Challenges of an AI-led agriculture

2022 UPDATE: I have to insist that anyone involved in such tech innovations have a think about why they are doing what they are doing. Just think & reflect:

1) Is it truly life-saving? Is it essential? Is it vital? For whom?

2) What are the true consequences of improving agricultural tech-innovation?

3) Are you aware of the historical record (see @wikipedia articles, I can recommend the relevant military history articles/key concepts)

4) Who handles the #data collected? Can they be trusted?

4) If unsure about any of the above, cease fire. Now.

Did you know that "cyber farming" makes basil taste better? According to MIT researchers, taste and other features can be improved by exposing plants to 24-hour light. They made their observation using networked tools to

“take a plant’s experience, its phenotype, the set of stresses it encounters, and its genetics, and digitize that to allow us to understand the plant-environment interaction” ( source )

This is simply one feature of a smart, AI-based agricultural solution. Artifical intelligence can also be utilised, for example, to find the best LED light recipes that maximise yield, plant development (vertical or horizontal, above- and underground) and reduce energy consumption to the minimum.

More broadly, a wide spectrum of smart agriculture is used as a means to conserve vital resources, most notably water through smart farmland irrigation, nutrients by reducing fertilizer use, available arable land and soil by hydroponics and aquaponics. Diseases can be predicted and yield maximised by AI through remote sensing, improving food security by looking at data from smallholder farms. Not to mention discovering and reducing the electricity cost of some cash crops in emerging markets.

IN particular, AI can be used to establish the optimal conditions for irrigation, as well as tasked with making the decision to irrigate. Another novel use for farmland AI, has been inspired by the success of image processing algorithms in spotting human disease. What if we could transplant the findings (deep learning) from finding human diseases to finding pests on farms, differentiating between plant developmental phases, as well as monitoring the health of crops on a large scale? Well, all of these are now possible.

While on the surface, the offer seems sweet, we also have to address the elephant in the room. If we are to facilitate the widespread use of deep learning (AI) as the modus operandi for precision agriculture, we must consider the environmental footprint of training deep learning algorithms.

"costly to train and develop, both financially, due to the cost of hardware and electricity or cloud compute time, and environmentally, due to the carbon footprint required to fuel modern tensor processing hardware." ( source)

In conclusion, any exponential technology has to be evaluated by looking at the complete cycle(s) it is part of. One barrier to AI-adoption in agriculture is the huge upfront cost, but in most cases, a strong business case can be made. However, if we are to advance as humans, we must also consider the runaway effects of our efforts. Training AI to save water - cost - energy is a noble venture, but its impact should be considered, anticipated, and dealt with.

What do you think? Let me know in the comments.