Combatting Food Shortages—Have No Fear, Machine Learning Is Here

What do you do when you have access to satellite imagery of the entire world and a mission to better understand the planet? You have the information needed to answer many important questions, but need a path to unlock those solutions: machine learning acts as the key to accomplish your goal.

What do you do when you have access to satellite imagery of the entire world and a mission to better understand the planet [1]? You have the information needed to answer many important questions, but need a path to unlock those solutions: machine learning acts as the key to accomplish your goal.

Founded in 2014, Descartes Lab had a simple vision of making, “it easy to build models on top of huge amounts of geospatial data” [2]. The lab collects five terabytes of satellite data daily; that is 2.5 billion printed pages, stacking to over 100 miles high [3][4]. Given the extensive nature of the data and the causal relationship the lab was seeking, machine learning was the ideal megatrend for Descartes Lab to build a business around [5].

Descartes Labs, in its four shorts years, through building this database, has partnered with clients to address some of the world’s most crucial issues. One of these problems is global food security.

Background on the Food Scarcity Problem

In the next 30 years the population of the world is set to increase by 40%, resulting in the need for 70% more food [6]. Until recently, addressing the issue of food scarcity in the context of population growth seemed unobtainable. Satellite imagery has been widely recognized as the third agricultural revolution and, by analyzing with machine learning, the main weapon to combat the food scarcity issue. Through combining many geospatial data sources, Descartes is able to create a complete picture of the state of the world today and by using machine learning better able to detect causal patterns in that dataset. [10]

Evolution of farming through time:



Descartes Approach

Short Term (next 2 years)—Finding the Right Projects/Partners

Descartes has worked on several key projects that are specifically targeted at addressing the food scarcity issue:

  1. Production Forecast (Initial Success)

The lab, starting with corn, spent much time analyzing crop data through machine learning to put together an algorithm that predicted crop yields. With the algorithm, Descartes was able to predict production yields within two percent of the USDA numbers, but was able to do so five months before the USDA numbers were released in 2016. The data is historically collected through USDA surveys sent monthly to farmers, but by moving the satellite date and analyzing through machine learning, Descartes is able to produce weekly reports.  [3]

  1. DARPA Grant (Current Project)

Defense Advanced Research Projects Agency, DARPA, has invested over $7M in Descartes to fund a project meant to predict food supply in Africa and the Middle East. Understanding that food scarcity leads to political unrest, the U.S. government has tasked Descartes with analyzing the regions wheat production to create models that can be proactive in predicting scarcity problems. [7]

What Next

  1. Developing more Sophisticated Predictions—Weather

Descartes realizes that a key next step is better predicting natural disasters.  At the end of July, the lab released a wildfire tracker as the first of hopefully many tools to help predict the movement of natural disasters and alert those most at risk [8].

What Else

Although extremely successful in its ventures thus far, Descartes has several other avenues they could choose to pursue that could further improve their ability to address food scarcity:

  1. Increase partnerships with firms that could leverage current findings

Descartes could partner with firms like Indigo, “a tech startup in Boston, Massachusetts, [that]makes seed treatments that help plants grow”, and other agriculture businesses to utilize product-information synergies to bring the lab’s crop findings to the market [9].

  1. Build capital

When information is virtually unlimited like it is for Descartes, the limiting resource then becomes capital. In a world where many institutional investors use the USDA production data to make trading decisions, with not much effort, Descartes could leverage its own predictive models to make better informed investment decisions.  If their models are correct, these decisions will offer greater returns and thus provide capital to fund the lab. This may relieve some of the pressures Descartes has to charge for its models, assuming these models are outside the ones fueling the investments.

Going Forward

There are still questions that remain about the potential negative consequences of Descartes’ predictive models from such a powerful, global dataset. Questions to think about:

  1. Knowledge is power—balancing profits vs. safety
    • How do you make sure Descartes machine learning findings in food scarcity issues are not used as tools to hurt societies?
  2. Market concentration
    • There is now a large competitive advantage for companies who have access to these findings: how do you make sure the data is not used to eliminate competition and create market inefficiencies?


Word count: 800


[1] “Descartes Labs: Advancing global food security,” Google Cloud,, accessed Nov 12, 2018


[2] Anusuya Datta, Meet the man who created first living atlas of the planet – Mark Johnson, Descartes Labs,” Geospatial World, May 19, 2018,, accessed Nov 12, 2018


[3] Daniel Oberhaus, “Trippy Satellite Imagery Could Forecast Food Crises Before They Happen,” Motherboard, Dec 10, 2016, accessed Nov 12, 2018


[4] “How to Visualize Data,” Simply Ted, Dec 8, 2005,, accessed Nov 12, 2018


[5] A. Fedyk. How to tell if machine learning can solve your business problem. Harvard Business Review Digital Articles (November 15, 2016).


