Location:  In-person : The Knox Shaw Room, Sidney Sussex College, Cambridge CB2 3HU

You are warmly invited to the 2023 Aviva-Cambridge annual partnership event, celebrating 5 years of collaborative research between the University and Aviva. It will be held at Sidney Sussex College, Cambridge from 2pm to 4pm. With talks on large language models and the use of AI in assessing the risk from climate change, it promises to be a topical and thought provoking occasion.

Agenda

2pm - Opening remarks -  Karen Kennedy University of Cambridge & Alessia Kosagowsky Aviva

2:05pm - Introduction and partnership summary - Mark Hayes University of Cambridge

2:10pm - Tools and inference in large language models - Paulius Rauba University of Cambridge

2:30pm - Using Natural Language Processing to enhance our understanding of climate risks and impacts- Helen Jackson, ClimateNode

2:50pm - Wildland-Urban Interface fires: very dangerous but not as unpredictable as we fear - Prof. Epaminondas Mastorakos  University of Cambridge

3:10pm - Future Infrastructure and the Built Environment - Dr. Brian Sheil University of Cambridge 

3:30pm - Panel: data science & climate change

4pm - coffee & tea served

Registrations

This event is invite-only for in-person participants. Please register using the link provided in your invitation email.  

Directions

See https://www.sid.cam.ac.uk/about-sidney/visiting-sidney for directions.

Talk abstracts

Tools and Inference in Large Language Models Paulius Rauba PhD student, van der Schaar group, Department of Applied Mathematics and Theoretical Physics, University of Cambridge

Large Language Models have recently demonstrated the emergent capabilities of using external tools to aid their decision-making process, as well as being able to perform inference and reasoning in complicated scenarios. This talk presents the (constantly moving) state-of-the-art in both areas with a short guide on how to know if the outputs/predictions of LLMs should be trusted

Using Natural Language Processing to enhance our understanding of climate risks and impacts Helen Jackson ClimateNode

Governments and businesses need information on evolving climate risks to plan and make adaptation decisions. However, as a recent report by the Institute and Faculty of Actuaries and the University of Exeter pointed out, modelling the likely economic and financial impacts of climate change in a comprehensive way is very difficult to do. This is almost certainly leading to underestimation and complacency. This talk will describe the work ClimateNode is doing on using natural language processing (NLP) – the branch of AI tasked with enabling computers to understand human language – to harness insights on climate-related risks and impacts from unstructured data spread across thousands of documents. As part of this, it will explore the potential for Large Language Models to extract this information, and reflect on potential applications of NLP to improving our understanding of climate risk, including in monitoring how economic and human welfare risks are materialising, and in building qualitative scenarios that complement models.

Wildland-Urban Interface fires: very dangerous but not as unpredictable as we fear  Prof. Epaminondas Mastorakos Department of Engineering,, University of Cambridge

A new modelling framework has recently been developed at Cambridge University, which takes fully into account most physical phenomena present in fires propagating in the interface between built and forest areas. Such fires can have devastating consequences and have hitherto been very difficult to describe theoretically. The new model has been applied to various WUI fires with acceptable accuracy. Its potential use for fire-fighting, de-risking at the planning stage, and insurance purposes will be highlighted in the seminar.

Future Infrastructure and the Built Environment  Dr. Brian Sheil University of Cambridge 

Infrastructure is the key to unlocking net zero” is the unequivocal assessment of the Skidmore Review published this year. This talk introduces the EPSRC Centre for Doctoral Training (CDT) in Future Infrastructure and Built Environment 3 (FIBE3): Unlocking Net Zero, which is at the forefront of research and training in the fields of sustainable infrastructure, net-zero emissions, and digital technologies. This talk will explore how FIBE3 is addressing the urgent challenges of our time, including climate change and the need for resilient and environmentally responsible infrastructure. Finally, some examples of Cambridge research in this area will be presented