“ They ’ re little microphones that you can put out in the field that create an enormous amount of acoustic data .”
Drew ’ s team has created an adaptable foundational model over the data that allows conservationists to use it for many purposes , including identifying species .
“ Say people that are working in the tropics have very long lists of very rare species but not many recordings . We ’ ve developed automatic ID systems by adapting that foundational model with a small amount of data . So that ’ s a key example of using AI to bring in more observations and data from nature , or monitoring nature .”
Google Deepmind is also developing a similar project with motivation activated cameras . The foundational models support the identification of the animals turning up in those images and vastly reduce the amount of data that needs to be manually trawled through .
Drew narrows AI ’ s use in acoustic monitoring cases into three stages :
• Deploying AI to bring in more data and monitor nature
• Integrating all that data together to be able to derive actionable information out of it
• Combining the information with other geographic information to predict where species are
92 August 2024