MIT Uses Machine Learning to Grow Super-Flavored Basil Plants
MIT researchers use machine learning to optimize hydroponic growing conditions, creating basil with significantly stronger flavors. The experiment shows how cyber-agriculture can automate and improve farming by manipulating environmental variables.
Researchers at MIT's Media Lab and the University of Texas at Austin use machine learning to optimize the growth of basil plants, resulting in significantly stronger flavors. This approach to "cyber-agriculture" aims to improve and automate farming by treating the growing environment as a complex system with numerous adjustable variables, such as light duration and watering frequency, that directly affect measurable outcomes like flavor-producing molecules.
The team limits the machine learning model to analyzing and adjusting the light regimens experienced by the plants to increase flavor concentration. After an initial round of growing basil with traditional light schedules designed by hand, a simple algorithm uses those results to suggest slight tweaks, followed by a more sophisticated model that is given greater freedom to recommend extreme environmental changes.
To the surprise of the researchers, the advanced model recommends leaving the UV lights on the basil plants 24 hours a day, a highly unusual condition that does not occur in nature. This unexpected strategy successfully produces basil with double the flavor concentration of normal plants, proving that artificial intelligence can discover highly effective growing strategies that human intuition would likely never consider.