How does this even work?
The lack of explainability of conventional machine learning and artificial intelligence is a well-known flaw of this otherwise powerful tool.
Most popular ML/AI models, that's Machine Language/Artificial Intelligence - including ChatGPT and the like, are black box models. As you may have seen yourself these sometimes spill out completely crazy results.
Because of the black box, we can't decipher how and why such results were produced or how they can be prevented in the future, even for the people designing the algorithms. You want answers, not more questions.
What is ML/AI
Explainable ML/AI is a special type of machine learning and artificial intelligence that aims to not only give an accurate prediction but also provide an explanation of the prediction, which is as important if not more so than the prediction itself, particularly in agriculture.
Most existing approaches to explainable ML/AI attempt to create another model to make sense of the black box model rather than making major changes to the black box model itself.
Crop Convergence takes a different approach
We use explainable ML/AI by completely replacing the black box modeling structure, which typically consists of millions or billions of parameters that don't have any physical meaning, with a few hundred carefully selected parameters that represent all the major physiological processes of the crop growth cycle.
As such, the Crop Convergence model is uniquely capable of predicting not only the final crop yield, as most of the other ML/AI models do, but also all the major physiological processes in every hour of the growth cycle that ultimately lead to the crop yield.
The applications of the this approach to explainable ML/AI are endless in agriculture. For example, when a conventional ML/AI model may predict low yield, the Crop Convergence approach can explain specific climate variables that will lead to the low yield and what changes to the recipe can most effectively alleviate the environmental stress and improve the yield.
Explainability can provide trust in the prediction results and enable growers to identify the most promising recipes that lead to higher yield, lower energy consumption, longer shelf life, and higher return on investment.
Comments