Multimodal AI and biodiversity

  • * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

  • Date
    24 June 2025
    Timeframe
    16:00 - 17:00 CEST Geneva
    Duration
    60 minutes
    Share this session

    Recent developments in multimodal AI have unlocked diverse and advanced capabilities across sectors, from healthcare to urban planning to education. In recent years, our community has begun to explore the transformative potential of multimodal AI systems in addressing global biodiversity challenges. By integrating diverse data streams, including visual imagery, acoustic recordings, genomic sequences, environmental sensor networks, and human language, multimodal AI promises unprecedented capabilities for monitoring, analyzing, and conserving biodiversity across terrestrial, marine, and freshwater ecosystems. However, these systems can also be brittle to use, fail catastrophically, be heavily biased, and be difficult to evaluate. We will summarize several recent and exciting works that explore the use of multimodal AI for diverse application in biodiversity monitoring and understanding, from enabling flexible and interactive search over large image repositories like iNaturalist, to training flexible multimodal representations across everything from satellites to sound that capture fine-grained geospatial variation, to evaluating the ecological knowledge contained in large multimodal generative AI models. We will conclude by looking to the future, with a discussion of short- and long-term priorities for the research community towards the goal of making optimal use of the many heterogeneous and complimentary data streams we currently collect to monitor biodiversity worldwide.

    Learning Objectives:

    1. Understanding the current landscape of multimodal AI for Biodiversity
    2. Analyzing the key challenges that must be faced to recognize the potential of multimodal AI in Biodiversity

    Recommended Mastery Level / Prerequisites:
    None