WMG Hosts Panel on AI Regulation and Safety at ADAS & Autonomous Vehicle Technology Expo

At the ADAS & Autonomous Vehicle Technology Expo in Stuttgart, Dr Siddartha Khastgir, Head of Safe Autonomy, WMG, the University of Warwick, and Dr Sarah Gates, Director of Public Policy at Wayve, participated in a panel discussion on “Safe AI for Automated Driving: Reality or Fiction?”

Moderated by Dr Khastgir, the panel brought together industry experts, including Richard Damm, President of KBA and Chair of UNECE GRVA; Matteo Oldoni, Senior Engineer at NVIDIA; and Dr Georg Stettinger, Senior Project Manager at Infineon Technologies. The panel delved into the necessity and feasibility of regulating AI in Automated Driving Systems (ADS), highlighting strategies for addressing bias in training datasets, the challenges posed by Large Language Models (LLMs), and a comparison of different regulatory frameworks.

During the discussion, Dr Gates emphasized the importance of a multi-faceted approach to AI regulation in automated driving. She noted that regulation, standards, and guidelines should work in tandem with legislation laying the groundwork for the safe deployment of ADS. This approach is already being implemented in the UK through the AV Act, which provides a framework for deploying self-driving technology. Dr Gates highlighted the critical role of developers in actively participating in the creation of these regulations and standards, as they have the deepest understanding of their technology.

The DriveSafeAI project, a collaborative effort between Wayve and WMG funded by the UK Government, was presented as a prime example of how industry and academia can work together to develop safety methods and tools that inform future regulation. This project aims to build the evidence base necessary to ensure AI safety in vehicles, contributing to the global conversation on AI regulation.

The panel also addressed concerns about the current regulatory landscape’s ability to handle AI in safety-critical applications like automated driving. Dr Gates acknowledged that while some in the AV community fear that regulation could stifle innovation, intelligent, sector-specific regulation, such as the UK’s approach, is crucial for assuring the safety of AI technologies before they are introduced to the market.

On the topic of AI explainability, Dr Gates explained that it is essential for building trust and confidence in AI-powered systems. Wayve’s approach includes understanding the data used to train AI models, developing new techniques for model introspection, and using tools like Wayve’s Vision-Language-Action model, LINGO, to predict driving paths and explain AI decisions. These efforts are part of an ongoing commitment to ensuring that AI systems in self-driving vehicles are safe, reliable, and transparent.