Mineral

Discovering the intelligence of plantkind to feed and protect humankind

Graduated, Alphabet Company/2023 - 2024

The Mineral team developed a range of breakthroughs in artificial intelligence and perception technology for the agriculture sector to build a more sustainable, resilient, and productive food system. The team incubated at X for five years, where they developed a novel learning platform and AI models. To understand the sector’s challenges and build agriculture-specific tools, the team worked closely with partners around the world including the Alliance of Biodiversity International and CIAT. In 2024 Mineral’s technology acquired by Driscoll’s and John Deere.

The Mineral rover was a learning prototype – it used breakthroughs in artificial intelligence, sensors, and robotics to find ways to grow more food, more sustainably.

Modern Agriculture’s Vicious Cycle

To feed the planet’s growing population, global agriculture will need to produce more food in the next 50 years than in the previous 10,000 – at a time when climate change is making our crops less productive.

Until now, the world’s approach to meeting this challenge has been to standardize what we grow and how we grow it. Modern agriculture practices focus on cultivating a few crops known to have high yields — today, rice, wheat, and maize provide nearly half the world’s plant-derived calories.

We also standardize how we manage the crops we grow — most crops are treated uniformly on a per acre or per hundred acre basis with chemicals for issues like pests, weeds, and disease and fertilizers. But an agriculture system that’s optimized for productivity and simplicity comes with risks.

Intensively growing just a few varieties of plants makes our food supply vulnerable to pests, disease, and a changing climate. Over time, it also depletes the soil of nutrients and minerals, reduces the diversity of the soil’s microbiome, and diminishes the soil’s ability to store carbon. Overuse of fertilizers and chemicals also negatively affects soil health, creating a vicious cycle that makes our farmlands less productive and the food we grow less nutritious.

The team conducting research in strawberry fields in Northern California

An “Operating Manual” for Plants

Project Mineral started with the insight that to sustainably grow more food on a global scale, new tools and insights will be needed to manage the staggering complexity of farming. The team started their journey by talking with breeders and growers around the world to learn about the challenges they face.

From soybean farmers in Argentina to kiwifruit breeders in New Zealand, they heard that they need to gather much more information on many more varieties of biodiverse plants—and quickly, if they are going to find varieties that are resilient and productive in the face of climate change.

Growers face hundreds of decisions every season, yet current tools aren’t equipping them to meet the challenges they face. Even though they use digital tools like sensors, spreadsheets, and GPS, their data is either siloed or doesn’t fully represent agriculture’s complexity.

Mineral saw an opportunity to build new software and hardware tools that can bring together diverse sources of information that until now were simply too complex or overwhelming to be useful.

Building on breakthroughs in artificial intelligence, machine learning, simulation, perception, and robotics, Mineral set out to build the world’s first detailed “operating manual” for plants. Just as the microscope changed how diseases are detected and managed, Mineral hoped that by seeing and understanding the plant world in a radically different way, we could usher in a new era of sustainable agriculture.

The team ran early experiments at the X lab, imaging individual bok choy plants to monitor features like leaf size and biomass.

The team moved out of the lab to capture plant images in the real world. The first plant rover prototype, made out of two bicycles and some Pixel2 phones, captured images of each plant as it was pushed through strawberry fields.

The rover soon evolved to capture high quality images of individual plants throughout the growing season, and to handle a wide range of conditions like muddy fields and bright sunlight.

Next, the team experimented with a remote controlled rover with controlled rolling speeds and more sophisticated sensors. It could photograph plants from multiple angles.

After collecting millions of images, the team created a rover that was resilient to the elements, had controlled illumination, and could collect terabytes of high quality images mapped to precise locations in the field.

The current version of the plant rover is designed to roll through fields with different kinds of crops. There are sophisticated cameras and machine perception tools inside the rover and it is equipped with solar panels.

Integrating Into the Industry

To uncover critical new details about how crops are grown and food is produced, the team developed a range of prototypes. These included the Mineral rover, which rolled through the fields gathering high-quality images of each plant. By combining the imagery gathered by the rover with other data sets like satellite imagery, weather data, and soil information, the team created a full picture of what happened in the field and used machine learning to identify patterns and useful insights into how plants grow and interact with their environment.

In 2024, Mineral transferred its advanced AI tools for crop phenotyping, yield forecasting, quality inspections, and food waste reduction to Driscoll’s, the world’s leading berry company. The transition aimed to maximize the impact of Mineral’s innovations on global agriculture to make it more sustainable and resilient.