Nick Schurch keynote speaker at Alan Turing Institute’s AIUK Fringe event on “AI and Sustainability”

8 July, 2024
Image
Man speaking on stage behind lectern, screen showing the talk title  "Reality Checking AI’s current capacity for environmental insight."

As part of the Alan Turing Institute’s AIUK Fringe Series the University of Aberdeen organised the event “AI and Sustainability” exploring the use of Artificial Intelligence (AI) in the drive towards net zero and environmental sustainability. Our Principal Statistician for Ecology & Environmental Science, Dr Nick Schurch, was the event’s keynote speaker with a talk on “Reality Checking AI’s current capacity for environmental insight.”

In his keynote Nick talk gave an overview of the history of AI models, taking us us to the present time when most of us have access to both high performance computing facilities and unprecedented amounts and variety of data, leading to immense potential for powerful and complex models. Alongside this, promises of positive impact in ecological and environmental science have been highlighted. While in some areas, an impact has been realised, some others have been slower to adapt AI approaches. 

Computational science, data science, and importantly, data richness, all three are key components of modern AI approaches to scientific research.” 

The talk also included an interesting comparison of a process model and a specific type of AI models (Long short-term memory neural network) for a hydrological case study, showing that the AI model outperforms the process model, but is purely predictive whereas the process model has an additional explanatory element that makes it potentially more valuable. Other examples included the great performance provided by AI random forests to map land cover, the significant contribution of AI pattern learning for habitat classification, and the data challenges faced when working with computer vision models in studies with ecological camera trap surveys.

Nick emphasised the importance of careful thinking about the training data that goes in our AI models to ensure that we are both doing the right thing and getting the best performance we can. He equally advocated for transparency and interpretability as good practice for science development.

The event which was held in March 2024 was organised by Prof. Georgios Leontidis (University of Aberdeen). You can watch the recording of the event here