Magazine - Computing in Science & Engineering
Important Dates Submissions due: 21 January 2026 Publication date: July - September 2026 The past years have seen an explosive growth in the use and adoption of artificial intelligence (AI) and machine learning (ML) frameworks across industries and applications, driven largely by advances in deep learning architectures, increased computational power, and the availability of massive datasets for model training. Emerging foundational models such as OpenAI’s GPT series and Google’s Gemini have revolutionized natural language processing by enabling zero-shot and few-shot learning capabilities, significantly reducing the need for task-specific training. Generative models like DALL-E, Stable Diffusion, and ChatGPT have transformed not only creative industries but also scientific research by enabling on-the-fly creation of new content, such as high-quality images, text and even code. Within the field of Scientific ML (SciML), deep learning (DL) models have demonstrated utility as fast non-intrusive surrogates for expensive high-fidelity models. They have also been used to enable data-driven physics discovery. While AI/ML have shown potential for advancing research in various science and engineering applications, applying AI/ML methods within these domains comes with some challenges, including: Data quality and availability: scientific data are often scarce, noisy and/or incomplete, which limits the training of accurate AI/ML models. Interpretability: many AI/ML models act as “black boxes,” making it difficult to understand or trust their predictions in scientific applications where explainability is essential. Generalization, robustness and trustworthiness: models trained on specific datasets may not generalize well to new, unseen conditions or scales, which is problematic in scientific applications requiring trusted predictions with quantifiable error bounds. Systematic refinement mechanisms: unlike traditional discretization methods, it is often not clear how to “refine” an AI/ML model to ensure it satisfies a specified error tolerance and converges to the sought-after physical solution. Physics and structure preservation: AI/ML models do not always respect the physics…
Submissions Due: 21 January 2026
