Both data analytics and artificial intelligence will play a bigger role in farming in the years ahead, but with the opportunities and benefits come great challenges. The key is how farmers harness all the data where it actually benefits the bottom line.
In the inaugural North Carolina Plant Sciences Initiative Symposium Dec. 6 at the Hunt Library on the North Carolina State University Centennial Campus in Raleigh, a panel discussed the best way to harness the data revolution to make agriculture more resilient.
The panel was moderated by Cranos Williams, platform director at the North Carolina Plant Sciences Initiative. The panel included Ranveer Chandra, managing director, research for industry, networking research, CTO, Agri-Food for Microsoft; Bill Danker, ag-tech principal with SAS; Tom Eickhoff, chief science officer with the Climate Corporation; and Isaac Boerema, owner of Boerema Farms in Hyde County, N.C.
In the discussion, Microsoft’s Chandra noted that many of the decisions made by farmers are based on intuition or guess work. He explained that Microsoft launched the FarmBeats project in 2014 to augment a farmer’s knowledge with data and artificial intelligence, or AI. He said the goal is to leverage the knowledge every farmer has with data-driven decisions.
“Some of the biggest bottlenecks of AI to agriculture are very unique to this field. The No. 1 challenge is how to get data from the farm. Most of these farms don’t have good internet access. The No. 2 challenge is connectivity; the No. 3 challenge is affordability. How do you make it so farmers can afford the AI you are building?”
Eickhoff, with the Climate Corporation, emphasized that in data science there is a tradeoff between precision and scale. The challenge is how precise can you be with the data and how wide can your decision making be?
“At the root of that is data privacy. For farmers, data can be an asset, and we want to protect that asset. We need to have a partnership that allows access and builds a trusting relationship with those individuals who are supplying data,” Eickhoff said.
Eickhoff pointed out that Bayer’s Fieldview Platform has been well adopted globally with more than 186 million acres of data coming into the platform. Eickhoff emphasized the agreements are very specific on who owns the data: The farmer owns the data.
“At any point in time if they want to disconnect from the platform and remove their data, it’s a simple as going to the website, making the request and the data is removed. Building that level of transparency and trust is absolutely critical,” Eickhoff said.
He said farmers can share the data with their vendors, which will help drive change in their operations based upon analytics that vendors can provide. The key is the vendors must bring value to the farmer to make it worthwhile.
“No one wants to hand over their assets and not get any return. First of all, you can trust us with the data; second of all, I am going to give you something that is going to make you better — sustainability, profitability,—whatever that may be,” Eickhoff said.
Boerema said the primary challenge for farmers in data management is ensuring they get good quality data. “We’re not making one stand-alone data collection pass. We’re collecting that data on the back of something else we are doing, whether it be planting, spraying, or harvesting. That’s the primary purpose.”
He said the challenge with AI and machine learning is how to move the data from the field to the computer or the cloud or the office or wherever it is needed, and then turning back around in real time to make an actual, actionable decision. Another major challenge is labor.
“We have really good people working for us, but they are trained to be operators or agronomists or something else, they’re not trained to be data scientists. If we are asking them to do cross disciplinary work, they need cross disciplinary training,” Boerema said.
“We look to industry, the university for solutions. It has to be easy on the front side to use and to capture that information. It’s got to be good information. On the backside, it has to be powerful enough to actually give us a good actionable item. If it’s just watered -down, weak information, we have enough intuition to make those calls already,” he said.
About the Author(s)
You May Also Like