Bringing Artificial Intelligence to Agriculture

By Elliot Meyerson and Risto Miikkulainen

Today, we’re pleased to share some promising early results from our newest project: working with the Open Agriculture Initiative (OpenAg), with the goal of optimizing crop yield and transforming the future of food production.

The folks at OpenAg’s lab have the full details, but the idea is this: what if you could create the optimal growing conditions for crops? Not only that, but what if you could contain those crops in controlled environments so you could not only measure every relevant nutrient and variable but also, eventually, grow optimized crops anywhere?

It works like this: First, the group used Food Computers (those contained growing environments we mentioned above) to grow generations of a crop (in this case basil). Elements like UV light, salinity, heat, water stress, and more are controlled in these experiments and the crop yield is analyzed. That’s where Sentient comes in. We use the data from those first generations of basil to model how different conditions would affect the plant in further experiments, coming up with new recipes to try.

Image Credit: MIT Media Lab Food Computer

How does the AI work? We’re glad you asked.

Our scientific approach is based on optimizing a surrogate model of how the plant grows under different recipes. First, when the number of samples is small and only a few actuators (UV light types in the first instance, etc.) are varied, Gaussian processes can be used to predict what the outcome would be given a new recipe, and Bayesian optimization used to create suggestions for good recipes. Later on, as the dimensionality and number of samples grow, we use a neural network as the model and evolutionary algorithms (population-based or EDA) as the optimization method. The suggestions are then tested in the FCs and the containers, and the results used to train the model further. In this manner, the model gradually improves, is more effective and novel recipes are developed. As a matter of fact, the AI has already rediscovered a known tradeoff between weight and flavor, and a surprising new result that perpetual light may help some plants produce more flavor!

An illustration of the surrogate model and the recipes suggested by the AI. The three axes correspond to the three actuators and the color of the small dots indicates their value predicted by the model (i.e. flavor; red > yellow > green > blue). The large dots are suggestions, and the darker dots are the most recent ones. They suggest utilizing long photoperiods and UV periods, the success of which was confirmed in growth experiments by the OpenAg team.

We’re really proud of this project. Please check out OpenAg’s post, for more color on the project and stay tuned for updates!