Bournemouth University receives over £500,000 to improve AI reliability
Bournemouth University has received 610,000 euros (£529,000) as part of a major EU-funded project to improve the reliability, trustworthiness, and security of AI agents.
FRAME is a 6m EUR project that aims to provide a shared software engineering environment to design, build and test AI agents effectively. This will lead to development methods and guidelines for future developers, helping them to produce agents that are more useful and reliable, even when they need to operate with incomplete or uncertain information.
An AI agent is a type of AI system that can analyse data and make decisions. Many people can experience this through large language models, such as Microsoft Co-Pilot or ChatGPT; they also have advanced commercial applications in industries such as healthcare, robotics, and computing. The most advanced agents can learn from experience, adapt to new situations and apply reasoning.
“We are all interacting with AI agents more and more in our daily lives, whether through using online customer services, planning our holidays, or managing our work emails," explained Hamid Bouchachia, Professor of Data Science and Intelligent Systems at Bournemouth University, who led the funding bid and will assume the role of technical and quality coordinator for FRAME. “The growing role of AI agents can be beneficial for the economy and for society, but it is important that they are developed responsibly and with suitable quality control,” he added.
The project will test the development of agents in real-world settings, including robotics, healthcare, and software development. FRAME will also help train industry professionals, postgraduate students, and researchers.
Professor Bouchachia and the team at BU will lead research into how AI agents can continuously learn and improve, as well as how they plan, reason, and make trustworthy decisions. This work will help ensure that the AI agents developed in FRAME are not only powerful but also reliable and safe to use in real-world applications.