Dr Simon Thorne is a Senior Lecturer in Computing and Information Systems at Cardiff Metropolitan University and an internationally recognised researcher in the reliability and governance of large language models (LLMs). He is the founder and co-lead of the Knowledge Observation Group (KOG), a multi-institutional research collaboration involving Cardiff Met, Cardiff University, Cambridge, Caltech, the University of Singapore and Microsoft. His work focuses on benchmarking LLM reasoning, understanding hallucination mechanisms, adversarial prompting, and the limits of language-based intelligence.

Title: Understanding the Reliability Limits of Large Language Models

Abstract: The accelerating deployment of large language models (LLMs) across research, industry and public-facing systems has placed renewed emphasis on understanding their reliability. This work examines the performance and limitations of contemporary LLMs through a structured benchmarking lens, focusing on correctness, reasoning stability and susceptibility to adversarial prompting. Empirical evaluations across deterministic logic tasks, spreadsheet reasoning, and structured problem solving reveal persistent vulnerabilities, including confident hallucination, inconsistent internal reasoning, and brittle behaviour under minimal perturbations. These limitations stem not only from training data and optimisation constraints, but from fundamental properties of language modelling itself: LLMs generate plausible sequences rather than truth-preserving inferences, and their apparent reasoning emerges from pattern prediction rather than grounded semantic understanding. The analysis argues that while such systems can effectively mimic aspects of human intelligence, decision making and knowledge use, they cannot replicate the underlying cognitive processes that guarantee verifiable reasoning. Recognising these structural boundaries is essential for safe deployment, meaningful benchmarking and the development of future reliability-focused methodologies.

Dr Angesh Anupam is Head of the Department of Data Science at the School of Technologies, Cardiff Metropolitan University, where he also leads the NetZero and Sustainability Research Group. He received his PhD in Nonlinear Dynamical Systems Modelling for Environmental Sustainability from the University of Sheffield, supported by the Grantham Centre for Sustainable Futures, an interdisciplinary research centre. Since 2014, Dr Anupam has built a broad interdisciplinary research profile that brings together data science, complex systems modelling, and machine-learning methods to investigate challenges spanning environmental, socio-economic, and infrastructural systems. His work has included ecological and wetland modelling, forecasting within transport and socio-economic contexts, and the use of AI-driven approaches to advance sustainability and ESG-related analysis.

Title: AI for a Net-Zero Systems: Modelling Complexity and Sustainable Transitions

Abstract: This keynote outlines how AI-driven nonlinear and complex-systems modelling can support the design, operation, and governance of net-zero systems. Dr Angesh Anupam presents a unified perspective in which AI, data science, and systems theory come together to analyse interdependent environmental, economic, and technological processes that shape sustainability outcomes. The talk highlights how AI-enabled dynamical modelling reveals hidden interactions, anticipates system responses, and offers predictive insights that traditional linear approaches often miss. By focusing on transferable modelling principles rather than single-domain case studies, the keynote underscores the broad applicability of these methods across diverse sectors. A central message is the essential role of interdisciplinary collaboration: meaningful progress toward net-zero futures requires integrating AI methodologies with domain expertise from multiple fields. The keynote concludes by showing how AI-supported scenario analysis and decision tools can help stakeholders design adaptive, resilient, and system-aware pathways to sustainability.