M is for Mineral

Editor's note: This post was first published on the Mineral blog.

After five years incubating our technology at X, Alphabet’s moonshot factory, Mineral is now an Alphabet company. Our mission is to help scale sustainable agriculture. We’re doing this by developing a platform and tools that help gather, organize, and understand never-before known or understood information about the plant world - and make it useful and actionable. Together with our partners across the food production system, we're hopeful that these tools will - over time - drive a deeper understanding of the complex interactions of plant genes, the environment, and farm management practices.

Why agriculture?  Why now?  Agriculture is increasingly believed to be a major contributor to the climate crisis - but it is also a victim of a changing climate.  There is no time to waste to find more climate-resilient crop varieties, to transition to less chemical- and fossil fuel-intensive practices, to improve soil health, and to restore biodiversity.  We are confident the foundational technology (such as generative artificial intelligence, machine learning and edge compute power) has matured to the point where we can solve problems at the scale needed for production agriculture and with the accuracy and speed demanded by farmers.

Our focus is in three key areas: developing sensing technology that can generate rich data sets about plants, organizing agriculture data from disparate sources for machine learning (ML) and building powerful software algorithms, and conducting research that can meaningfully advance our fundamental understanding of plantkind.

_{Mineral’s perception technology individually segmenting sugar beet leaves}

We’re partnering with companies across the food production system, alongside leading agribusinesses, research organizations, and farmer groups with deep domain expertise. These partners help shape and inform our focus and bring speed, precision, and scale to our efforts.

We're early in our journey, but have already learned lessons we think are worth sharing:

• We found that most companies are not collecting the quantity, diversity or quality of data needed to take full advantage of machine learning.  That’s why we built tools to better capture, curate, clean and augment multimodal data; and assembled our own bootstrap ag dataset.
• There is no single mode of data collection suited to every agriculture task or crop.  We began with a plant rover that could capture huge quantities of high quality images, and over time expanded to building generalized perception technology that can work across platforms such as robots, third party farm equipment, drones, sentinel devices, and mobile phones.
• ML in agriculture is empowering and augmenting human experts - not replacing them.  We have seen how patterns and correlations generated by our algorithms are helping our partners discover new knowledge and gain a deeper understanding.
• Software can help overcome the technology accessibility barriers that have historically held back farmers worldwide.  While high performing seeds, fertilizer, and equipment can be hard for farmers to find - data, advice, and the power of ML will reach anyone with access to a smartphone.

The passion and urgency in our work is because the clock is ticking on improving farmland productivity and sustainability. Through the Mineral blog we'll aim to share our progress (and learnings) and perspectives, so please follow us if you’d like to stay informed.

Editor's note: Read about Mineral's incubation and journey at X here.