Living Data, Living Systems

Members of the AI and Biodiversity Change (ABC) Global Center took the stage at Living Data 2025 (Oct. 20–24, 2025), a community convening focused on making biodiversity data FAIR interoperable, and actionable for science and policy.

The team contributed across sessions: François Leroy presented on data-driven workflows and standards; Catherine Villeneuve spoke in the modeling track; David Rolnick joined the AI session to discuss Antenna and scalable ML for ecology; and Mélisande Teng (Mila) earned a Best Talk award for a machine-learning approach to species distribution modeling that combines Sentinel-2 satellite imagery with citizen science observations (eBird, eButterfly). Teng’s talk highlighted how multi-source data can fill gaps in where species live, how they move, and how habitats change.

The conference emphasizes shared infrastructure and standards—the connective tissue that allows images, audio, and remote sensing to flow into models and back into decisions. That aligns with ABC priorities: explainable AI, reproducible pipelines, and data practices that support both discovery and conservation.

The group also engaged discussions on interpretability (how models make predictions), ethics and authorship in cross-disciplinary publishing, and practical steps to make datasets AI-ready—from metadata and licensing to benchmark design.

Building on Montreal momentum, the teams will continue collaborating on standards, open datasets, and teaching materials that lower barriers to entry—so ecologists, data scientists, and communities can co-produceknowledge. The message from Living Data 2025 was clear: better data, better models, better outcomes for nature.