You are born. Life gets more and more complicated. Then you die.
Okay. Maybe it’s not as simple as that. But simple is pretty hard to come by these days and more and more complicated covers a lot of our mutual experience. New ideas, methods, technology and players are springing up in every field. One of those fields is my own back yard. I have been writing about regenerative agriculture for 5 years now. It seemed pretty simple in 2019. Don’t plow. Plant cover crops. Be organic. Rotate grazing and you will be alright.
Ever notice how the more you know about something, the harder it becomes to understand, let alone explain? That happens to me a lot. When I was in an office showing people how to do basic things on computers, I would gladly demonstrate how with a few easy actions you could get a report to print or something like that. It didn’t work out so well. People mostly learned “Oh, Joe knows how to do that. I’ll get him to do it.” I ended up with more things to do by trying to show people how to do things themselves.
Farming regeneratively can get complicated: Arbuscular mycorrhizal fungi attach to roots and form a symbiotic relationship that vastly increases the water and minerals available to the plant while using some of the plant’s excess gluten to extend their hyphal network to harvest nutritional exudates from bacteria and other microorganisms. So, don’t break the fungal network or let your soil become bacteria rich and fungi poor. See. Things can get a lot more complicated.
The solutions offered by regenerative agriculture depend on biology: microbes, bacteria, viruses, protozoa, fungi, nematodes, arthropods and earthworms. Sure. Chemistry is complicated but let’s not pretend that biology is simple. One of my guiding lights learning about regenerative agriculture has been John Kempf, founder of Advancing Eco Agriculture (AEA) in 2006. AEA wants to make regenerative farming the norm by 2040. That is not something one person or one company can do alone. AEA currently provides crop plans on more than four million acres globally. The real genius of the operation is to bring in agronomic experts from around the world and spread information through books, seminars and podcasts plus a large staff of in-house consultants.
I only farm one acre of grapes. Hiring consultants, doing soil and sap tests, doing extensive field work with employees and machinery is usually out of my league and beyond my budget. Making detailed crop plans is basically over my head. But, out of the blue, AEA has introduced Fieldlark AI. They took over two decades of tests, techniques and results from all of their consulting work and fed it into an artificial intelligence engine. Right now, it is free to use. There’s a fee after a certain number of questions. I had to check it out.
I asked it about a pest problem in one variety of grapes. The very immediate response asked a few questions about my soil etc. Then came a long response broken into 6 parts: 1) Short Term Treatments 2) Biological Control 3) Long -term prevention 4) Optimizing water management and carbon 5) Promoting Synergistic Biodiversity and 6) More questions to explore with ready answers. You will notice that these arranged themselves from immediate local actions to larger responses spread over time and area. Another query into Fieldlark about a weed in my field brought back another 6-part response arranged in the same order. Within each of these 6 parts there are 3 sub-categories: a) recommendation b) biological basis and c) Expected outcomes. Those outcomes cover plant performance, soil outcomes and yield improvements.
Granted, this is a very mechanical arrangement, but it is also helpful in terms of the most urgent actions to take first and what specific ingredients are needed to perform them. Embedded in the expected outcomes part is a feedback loop that presumably would fine-tune the response if the outcomes don’t come out. There are also recommendations to check with an agronomist to verify recommendations for your context. I’m not sure if that absolves the AI of responsibility but it might.
Discussing this new twist in agriculture with a farmer friend I found out that he has also used an AI to research farm problems. He used GROK which turns out to be a product from Elon Musk. It also turns out to cost money but not a huge amount. I bit the bullet and bought a month’s worth of advice for $20. GROK’s advice was much shorter and very product oriented. The first recommendation was to use yellow sticky tape – apparently lots of it if there are many rows with the issue. Next was to use insecticides with very poisonous ones listed before more organic treatments. Actually, I had already tried these particular organic products with limited success. Then it suggested biological controls, monitoring the sticky traps to see if they work and getting local advice perhaps again for legal reasons.
Since GROK is not based on a specifically curated set of regenerative agriculture data, it is helpful to see that each recommendation has a specific reference to see where it came from. As it turns out, the source of the recommendations was usually also the company making the product recommended. I don’t know if GROK gets a kickback from each company for each recommendation, but that seems likely. To be fair, Fieldlark also recommends specific products from AEA.
What really sets these two approaches apart from each other is the general assumption that the best way to control pests and disease in the source material used by GROK is with products and chemicals and the basis used in AEA is that the best approach is by improving the health of the plants. In Quality Agriculture by John Kempf, agronomist Tom Dykstra writes “When you have a healthy plant, you don’t have to use, for example, all of the pesticides that are being used today: herbicides, insecticides, fungicides, nematicides etc.” This point is reiterated in a field guide backed by the Australian government that states “Nutrient-dense plants are more resistant to pests and diseases.” Moreover, it goes on to note that pesticides and nitrogen-fertilizers may worsen the problem by making plants more attractive to disease and insects. In the long run, I hope that these AI’s can learn from interacting with their users. In my example of teaching computer use I became less effective and more burnt out when lessons failed to take hold. A decent AI should learn as it goes and become even more effective based on evidence not just a large language model.