Reuben College welcomed two distinguished scientists and accomplished communicators to its regular Tuesday Talks/DwD dinner in the last full week of Hilary Term 2025. Dr Peter Coveney is a Professor of Physical Chemistry at University College London and has made significant contributions to the fields of computational medicine and high-performance computing. Dr Roger Highfield, Science Director at the Science Museum Group, is a prolific science writer and editor.
In 2023, the two speakers co-authored Virtual You: How Building Your Digital Twin Will Revolutionize Medicine and Change Your Life, a book exploring the implications of digital twin technology across various fields. A digital twin is a virtual model of a physical entity - whether an object, system, or potentially even an entire organism. These techniques are already used in engineering and manufacturing, but they are now being extended to human biology.
Roger Highfield opened the talk by discussing real-world examples illustrating the potential of in silico simulations of organ systems. One prominent example was the digital heart, which allows researchers to simulate various pathologies, evaluate potential drug targets, and explore different treatment modalities. Similar efforts are underway for other organs and biological systems.
Peter Coveney then discussed the numerical challenges involved in running simulations on digital computers, drawing on his 2019 paper with Boghosian and Wang. This paper demonstrated how systematic and potentially misleading distortions can arise when modelling complex, chaotic systems using floating-point numbers. Coveney emphasised the risks of blindly trusting numerical simulations. These challenges are even more pronounced when modelling biological processes, which operate across multiple scales—from molecular interactions and cellular processes to tissue dynamics, whole-organ functions, and broader environmental influences.
The talk also addressed key limitations of current artificial intelligence models, particularly large language models (LLMs). Coveney pointed out that these models still struggle to understand and predict phenomena outside their training data and therefore lack the deeper causal reasoning required to make independent scientific discoveries. Another concern is that AI can reinforce biases embedded in its training data. This is particularly problematic in scientific research, where AI might replicate flawed assumptions rather than challenge them. Ensuring fairness and accuracy in AI-driven simulations will require careful curation of data sources and continuous scrutiny to prevent the reinforcement of existing biases.
Beyond bias, AI’s impact on scientific research is further constrained by issues of reproducibility, interpretability, and explainability. Ethical and security considerations also arise, especially in applications involving sensitive health data. Another critical challenge discussed was the environmental impact of AI and digital twin technologies, given the vast computational and consequently energy, resources required to power these models.
The evening concluded with a panel session, where Peter and Roger were joined by Lord Tarassenko (College President) and Professor Anisoara Calinescu (Official Fellow and Theme Lead for AI & Machine Learning). The Q&A with the panel was led by Professor Chrystalina Antoniades (Official Fellow and Theme Lead in Cellular Life). This panel discussion sparked lively debate among the very large and engaged audience.
Some key messages from the talks, panel discussion, and “local” discussion at our table included reflections that while digital twins offer promise for advancing our understanding of complex medical systems, their development faces significant challenges, including computational limitations and patient acceptability.
Crucially, regardless of how innovative or apparently powerful new technologies may be, their application in patient care must continue to adhere to the principles of evidence-based medicine.
It was exciting to conclude the academic programme for this term with an interdisciplinary collaboration between members of our Public Engagement, Cellular Life, Machine Learning and Ethics & Values theme. We look forward to more exciting events next term.
Missed out?
You can now watch the recording here.
