Introduction

Artificial Intelligence: State of the Art and Current Directions

Artificial intelligence has moved, in barely a decade, from a research curiosity to a technology that is reshaping the economy and society. This talk traces that trajectory, from the early rivalry between symbolic and connectionist approaches to the deep learning revolution that finally made machines competent at exactly the perceptual tasks Moravec identified as the hardest to program. Four developments — deep convolutional networks, deep reinforcement learning, diffusion models, and transformer-based language and vision-language models — account for most of what has surprised us, and their applications now run from medical imaging and drug discovery to marketing, finance, transportation, design, and scientific research itself. What is changing now is less the models than what is built around them: context engineering, retrieval, tool use, and explicit reasoning are turning models into systems, and those systems into agents that plan, remember, and act in the world with growing autonomy. Several major laboratories are now explicitly pursuing artificial general intelligence, raising the prospect of machines as capable, as powerful, and perhaps even as conscious as ourselves. Long before that point is reached, the systems already being deployed raise hard questions about employment, inequality, privacy, and the concentration of power. The talk closes with the questions that matter most: if such systems do come into existence, what will their social, legal, and ethical implications be, and how much time do we have to prepare for them?

Arlindo Oliveira received his BSc and MSc in Electrical and Computer Engineering from Instituto Superior Técnico (IST) and his PhD in the same field from the University of California at Berkeley. He was a visiting professor at MIT and a researcher at INESC, CERN, the Electronics Research Laboratory at UC Berkeley, the Berkeley Cadence Laboratories, and the University of Tokyo. He is a distinguished professor at IST, president of INESC, visiting professor at the Macau University of Science and Technology, non-executive director of Caixa Geral de Depósitos, and a researcher at INESC-ID. He has published five books, translated into several languages, and hundreds of scientific articles in international journals and conferences in the field.

How to navigate a mathematical modelling module through GenAI?

ChatGMP: Using Domain-Specific AI Chatbots to Support Learning in Good Manufacturing Practice

The increasing availability of large language models (LLMs) creates new opportunities for engineering education. This presentation introduces ChatGMP, an AI-powered chatbot developed to support teaching and learning in Good Manufacturing Practice (GMP) courses. The chatbot was trained on course-specific materials and designed to provide students with immediate feedback, guidance during audit exercises, and access to relevant GMP knowledge. ChatGMP enables a more interactive and scalable learning environment while reducing repetitive teaching tasks in large classes. Experiences from implementation at the Technical University of Denmark (DTU) will be presented, including student feedback, pedagogical benefits, and challenges related to AI adoption in higher education. The work demonstrates how domain-specific AI assistants can complement traditional teaching methods and contribute to the digital transformation of Chemical Engineering education.

Ulrich Krühne is Professor at the Technical University of Denmark (DTU) working in the areas of process systems engineering, bioprocess technology, digitalization, and engineering education. His recent work focuses on integrating artificial intelligence, chatbots, virtual laboratories, and digital learning tools into Chemical Engineering curricula to prepare students for an increasingly data-driven and AI-enabled industrial landscape.

More than just modelling and programming: educating life scientists in the age of AI

Modelling- and programming skills are essential for engineers. Most courses focus solely on teaching these technical skills. For successful application in industrial or scientific practice, ancillary skills regarding collaboration, project management, and responsible application are equally important. With LLMs rapidly taking over menial coding tasks, these aspects become even more important.

The programming and modelling content of the life science & technology program at TU Delft and Leiden University is continuously evolving to reflect these developments. I will discuss how we provide students with a basis in programming, modelling and machine learning during the bachelor, which we expand upon in the master with application in practice. For this, we have developed a collaborative, staged, project-based approach aimed at introducing a realistic project context, and we have introduced an elective on AI/ML for (bio)chemical engineering to familiarize students with the latest model developments. Current developments are aimed at including a stronger focus on responsible AI/ML within the ERASMUS+ FAIR PLAI project.

Dr. ir. Cees Haringa is associate professor in bioprocess engineering, focusing on model-driven bioprocess development. His research work leverages computational fluid dynamics (CFD) to gain insight in the environment cells experience in bioreactors, with application in scale-down and scale-up. He holds a 50/50 research/education position; his education work and education innovation are centered on effective and responsible use of modelling in biotechnology.

Closing discussion and next steps