This seminar will delve into various aspects of building and managing trustworthy systems, focusing on the following topics:
- Autonomous Systems: Advancements and challenges in the deployment of autonomous vehicles, vessels, drones, and robots, emphasizing safety and trust.
- AI Agents: Issues related to trust and safety of agentic AI systems.
- Complex Industrial Systems: The use of modelling, simulation, and AI for understanding, controlling, or managing complex systems.
- Multi-Agent AI: Trust and safety issues in systems involving multiple AI entities.
- Ethical and Social Implications and Governance: Societal impacts and governance of intelligent systems, including privacy, bias, and ethical considerations.
The seminar is also meant as a starting point for the newly established ESReDA Project Group How can AI and other Emerging Technologies strengthen Safety and Security, where more than 10 European institutions will collaborate the next three to four years to make progress in this important area. Additional members are still welcome.
Keynote speakers:
-
Prof. Ingrid B. Utne is a professor at the NTNU Department of Marine Technology, and an expert on risk analysis of complex systems. In recent years, her focus has been on implementation of risk-aware AI/mission planning and decision making in robotics, particularly autonomous ships, drones ASVs and AUVs.
-
Prof. Hector Zenil is an associate professor at King’s College with a prolific academic and entrepreneurial career spanning many disciplines, including complexity science, artificial intelligence, computer science, precision healthcare and ageing research. He contributed to the computational linguistics’ code behind WolframAlpha, Siri, Alexa, and GPT4. He has authored 130+ academic papers and multiple books, including ‘Algorithmic Information Dynamics’ and ‘A Computable Universe’, and is Managing Editor of the Complex Systems journal.
-
Prof. Gaute T. Einevoll is a professor of physics at the Norwegian University of Life Sciences and the University of Oslo, with focus on computational neuroscience. He is a partner in the EU Human Brain Project and co-leader of the Norwegian node of the International Neuroinformatics Coordinating Facility (INCF). He also hosts the Norwegian science podcast ‘Vett og vitenskap med Gaute Einevoll’ and the ‘Theoretical Neuroscience Podcast’ and has authored several books.
PROGRAMME:
25 September
10:00–10:15 Welcome (Astrid Rusås Kristoffersen, Group Director Research and Development, DNV)
10:15–11:00 KEYNOTE: Supervisory Risk Control for Improved Safety and Trust in Embodied AI (Ingrid B. Utne, Professor, Norwegian University of Science and Technology)
11:00–11:15 Break
11:15–11:45 Safety Assurance for AI-Based Components: Possible Coping Strategies (Eric Marsden, Programme Manager, FonCSI)
11:45–12:15 An Approach to Structured Safety Arguments for Safe Use of Black Box Reinforcement Learning in the Control Loop of Real-Time Control Systems (Rune Winther, Principal Scientist, Risk & Safety, Institute for Energy Technology; Emilie Greaker, Master's Student, Engineering)
12:15–12:35 Industry 5.0: Navigating the Regulatory Frontier for Trustworthy AI (Anna Lisa Vetere Arellano, Senior Scientific Officer, European Commission)
Towards a Taxonomy of AI Tools Applied in the Process Industry (Tuuli Tulonen, Senior Officer, Finnish Safety and Chemicals Agency; Sever Paul, Investigator, Romanian Railway Investigation Agency)
12:35–13:20 Lunch
13:20–13:50 Trust in AI: Interpretability and Beyond – A Review (K. Darshana Abeyrathna, Senior Researcher, DNV; Tita Alissa Bach, Principal Researcher, DNV)
13:50–14:20 Towards a Taxonomy of AI Tools Applied in the Process Industry (Peter Kouwenhoven, MSc Chemical Engineering)
14:20–14:50 Building Trust in Autonomous Vessels by the Law (Erik Røsæg, Professor, University of Oslo)
14:50–15:05 Break
15:05–15:25 The Role of Simulations When Assuring Complex Systems (Siegfried Eisinger, Senior Principal Specialist, DNV)
15:25–15:55 Knowledge-Driven Maintenance for Fusion: An IFMIF-DONES Ontology Enabling Simulation and Decision Support (María Megía Cardeñoso, Post Doc, Engineering, University of Grenada)
15:55–16:15 On the Failure Probability of Silicon Crystals Related to Particle Collisions and Scattering (Elena Bylski, Ruhr West University of Applied Sciences)
16:15–16:35 Extending Maintenance Models Through Semantic Integration: A High-Level Petri Net Perspective for IFMIF-DONES (Antonio Sánchez Membrives, University of Grenada)
16:35–16:40 Info about dinner (Siegfried Eisinger, DNV)
19:00– Dinner
26 September
09:00–09:15 Opening of Day 2 (Frank Børre Pedersen, Programme Director, Maritime Research, DNV)
09:15–10:00 KEYNOTE: From BI (Biological Intelligence) to AI and Back (Gaute T. Einevoll, Professor, University of Oslo)
10:00–10:45 KEYNOTE: Not All Patterns Are Equal: How Complexity Can Evaluate and Improve AI Frontier Models (Hector Zenil, Associate Professor, King's College London)
10:45–11:00 Break
11:00–11:30 Towards Trustworthy Knowledge-Augmented Language Models for Maintenance Decision Support in Fusion Facilities (Mohammed Anouar Amzert, P.D.I., University of Grenada)
11:30–12:00 The Development, Use, and Management of AI to Ensure Safety and Security in Socio-Technical Systems (Sever Paul, Investigator, Romanian Railway Investigation Agency; Tuuli Tulonen, Senior Officer, Finnish Safety and Chemicals Agency)
12:00–12:30 From Static Certification to Continuous Assurance: Adapting Governance for Learning AI Systems (Eric Marsden, Programme Manager, FonCSI)
12:30–13:15 Lunch
13:15–13:35 What is the Link Between Complexity and Risk? (Andreas Hafver, Principal Research Scientist, DNV)
13:35–14:05 The Shape of Complexity (Manuel Guillermo Reta Palacios, Senior Research Engineer, DNV)
14:05–14:35 The AI Trust Paradox: Using Intelligent Systems for Risk Management and Compliance in Critical Infrastructures (Juan Camilo Guevara Gomez, Researcher, DNV)
14:35–15:05 Discussion: Research Gaps and Future Research Areas
15:05–15:15 Closing Session – Next ESReDA Seminar (TBD)