I am interested in the use of Large Language Models (LLMs) for conservation, specifically in how they can be improved with Retrieval-Augmented Generation (RAG) and Graph-based RAG.

• LLMs (Large Language Models) are AI systems trained on vast amounts of text data, but they often lack real-time access to specific, reliable knowledge—which is critical in conservation.

• RAG (Retrieval-Augmented Generation) helps LLMs by allowing them to fetch relevant, up-to-date information from trusted sources instead of relying solely on their pre-trained knowledge.

• Graph-based RAG goes further by structuring knowledge as a graph (a network of interconnected entities like species, habitats, and policies), making AI more context-aware and better at reasoning over complex ecological relationships.

I also believe that in the future, traditional interfaces (such as websites and apps) will be replaced by AI agents—intelligent systems that interact autonomously with users and the environment. My concern is:

How do we ensure these AI agents serve biodiversity rather than just human or corporate interests?

To make AI trustworthy and ethical in conservation, we need to:

1. Ensure AI is aligned with biodiversity goals, not just human-centered objectives.

2. Incorporate ecological knowledge from diverse sources, including scientific data and indigenous wisdom.

3. Design AI that is transparent and explainable, so users can trust how it reaches decisions.

4. Minimize AI’s environmental footprint, making it energy-efficient and non-extractive.

Finally, how can we build AI that respects all living beings, not just humans? What principles should guide the development of ethically sound, biodiversity-centered AI?

Kind regards