The (Green) Robot Revolution

John Deere Leaps Into Silicon Valley to Feed the World.

Hello, World

“Fire gave us power. Farming made us hungry. Science made us deadly.”

– Yuval Noah Harari, Author of Sapiens and Homo Deus

     The often esoteric, increasingly prevalent topic of machine learning has dominated headlines for companies like Facebook, Netflix and Tesla. But seldom do the implications of artificial intelligence overlap with old-fashioned industries like agriculture, dominated by manufacturing stalwarts like John Deere, best known for its iconic green tractors. And yet with the population of our planet expected to surpass 9 billion people by 2050, generating a 70% required increase in caloric output[i], perhaps there’s been no better time to step back and do just that. 

     Why does this technology represent the most drastic step function shift in the ten thousand years since the establishment of the agricultural economy? In addition to the increased number of mouths to feed, “climate change disrupts growing seasons, turns arable land into deserts and floods once-fertile deltas with seawater”.[ii] The resulting increase in variability of conditions manifests in even lower yields and profits in an already notoriously low-margin industry. There are “2 million farms in the continental U.S. and the decisions their proprietors make impact the entire food supply chain”.[iii] As of today “almost 35% of the food produced is wasted between production and consumption”.[iv] In the near-term, the farming industry is looking towards artificial intelligence to “optimize their land and their operation by being able to make decisions in real time based on in-field conditions”.[v]

 

Agricultural Intelligence

There’s not a whole lot of additional land to bring into production, so we need to do more with less. Seventy percent of it is going to come from technology.

– Alex Purdy, Director of John Deere Labs

     After 180 years of providing traditional farming equipment, last year John Deere took the leap into Silicon Valley with its $300m acquisition of Blue River Technology, a startup with “computer vision and machine learning technology that can identify weeds, making it possible to spray herbicides only where they’re needed, reducing chemical use by about 95%”.[vi] The result is improved yield, decreased labor cost and healthier produce. The key technology is called see and spray[vii], which uses a system of cameras atop spraying equipment, in conjunction with intelligent software that uses deep learning to identify different types of plants: if it recognizes a weed, the robot sprays it with pesticide and if it recognizes a crop, it nourishes it with fertilizer. Furthermore, all parameters can be toggled by farmers enabling a degree of flexibility for various scenarios.  

     Ultimately this is just the first step in what is likely a multi-decade evolution for Illinois-based John Deere, as the development of more advanced but manually-operated machinery will generate diminishing returns. The product development cycle will rebalance, shifting towards computer science in conjunction with mechanical engineering. Deere plans to leave the 60-person Blue River team in San Francisco, to continue its rapid expansion and grow its innovation pipeline. But to provide a fully integrated solutions, the two teams cannot operate in isolation. Given access to talent, culture differences and market optics, the siloed approach right now is logical. Eventually, however, management must decide if it wants remain an industrial manufacturer or grow into a technology-enabled robotics company. 

 

Green Is Good 

Farming is on the cusp of a major change. The industry will be transformed by data science and artificial intelligence.

– Gayle Sheppard, Vice President of Intel AI

     The combination of robust training sets, sharpened algorithms and connected hardware will result in an increasingly smart farm, trading labor for data. And this will manifest throughout the entire supply chain. For instance, Blue River is also developing a robot for “precision lettuce thinning and a drone imaging system that collects data from fields”.[viii] Over the next decade, artificial intelligence startups, equipment manufacturers and agricultural producers will have to collaborate to locate inefficiencies in the farming process and use technology to boost productivity. There will likely be a wave of consolidation and vertical integration across the industries, leading to more streamlined operations. 

     So what does farming look like in 2050? Broadly speaking, digital imaging and machine learning can be utilized “in every other step of farming: tilling soil, planting seeds in the optimal locations, spraying fertilizer or nutrients and harvesting”.[ix] The concept of fully-automated agriculture is not out of the question. The benefits are clear: higher yields, lower costs and no empty stomachs. But what does this imply for the millions of people currently employed by farming? What happens to 1.5 billion hectares currently used for crop production[x] as vertical greenhouses pop up outside major cities? Perhaps most importantly, is farming really the future… or will another revolution shift the way we feed ourselves?

 


(Word Count: 780) 

[i] Office of the Director, Agricultural Development Economics Division, “How to Feed the World in 2050”, Food and Agriculture Organization of the United Nations (October 2009), 2.

[ii]The Future of AI in Agriculture”, Intel AI

[iii] Kyle Wiggers, “IBM’s Watson Agriculture Platform Predicts Crop Prices, Combats Pests, and More”, Venture Beat (September 2018).

[iv] “IIoT in Agriculture”, Frost & Sullivan (December 2017), 15.

[v] Errol Barnett, “Farmers Look to Artificial Intelligence as Workforce Declines”, CBS News, (April 2018).

[vi] Adele Peters, “How John Deere’s New AI Lab Is Designing Farm Equipment For A More Sustainable Future”, Fast Company (September 2017).

[vii] James Vincent, “John Deere is Buying an AI Startup to Help Teach its Tractors How to Farm”, The Verge (September 2017).

[viii] Michael Lev-Ram, “John Deere Is Paying $305 Million for This Silicon Valley Company”, Fortune (September 2017).

[ix] Adele Peters, “How John Deere’s New AI Lab Is Designing Farm Equipment For A More Sustainable Future”, Fast Company (September 2017).

[x] Rajender Thusu, “Sensors in Agriculture – Smart to Intelligent”, Frost & Sullivan (April 2018).

Previous:

Data – The Farmer’s New Edge

Next:

Can ML replace Human Resources?

Student comments on The (Green) Robot Revolution

  1. I agree with you that farming will drastically change during our lifetime. While I was on active duty, we used Persistent Systems Wave Relay radios which are also used by farmers experimenting with autonomous and remote-controlled farming equipment. I am unsure of when vertical greenhouses will become economically viable. In general, the cost of land and building costs have grown a lot over the last real estate cycle making even traditional multifamily and office development challenging. Secondly the rise of e-commerce has made industrial buildings, which had traditionally been the cheapest property type both by land value and building cost, much more sought after for last mile distribution warehouses. This being said I do see small scale vertical greenhouses growing in popularity as the farm to table and locally sourced movement grow in popularity.

  2. I have been interested in how sleepy industries, like agriculture, are adopting machine learning for quite a while now. While it is clear that machine learning can bring great productivity to such industries, I think you hit upon one of the bigger impediments when you explained, “Over the next decade, artificial intelligence startups, equipment manufacturers and agricultural producers will have to collaborate to locate inefficiencies in the farming process…” I am curious to know how agricultural producers capture data and, if they do capture such data, how granular and reliable is it. At the end of the day, the application of expensive technology like ML and robotics comes down to an ROI question, and you can only answer this question by analyzing how expensive of a pain-point is the problem compared to the cost of the solution. I have no doubt AI startups are excited to explore this question, and you show that John Deere is also excited to collaborate; but will agricultural producers be incentivized and excited to put in the time and effort to find such inefficiencies?

Leave a comment