Foundation Models For Astrobiology – NASA-DARES 2025 White Paper

A white paper for the 2025 NASA Decadal Astrobiology Research and Exploration Strategy (DARES)

In this RFI response we present preliminary recommendations from a workshop on Foundation Models (FMs) in Astrobiology that was hosted by the NASA Ames Research Center and the SETI Institute in Mountain View, CA, February 24-26, 2025. The goal of the workshop was to assess FMs towards applications in astrobiology and to produce a number of white papers to guide the community.

A FM [1] is a large-scale machine learning (ML) system typically trained on enormous datasets to encode fundamental, general information and relationships, enabling it to serve as a foundation for various downstream applications with limited additional training and fine-tuning.

While building such large-scale FMs requires a great deal of ML expertise and effort, adapting them for specific downstream tasks is often fast and requires minimal ML expertise, making them excellent tools for rapid development in fields that call for a wide range of applications. For astrobiology, FMs may offer unique and critical opportunities for advancing efforts in life detection and characterization.

In the last few years, FMs have emerged as a new paradigm that can dramatically accelerate the application of ML to specialized tasks and a wider range and types of data [2, 3].

Full paper below

Astrobiology