67 Leveraging Ai/Ml to Address Critical Challenges in Livestock Research

Abstract

Abstract In recent years the scientific community has witnessed extraordinary growth in the development and application of techniques from Artificial Intelligence/Machine Learning (AI/ML) to address pressing societal challenges. Contrary to more traditional linear models, some AI/ML approaches suffer from a general lack of interpretability and in disciplines/industries where understanding biological relationships among predictor and predictand is important, utility and even trust in resulting predictions is a potential issue. Regardless, the livestock research community has shown considerable interest in the utilization of AI/ML methods, coupled with advanced cyberinfrastructure, to address large-scale data challenges from the use of sensors and internet of things (IoT) to solve problems related to animal identification, welfare, behavior detection, genome-to-phenome, and data quality concerns from large-scale non-research sources (to name a few). The effective application of AI/ML to problems in livestock research requires an understanding of what methods are available; the benefits and pitfalls of these methods; what resources are required including data and metadata, cyberinfrastructure, and interdisciplinary skills; best practices; and how AI/ML methods compare with alternative approaches. Awareness of the above-mentioned items is critical to avoid reproducibility issues and the possibility of using overly complex models for simple tasks. Each of these topics will be addressed, including a gentle introduction to AI/ML, in the context of examples from livestock research and a view towards future developments and application.