The agricultural sector stands at a crossroads of technological advancement with the emergence of Artificial General Intelligence (AGI) specifically targeting crop data. This comes on the heels of the rapid introduction of AI platforms like ChatGPT, which amazed the world with its impressive intelligence. However, that intelligence was partly built on stolen data from the internet, including content from news outlets. Recently, it also came to light that companies like Apple and Nvidia used datasets from over 173,000 YouTube videos, including subtitles, without permission to train their AI models. This begs the question: Is the agricultural sector facing a similar scenario?
In a few years, it may turn out that major AI platforms quietly walked away with farmers’ business data, only to then charge them hefty fees to access what was once their own information. Let me be clear: I am an optimist and a proponent of technological progress. Generative AI, or Gen AI, has the potential to support farmers as a virtual assistant that knows everything about crops, soil, weather, and machinery. It continuously learns from vast amounts of data, as if the farm has an extra pair of eyes and brains that are always ready to help.
There are many examples of this. In May 2023, Norm, an AI from the rapidly growing American platform Farmers Business Network, was one of the first. Recently, platforms like Taranis, AgriSearch Assistant, and Sage from Cropin have joined the ranks. Sage, for instance, collects detailed crop data on a global scale down to a 3×3 meter grid, allowing for accurate yield predictions. However, instead of creating a level playing field, these systems threaten to further skew the balance of power in agriculture.
Krishna Kumar, CEO of Cropin, stated: “The information provided by Sage will primarily benefit large agribusinesses that can afford the expensive subscriptions.” For the individual farmer, this means nothing less than a growing information gap between them and their suppliers and buyers. While large companies reap the benefits of the precision and predictions offered by these AI systems, farmers are left empty-handed, forced to pay for information they helped generate.
This development is an inversion of the natural order; large companies use hyper-powerful AI to swipe farmers’ data right from under their noses and get richer in the process. This is strongly reminiscent of how AI companies have used public data from the internet over the past few years to train their models. Now, the same thing is happening on a large scale in agriculture, with potentially disastrous consequences.
One often overlooked aspect is the abundance of public data available. Advanced satellites capture images of every square meter of agricultural land, which can expose everything like X-ray images, from crop growth to the timing of farming activities. As a result, the agricultural sector has become one of the most transparent and exposed industries. However, for AI platforms, gaining access to specific business and crop data is even more valuable. This data is being shared on a massive scale worldwide through platforms like My John Deere, offered by major agricultural machinery manufacturers.
But who ensures that this data remains secure? Who says that smart AI algorithms won’t find backdoor access to this valuable business data to further refine their models? This is not a far-fetched theory; it’s a real risk that farmers worldwide face. If farmers do not take action themselves, we might witness the largest data heist in agricultural history. In broad daylight, no less. Farmers would then not only lose control over their data but also their knowledge advantage, becoming dependent on the advice of AI platforms—advice based on their own data, but for which they have to pay dearly.
With data from a single farmer, you can’t do much, but if you combine it with data from all farmers, it becomes invaluable. I am certainly in favor of data sharing, but it’s crucial to consider with whom this data is shared. And more importantly, who processes and analyzes it, and for what purposes.
It is crucial that farmers become aware of this threat and organize themselves—not only to protect their data but also to process and use it for benchmarking purposes, such as evaluating product performance and identifying which farming practices yield the highest returns. Over time, this can lead to the creation of a large virtual test farm, also known as a digital twin farm. Farmers could manage their own data cooperatives, where they share insights from their collective crop data and only grant third parties access for a fee. This is the only way for farmers to retain their autonomy and knowledge, and even gain an edge over their suppliers, buyers, and governments. It’s essential to prevent being swallowed up by the technological giants of the world.