Interview with Bertie Müller

John Bensalhia talks to Dr Bertie Müller about the new EU AI legislation and the benefits and challenges of using AI for cyber security.

Dr. Bertie Müller, Senior Lecturer in Computer Science – PGT Programme Director & MSc Co-ordinator at Swansea University will be appearing at Big Data & AI World Frankfurt 2023. Bertie's forthcoming presentation will address generative AI in the light of forthcoming regulation of AI-driven methods. “This will include insights from current projects on AI evaluation in the healthcare and medical domain,” says Bertie. “The presentation will also summarise opportunities and risks of (generative) AI in a variety of sectors.”

Bertie holds a PhD in Computer Science from the University of Hamburg, Germany. Before moving to Wales, he worked in the Computer Science Department at Durham University. “My background is in formal modelling and analysis of distributed systems, including autonomous agents. This involved mainly the verification of systems properties. In 2006, I began applying these methods to multi-agent systems and AI in a broader sense.”

Since 2014, Bertie has been chair of the world's oldest AI society, AISB (The Society for the Study of Artificial intelligence and Simulation of Behaviour, (www.aisb.org.uk). “As an academic I have always strived to collaborate with industrial partners (e.g., in aerospace, finance, and med tech), and to influence policy making at UK (Home Office Digital Strategy, Welsh Government Panel on Digital Technology and the Future of Work in Wales) and European levels (European AI Landscape Initiative).”

At Swansea University, in his role as PGT Programme Director, Bertie oversees the postgraduate taught programmes, such as the MSc in Data Science, The MSc in Cyber Security, and the MSc in Advanced Computer Science. “We also offer an MSc in Computer Science for students not having a first degree in computer science. My research focus lies in the intersection between AI and cyber security with a flavour of technology ethics and responsible design of data-driven algorithms.”

Discussing his thoughts on the new EU AI legislation, Bertie comments:  “I believe that AI regulation is necessary to complement existing regulation of critical sectors, such as the medical and finance sectors. Responsible handling of ever-growing repositories of data is crucial to protect individuals' rights, as well as intellectual property. Growing concerns over misinformation and cyber-security risks call for transparency and explainability.”

“One thing that any AI legislation needs to get right, though, is the problem of capturing the correct set of technologies in its definition of AI. Being too narrow in such a definition will exclude technologies related to AI, that also need to be treated with the same precautions, while being too general will affect the effectiveness of rules.”

Unlike some opponents of AI legislation, Bertie does not believe that such legislation will slow down innovation. “There is plenty of scope for reliable, responsible, and resilient systems design, leading to systems and processes that are accepted and trusted by users.”

With respect to using AI for cyber security, Bertie says that we can no longer achieve meaningful intrusion detection and protection from cyber attacks without the use of AI. “On the other hand, AI is part of a weapons race where successful use of AI for detection of data breaches is countered by AI-optimised attacks and prosecution evasion. Most interesting to me are vulnerabilities that the use of ubiquitous AI technology introduce. These include the possibility of subtle and hard to detect long-term data pollution or poisoning. There are many challenges ahead of us ...”