Foundation models for agricultural sciences - Challenges and opportunities

Ioannis N. Athanasiadis

Keywords: agricultural sciences, foundation models

TL;DR: Foundation models for advancing agricultural sciences

Abstract:

Foundation models, developed with self-supervised methods using unlabeled and multimodal multidisciplinary corpora, offer a new pathway for AI-driven discoveries in agricultural sciences. In this paper, we identify key challenges inherent to AI research in agricultural sciences, and advocate that an emerging new generation of agricultural AI, in the form of foundation models is necessary to kickstart data-driven discoveries in agricultural sciences.
We argue that foundation models have the potential to act as jumpboards and enable a new generation of AI models for agricultural sciences, offering power and adaptability to accelerate innovation. By reflecting on agricultural science challenges, data, and theory, we identify pathways for model development and evaluation that will enable agricultural researchers to build on shared foundations instead of training new models from scratch for each application and location.

Submission Number: 123