[6] Rob Dongoski and Andrew Selck, Digital agriculture: helping to feed a growing world,” Feb 23, 2017,, accessed Nov 12, 2018


[7] Emma Kennedy, “AI companies spot a business opportunity in space,” CNN, Apr 6, 2018,, accessed Nov 12, 2018


[8] Joe Evans, “ARPA SBIR: Descartes Lab Satellite Imagery Analysis,” YouTube, Oct 11, 2018,, accessed Nov 12, 2018


[9] “@WildfireSignal Tracking System Now Live on Twitter,” A Medium Corporation, July 18, 2018,, accessed Nov 12, 2018


[10] Jeff Kart, “Indigo Ag’s Natural Seed-Coating Technology Helps Plants Thrive,” Forbes, Sep 24, 2018,, accessed Nov 12, 2018


Predix at GE: Machine Learning in Industrial IoT


Under Armour: Can it Continue to Customize and Expand its Line of 3D-Printed Midsole Performance Footwear?

Student comments on Combatting Food Shortages—Have No Fear, Machine Learning Is Here

  1. What an interesting piece! I wonder how this knowledge would affect the individual farmers who own and lives off their own farmland, would this the new knowledge enabled by machine learning allow farmers to grow and farm more efficiently on their own farmland?
    To the question posed, in my view, one way to combat potential negative consequence of findings being used as tools to hurt societies is to make this information public. If combating global food shortage is the main objective, sharing this data openly could foster collaboration between companies and governments to address global food challenge. Descartes could potentially become a service provider in this space, receiving less profit from a bigger customer group rather than selling at a high market to selected few.

  2. Great piece – informative and also changed my perspective on the issue. In terms of market concentration, I agree with the author that this technology will only increase the leverage and advantages of large farms in the agriculture industry. Consolidation is happening rapidly in the US and spreading to the developing world as companies continue to learn how economies of scale impact the business. While many folks would argue to protect “small farmers” as a market economist who believes in efficient markets, if geospatial data enables large farmers to realize higher yields and lower risks, I argue in favor of the consolidation. The role of regulators will be to make sure that those efficiencies are passed onto consumers and not used to price gauge consumers.

  3. This is a really cool application of AI. I see enormous potential in your point around partnerships with other companies looking to solve similar social problems using AI, for example with “precision agriculture” which is the technology that links sensor data on the ground to cloud service providers who analyze it. This could help in not only the food space to help farmers increase yield, but also in the space you mentioned around predicting and preventing natural disasters. I think the government will have to play a huge role to make sure the public policy environment’s rule will fit the transition to an AI future

  4. Very interesting! Given that machine learning identifies probabilistic rather than causal relationships, I wonder if Descartes will face challenges in forming partnerships or getting product adoption as it transitions from being a predictive product to a prescriptive one

  5. Really enjoyed your article. I found the bit about DARPA funding particularly interesting. I think it speaks to the increasing need for the United States to proactively deal with global food security issues in order to mitigate conflict.

  6. Fantastic piece on a very interesting topic. I think that there are several ways to ensure that machine learning findings in food scarcity issues are not used as tools to hurt societies. The two main focus points should be the democratization of information and government regulation. To elaborate further, information on such critical undertaking should be made available to the public. This reduces the risk of the improper use of information, but also invites an element of crowdsourcing to solving one of humanity’s most pressing problems. Government regulation, while often derided, could be another effective tool considering the wide impact food scarcity will have on society.

  7. Interesting and clearly organized essay on a hugely important topic! I didn’t know the scale to which farmers will have to grow in terms of feeding people, and before TOM it wouldn’t have been as intuitive to me how important forecasting is to efficiency. The point on partnerships was also particularly insightful — Beyond Indigo I wonder if there are also other ways to partner with organizations that have shorter lead-times for producing food and can be more dynamic in adjusting to demand and unexpected events like disasters.

  8. Interesting article! I think the last point you mentioned on using this tool for good, not evil, is a really pertinent one. I’d like to believe that the solution is transparency. I think Descartes should make a choice to disclose what companies/governments/organizations are using their findings. If the goal of their company is truly to make this data available to improve societal issues, ideally management will want to ensure that their customers (whoever purchases their data/findings) are operating in line with that. To that end, I think making it publicly known who has access to this information will incentivize those customers to use it wisely and ethically.

  9. This was very interesting to read – what a creative way to use satellite imagery! It is very impressive that the company is able to aggregate so much data and distill the information into predictive models. I can see how this could help societies plan for potential food shortages, but I am having a harder time understanding how this technology could actually improve the food yield of farmers. The statistics you have shared about the increase in food production required for our growing population are shocking! I think that this is a very important issue that we need to solve so I am glad to know that companies like Descartes are working towards a solution.

  10. Awesome article! The topic of food security is near and dear to my heart. The Descartes solution reminds of a similar company in the space – Gro Intelligence. That company has found commercial success by finding a wide variety of clients for its satellite-driven data analytics on weather patterns’ impact on production. It helps commodity traders predict prices, governments form agricultural budgeting policy, and financial institutions lend to farmers, among others.

Leave a comment