My project will merge the burgeoning field of artificial-intelligence-driven language models, and sustainable agriculture. In the face of climate change, a growing global population, and increasingly detrimental agricultural practices, breeders of staple crops are at the helm of making sure that farmers have the cultivars they need to respond to changing climate conditions, increase the yield of their crops, and limit their use of harmful fertilizers, pesticides and herbicides. For several decades, plant breeders have used genomic data to estimate the value of specific plants in terms of its use in developing a new cultivar. This process is called genomic prediction, and breeders of staple crops have been trying to use neural networks and machine learning for the task of genomic prediction with limited success. In the hopes of achieving greater success using neural networks for genomic prediction, I would like to turn the task of genomic prediction into a translation question. Can we take genetic marker data and effectively translate it into a description of estimated phenotypes for use in plant breeding programs, to help meet the needs of farmers around the world?