AI for Climate & Nature

Cambridge researchers at Conservation Evidence and the Computing Department are using the UK’s most powerful supercomputer to create an AI-driven “Conservation CoPilot” to guide decision-makers towards the best solutions for conservation projects aimed at halting the alarming loss in global biodiversity. 

The CoPilot will harness the power of artificial intelligence technology (AI) to distil hundreds of millions of research papers into the Conservation Evidence database, which aims to capture the results of projects specific to as many landscapes and eco-systems as possible, from re-wetting peatland to sowing seeds for wild bird species. 

The United Nations Environment Programme (UNEP) says populations of fish, birds, mammals, amphibians and reptiles declined by 58 percent between 1970 and 2012, and some 25 percent of mammal species and 41 percent of amphibians are threatened with extinction right now. 

Conservationists are grappling for effective responses to the loss of habitats and biodiversity, fuelled by changes in land and sea use, pollution, global warming and invasive species. The global decline of biodiversity requires informed actions. 

“If you are going to carry out actions that will benefit biodiversity, or to reduce the impact of climate change, then you want those to be effective,” said Professor Bill Sutherland CBE FRS, who is Director of Research at the University of Cambridge Department of Zoology. 

In the past 30 years, their panels of expert scientists at Conservation Evidence have screened more than 1.6 million scientific papers, in 17 languages, to determine the effectiveness of 3,600 conservation measures. Building this kind of database currently requires a lot of human effort, approximately 75 years of human researchers’ time so far. 

The Conservation Evidence team, collaborating with members of the Computing Department, are harnessing the power of Cambridge’s Dawn supercomputer to build an AI model to speed this process up. 

A potential 300 million academic papers are available for the AI model to analyse, in collaboration with Office for Scholarly Communication. 

The AI model will be able to determine if certain research is relevant and condense it into an accessible summary so that the Conservation Evidence team can consider quicker whether it should be added to their website. 

“The aim of our work is to build this evidence base and create a culture of evidence use in conservation – we synthesise global evidence on actions for biodiversity, sorted by final outcome, for easy-to-access conservation recommendations.” said Dr Sam Reynolds, Department of Zoology, and member of the Conservation Evidence team. 

The team are going to build a conversational chatbot, they call the “Conservation CoPilot”. This tool would allow users to interact with the Conservation Evidence database through a chat interface – guiding decision-makers through the conservation process, providing relevant evidence at each step. 

“You will be able to say, ‘I have a site of this type with these species of interest and want to change the management to improve it’, then the chatbot will work out what is relevant for you in your particular area — whether you are in the uplands, or the lowlands, what species are there, what the soil type is — all of which will affect the CoPilot’s recommendations,” Sutherland said. 

This project is one of several projects under a new Cambridge focus on AI for Climate and Nature, which has received seed funding from ai@cam, the University’s flagship AI mission. 

“Our CoPilots are intended to augment human decision making rather than replace it. They will help conservation experts to categorise interventions much more quickly and accurately and accelerate the rate at which impactful conservation actions can be replicated and scale worldwide,” explained Professor Anil Madhavapeddy, Department of Computer Science, and a member of the team. 

AI for Climate and Nature is an interdisciplinary collaboration between Cambridge Zero, Cambridge Conservation Research Institute, Conservation Evidence, Institute of Computing for Climate Science, Centre for Landscape Regeneration, Cambridge Centre for Carbon Credits (4C), Cambridge Conservation Initiative, Cambridge Centre for Earth Observation and the Dawn supercomputer.