How AI is unearthing hidden scientific knowledge | Sara Beery

Episode Summary

Executive Summary: Ecologist and AI researcher Sarah Beery discusses how AI can help unlock hidden insights from vast ecological databases, such as iNaturalist's 300 million images, to accelerate scientific discovery and conservation. She introduces 'Inquire,' a system that allows scientists to ask questions directly to databases, reducing analysis time from thousands of hours to just a few hours, and emphasizes the urgency of understanding and protecting the 80% of Earth's species that remain unknown.

Main Topics: The Biodiversity Crisis and Knowledge Gap (Priority: 5/5): Scientists estimate 10 million species exist, but only 2 million have been observed, leaving 80% unknown. Extinction rates are 100-1,000 times higher than historical baselines, and species like the Tapanuli orangutan are discovered only to be critically endangered. AI as a Tool for Ecological Discovery (Priority: 5/5): AI can analyze massive datasets (e.g., 300 million iNaturalist images) quickly, but traditional AI requires time-consuming training. Beery's team developed 'Inquire,' which uses AI to compare images with scientific language, enabling direct queries without coding or example collection. Hidden Knowledge in Ecological Databases (Priority: 4/5): Beyond species identification, images contain data on individual animals, behaviors, vegetation, and food chains. Platforms like iNaturalist, Xenocanto, and FathomNet hold untapped ecological goldmines that AI can help access. Case Study: Bird Diets and Seasonal Changes (Priority: 4/5): Using Inquire, a team found thousands of examples of birds eating various foods in hours, revealing that some birds (e.g., American robin) eat insects in winter but less than summer, while others (e.g., American tree sparrow) stop entirely. This matched a manual study that took 1,560 hours. Future Potential and Public Participation (Priority: 3/5): The approach can extend to bioacoustics, satellite data, and GPS tracks. Public contributions to platforms like iNaturalist are crucial for building a complete picture of life on Earth and supporting conservation actions.

Key Arguments: Traditional data collection is too slow to address the current biodiversity crisis; AI can exponentially increase ecological knowledge. Current AI methods require collecting hundreds of examples for each new query, which is still too slow; interactive systems like Inquire allow direct questioning. Inquire reduces analysis time from 1,560 hours to 3 hours for similar studies, demonstrating efficiency and accuracy. Maximizing value from existing data helps identify knowledge gaps and strategically allocate resources for new data collection. Public participation in data collection (e.g., uploading photos to iNaturalist) is essential for conservation.

Data Points: Estimated total species on Earth: 10 million - Scientists estimate this number, but only 2 million have been observed. Unknown species percentage: 80% - 8 million species remain unknown to science. Extinction rate increase: 100 to 1,000 times higher - Compared to historical baselines based on past data. Images on iNaturalist: 300 million - Uploaded by volunteers, each with species identification. Time to view all iNaturalist images manually: 40 years - Assuming one second per image, full-time work. Time for Inquire bird diet study: 3 hours - From question to answer, finding thousands of examples. Time for manual bird diet study: 1,560 hours - Manual curation for a similar study.

Pivotal Quotes: "Imagine you're a doctor and you're trying to save the life of a patient, but you can only see a fifth of their body. How are you going to prescribe medicine? How are you going to do surgery? See, this is exactly the situation we're in with nature across the planet." — Sarah Beery: Opening analogy to illustrate the lack of knowledge about Earth's biodiversity. "We stand at a unique point in history. We have both an unprecedented biodiversity crisis, but we also have unprecedented tools to address it." — Sarah Beery: Conclusion emphasizing the opportunity to use AI for conservation. "The future of conservation doesn't just lie in remote rainforests or deep ocean trenches. The future of conservation is hiding in our ecological databases, both the ones we have now, but also the ones we have yet to collect." — Sarah Beery: Call to action highlighting the value of existing and future data.

Implications: This talk empowers scientists and the public to leverage AI for faster ecological insights, potentially transforming conservation strategies. It highlights the need for accessible AI tools and public participation in data collection to address the biodiversity crisis efficiently.