only to replace grey hydrogen use in existing chemical industries but also to decarbonize hard-to-abate sectors. The main objective of this work is to assess the techno-economic viability of using an MW-scale electrolysis-based green hydrogen plant as a supplier for an industrial heating furnace. HYTECSIM simulation tool is employed to physically model alternative plant configurations and to estimate both onsite footprint and economic metrics, including the levelized cost of hydrogen (LCOH). Consumption measurements from an internal zone of an ingot heating rotary furnace are used as demand profiles in the simulations. Under these premises, three plant configurations are sized and simulated in order to quantify the capital expenditures (CAPEX), footprint, electrolyzer performance, and operational expenditures (OPEX) and to evaluate their combined impact on the LCOH. Results reveal that extending the electrolyzer's stack lifespan by optimizing its operation has great potential to achieve competitive LCOH values.
