Location: Office of Associate Administrator
Project Number: 0500-00110-001-001-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Sep 30, 2022
End Date: Sep 29, 2024
Artificial intelligence (AI) technologies, and deep learning in particular, have numerous potential agricultural applications and are poised to transform many areas of agricultural research. However, successful, rigorous, reproducible, and ethical application of AI technologies to scientific research requires specialized training in a variety of skill areas. As part of a new, campus-wide AI initiative, the Cooperator’s Research Computing group has developed and continues to expand an AI training curriculum that is designed to help professionals, including research scientists, with no prior experience build the skills needed to use applied AI technologies in their own research and other work. The primary goals of this cooperative agreement are to 1) provide ARS staff with access to the AI trainings mentioned above; and 2) customize extant training courses to better meet ARS needs and develop new AI trainings on topics of particular interest to ARS staff. The resulting training courses and modules will be of benefit to both ARS and the Cooperator.
ARS has deep scientific expertise across a wide spectrum of agricultural research domains, and ARS scientists are pursuing numerous research questions that could benefit from modern AI techniques. ARS also has the scientific computing infrastructure to support compute-intensive AI methods, including multiple high-performance computing (HPC) clusters. The Cooperator has major research programs in agriculture and natural resources and, since 2020, a campus-wide initiative to integrate AI across the Cooperator’s research and education programs. This initiative includes the development of new AI-related training opportunities with a focus on professional development. Specifically, the Cooperator has developed a hands-on, Jupyter Notebook-based, AI training program called Practicum AI. The modules for this program specifically target newcomers to the field and those with limited coding experience and mathematical background. Practicum AI currently has two main series of courses: The Practicum AI Beginner Series: * Ethics and AI (about 1.5 hours) * Introduction to Python (about 4-5 hours) * Reproducibility with git and GitHub (about 1 hour) * Deep Learning Foundations (about 4 hours) The Practicum AI Intermediate Series: * Convolutional Neural Networks (3-4 hours) * Transfer Learning (1 hour) * Natural Language Processing (2 hours) * Recurrent Neural Networks (2 hours) * Transformers (3-4 hours) * Generative Adversarial Networks (2 hours) The Cooperator has offered these modules multiple times in-person, and continues to enhance their content based on student and instructor feedback. The Cooperator is in the process of beginning the incorporation of these courses into the offerings available through the Cooperator’s Professional and Workforce Development (PWD) online learning platform. After completing the courses in a series, students will earn a microcredential for their completion. The structure of Practicum AI is such that students can start with little background knowledge and, in the Beginner Series, gain the skills needed to start learning AI methods. Students can then progress to the Intermediate Series, where they are exposed to the variety of neural network architectures and datasets used today. More advanced modules will build from these foundations and keep learners engaged by further developing the skills needed for applying AI technologies to scientific research and other application areas. Work on this agreement will have four main components: 1) offering the Beginner and Intermediate Series courses as live trainings for ARS staff at least once per year; 2) providing ARS staff with access to course materials for asynchronous learning and self-study; 3) customizing extant training courses to better meet ARS needs; and 4) developing new, advanced Practicum AI courses covering topics of particular interest to ARS scientists.