Decoding nature’s hidden patterns

What It Covers

MIT professor Sarah Biri and computational linguist Jeff Reed explain how AI tools process 200 million iNaturalist images and thousands of hours of Yellowstone wolf recordings to extract ecological data at scales previously impossible, accelerating conservation decisions for species from salmon to apex predators.

Key Questions Answered

• •Citizen Science Data Scale: iNaturalist's 200 million images now constitute roughly 90% of all biodiversity data ever collected, feeding into a global clearinghouse of over 3 billion species occurrence records. Researchers can leverage this by uploading observations through iNaturalist or eBird — every photo automatically becomes a georeferenced scientific record exportable to global databases.
• •InQUIRE Query System: MIT's InQUIRE tool lets ecologists search massive ecological image databases using plain-language questions — no training examples or code required. A researcher studying post-wildfire forest recovery used it to discover that severely burned areas regenerate more deciduous trees while less severe burns favor coniferous regrowth, informing climate-change forest modeling.
• •Energy-Efficient AI Design: Conservation AI is deliberately built small and computationally lean — designed to run on battery-powered field devices in remote locations without cloud connectivity or GPUs. This makes it far less energy-intensive than large generative models like ChatGPT, countering the assumption that AI-assisted conservation necessarily worsens the carbon footprint of research.
• •Wolf Pack Decision-Making via Audio: AI analysis of Yellowstone chorus howls reveals that dominant wolves trigger collective territorial howling — once the alpha joins, the pack follows. Separately, wolves assess rival pack size acoustically and decide whether to approach or retreat based on numerical advantage, a behavior now quantifiable through AI-assisted audio classification.
• •Targeted Conservation Timing: Bird migration through the entire United States concentrates into just a few days per year, during which hundreds of millions of birds die striking city windows. AI-powered migration detection enables city-wide lights-off campaigns timed precisely to those specific nights, delivering significant conservation impact with minimal disruption to residents and businesses.