Crop yield depends on traits that interact in multiple dimensions -growth stage, plant architecture, growing environment, just to name a few-. Until now this has precluded ‘precision breeding’ for genetically complex trats to boost, for example, abiotic stress tolerance, input use efficiency and yield potential. Significant genetic variation for such traits among elite breeding lines testify to this. However, by combing ever more accessible and powerful technologies -remote sensing, gene sequencing and AI-, deterministic breeding is feasible.

A series of wiring diagrams (WD) representing trait interactions over time for wheat -based on comprehensive review of empirical data and hypotheses consistently reinforced by crop models- provide a starting framework for AI-assisted simulation of different crossing strategies, at a scale previously unimaginable. With so many traits and interactions involved, AI is the technology most likely to enable genetic control of complex traits. Iterative AI-assisted simulations could also be trained to identify key knowledge gaps, as well as ‘rate-limiting’ traits to target genetic resource mining.

A vast reserve of genetic diversity is available in collections worldwide and even the small fraction characterized to date have generated massive benefits for farmers and consumers worldwide. Crop wild relatives, having survived millions of years of environmental flux, are probable sources of badly-needed adaptive traits.

AI-driven crop modelling approaches that integrate trait, gene and environmental data as proposed, is currently our best-bet to ‘future proof’ crops given the wealth of untapped genetic diversity available.