84 The Adoption of AI in the Core Scientific Cycle of Feed Research

Abstract The Gartner Hype Cycle methodology highlights five self-explanatory steps through which a technological innovation needs to go: (1) innovation trigger, (2) peak of inflated expectations, (3) trough of disillusionment, (4) slope of enlightenment, and (5) plateau of productivity. It is between step 4 and 5 that any tech innovation really starts to commercially pay off. For Artificial Intelligence (AI), the possibilities and challenges are so diverse that several separate cycles now exist for distinct parts of AI development. For the Animal Sciences / Feed industry the application of AI is not straightforward. In fact, in most industries, application is not straightforward because applications equals implementation. Hence, being able to translate the pains and gains of your customer into the models that you create, and finding a way to implement it, is key and that is more than just applying AI. The majority of (animal) models in the feed industry are mechanistic by nature. Parameters are generated/updated via controlled experiments, and stochasticity is allowed via Monte Carlo simulations and scenario analyses. The validity of the model is its ability to provide usable growth/health predictions, enable least cost formulation, and provide a sustainability footprint. To further support such models, and to offer new services, we have recently begun to combine our near-infrared spectroscopy(NIR), laboratory information management system (LIMS) and mycotoxin databases with climate and geographic data to (1) predict the nutrient composition of raw materials over time, (2) enable risk assessment of mycotoxin co-contamination, (3) improve the feed evaluation of silages, (4) estimate a full nutrient profile, and (5) improve the precision of net energy estimation. In the end, if nurtured carefully, AI is just another technical leap that needs to be integrated into the core scientific cycle.