Professor Plamen Angelov, of Lancaster’s School of Computing and Communications, has recently undertaken two trips – to the Jet Propulsion Lab (JPL), NASA and to the Phi Lab based at the European Space Agency, respectively – to give talks on Deep Learning and AI for Earth Observation.

Professor Angelov – who holds the status of Visiting Professor at the Phi Lab in the European Space Agency – was invited in December to JPL NASA in Pasadena, California, to give a talk entitled “Learning from Data in post-Foundation Models Era: bringing learning and reasoning together”, which delved into the nature of explainable AI, and how Deep Learning can be developed in a way that is trustworthy and interpretable.

“Foundation models” are a newer form of Deep Learning and are utilised by technologies such as ChatGPT, Claude, Gemini, and many other cutting-edge AI tools. They are characterised by their ability to perform a wide range of tasks, and are typically trained on incredibly large data sets. However, many of these models are “opaque” in how they reach their conclusions – with some models even making incorrect predictions with high levels of confidence, eroding public trust. Many of these models even “forget” previously held information when faced with continual learning, rendering them only useful for the small subset of tasks they were originally designed for.

Professor Angelov’s talk considered how we can get “the best from both worlds” – high accuracy from Deep Learning models and interpretability through knowledge-preserving representation and decision making, and providing examples of how such systems might be used for Earth Observation, amongst other applications. Whilst in Pasadena, Professor Angelov was also privileged to receive a tour of the control room for unmanned missions.

Professor Angelov also gave a talk at the European Space Agency last month as a part of his Visiting Professor role, in addition to spending a week there collaborating with academics based within the Phi Lab.

On his visits to these notable institutions, Professor Angelov commented: "Giving a talk at JPL-NASA in Pasadena and discussing with experts who are behind Foundation Models such as Prithvi provided many opportunities for collaboration, and exploring the ways to improve the current state-of-the-art is both enriching and exciting as well as promising and a thoroughly worthwhile part of my sabbatical. Particularly exciting was the control room, where I got to witness the live data streams that we aim to process coming in real time from the outer space, as well as visiting the lab where robots such as Perseverance rover exploring Mars were designed, assembled and tested.”

“The one week stay at Phi-Lab, ESA as part of my Visiting Professorship role gave me the opportunity to directly discuss with, present to, set up collaborations with a range of early career postdocs and more experienced experts. Phi-Lab is the AI centre of ESA and is addressing a range of Earth Observation problems spanning a range of natural disasters such as floods, wildfires, draught, and overall climate health.”