The many faces of artificial intelligence

Artificial intelligence has many faces, and its diverse application does not exclude the agricultural sector. In order to make a strong case for an increased implementation of AI in agriculture, we present five examples of how AI could be or is already being a benefit to us humans.

In March 2020, when Germany plunged into the quagmire of the Corona crisis, I happened to come home one day from a strawberry farm. The usual seasonal workers could not cross the Romanian border, strawberries were choked with weeds, and the farm’s owner complained about new challenges. I, completely exhausted after 8 hours of weeding, was listening to her drowsily and thought to myself: Why is it still cheaper in such a prosperous country like Germany to hire broke students and Romanian workers for this physically demanding job rather than to buy a robot that will do it much more efficiently and without any border issues?

A year has passed, strawberries were harvested and eaten, things are carrying on as usual, and only the Hohenheim student is still trying to get her mind around agricultural robots. Why aren’t we seeing them here and there, in each and every field? There are tons and tons of articles and scientific papers describing direct advantages of artificial intelligence (AI) in agriculture and probably for most people, it is already clear that agricultural robots (or agribots) will become more popular in the future. For greater certainty, we decided to think about other, less obvious but still important possible features of agribots.

1. Farmers are exposed to various health threats: They stay in the sunshine all day long, often in poor sanitary conditions and inhale pesticides and grain dust which might lead to the development of the Organic Dust Toxic Syndrome or different parasitic diseases. Add oftentimes limited access to healthcare thereto and you end up with farmers’ overall vulnerability to health disorders. Implementation of automated systems applying herbicides/pesticides very precisely and only locally will mitigate exposure of farm workers to hazardous substances. Agricultural jobs won’t be gone completely since people still need to operate the AI machines, while new jobs will be generated for people developing AI systems.

2. While having originally been designed for practical farming purposes, such as defining fruit ripeness, agribots can be very useful to researchers as well. These machines are able to precisely measure canopy height, leaf area index, stem thickness, and other parameters which are used as indicators of various physiological traits. Depending on the research question, a scientist may have to measure thousands of leaves, shoots, or canopies. This is a time-consuming, tedious, and exhausting work requiring a lot of manual input from technicians and assistants. How much time can be saved by using agricultural robots and spent for other tasks that cannot be done by AI? Through the participation in scientific research, agribots may eventually benefit the development of better cultivars and promote biodiversity in agricultural ecosystems.

3. While initial agribot projects focused mainly on large-scale, industrial agricultural production, further development has led to the invention of smaller and more agile niche machines. Besides the straightforward benefit of enabling farmers to explore new profitable market niches (such as gentle harvesting of strawberries or management of compact vertical farms), they also allow for sophisticated agronomic approaches which eventually benefit biodiversity. For example, intercropping, a.k.a alternating crop stripes, is often criticized for its complexity (in seed beds preparation, varieties choice, adequate machinery etc.) and labour intensity. The same applies to agroforestry - growing fruit trees and agricultural crops on the same land. A farmer just cannot handle these farming systems with traditional big machines but might also not be committed to do it manually. Clever and compact robots can solve this problem. As a result of such intelligent machines allowing for eco-friendly farming management, less weeding is needed, and thereby less herbicides required. Thus, it benefits the farmer, the consumer, and not least, invertebrates, leading to higher biodiversity in agricultural landscapes.

4. Another not so obvious but very welcoming impact of AI’s on agro-ecosystems is its potential for the protection of segetal vegetation, which comprise all non-crop plants that grow in agricultural landscapes and have evolved to coexist with crops. Due to intensive agricultural practices, the diversity and abundance of these plants are decreasing significantly, sometimes even to the point of extinction. Therefore, many segetal plants have been put on the red list. Organic farming practices and environmental programs are able to protect the biodiversity of segetal plants, but this often comes at the expense of yield in comparison to conventional farming. Also, many farmers simply do not want to switch from conventional farming methods to organic farming. Thus, a feasible solution to this problem might be the replacement of herbicides with intelligent robots that can differentiate between non-threatened weeds and red-listed weeds on conventional farms. That way, conventional farmers would largely be able to maintain a weed-free farming practice while sustaining those weeds which need to be protected from extinction.

5. Finally, who doesn’t hate throwing away food? Globally, around one third of the food produced for human consumption is wasted each year, for many different reasons. This is not only disastrous for ethical reasons, but it is also damaging the environment since all the energy and resources used for food production go to waste with the food itself. The loss and waste of food along the supply chain is therefore the world’s third largest emitter after China and the United States. Among the reasons for food waste is the situation that “[shoppers] want lots of options, and retailers don’t want to run out of anything,” leading to a surplus of food that goes bad on the shelves. In that sense, when, where and how much of a certain crop is demanded by the market plays a role in what and how much of it is grown at a particular time. And since crops usually take several months to develop, it is all the more important to be able to predict the demand of a certain plant. Artificial intelligence is increasingly being used for just that purpose. The behavior of customers can be analyzed and predicted, and changing patterns can be detected by machine-learning. Thus, the global food market can match its supply much more precisely with the demand of the population, thereby decreasing food waste.

The application of AI in agriculture seems almost limitless since each issue requires a very specific solution - there is no such thing as the AI technology for agriculture. Thus, we’re looking forward to seeing more and more agricultural AI applications emerge in the future.

Authors: Thomas Köhler, Maria Kunle

Comments:

dnpatrice2001

Robots are mostly built of critical raw materials (CRMs). For many CRMs, the upstream steps of the value chain are not présent in the EU: beryllium, antinomy, borates, magnesium, niobium, PGMs, phosphorus, rare earths, scandium, tantalum, vanadium and natural rubber. This is due to either the absence of those materials in the European ground or to economic and societal factors that negatively affect the exploration or the extraction.

To enhance robotics application in agriculture, it is important to secure access and diversify the supply of those materials. Going forward, actions are needed regarding the robotics supply chains such as:

- Incentivizing the adoption of sustainability software by extractive industry

- Diversify the material supply

- Support the eco-design of robotics products through recycling and reuse, substitution

- Promote R&D, develop skills and competences

- Unlocking technology transfer and investment opportunity in third countries in order to promote secondary sources of CRMs.

dnpatrice2001

In economics, demand is the quantity of a good that consumers are willing and able to purchase at various prices during a given period of time. The demand functions are derived mathematically through utility maximization (Lagrange method) and theoritical properties of demand systems. However, the demand function lacks robustness because human actors often make sequences of choices that are incompatible with any consistent utility function. The standard demand function fails to consider the underlying factors of purchasing decisions and therefore does not represent consumers' behavior in a real economy. The purchase decision-making process and the interactions among consumers based on this process generate market dynamics which are not handled by the standard demand functions. Those shortcomings prevent the global food demand to match more precisely with the supply and therefore increase food waste and the related environmental and social impacts. The transition from standard towards real demand functions requires modelling the purchase decision-making of each consumer. This issue involves research in the fields of psychologue, economics, sociologue and marketing, which is in line with the research of the agent-based computational simulation of complexe social systems. The combined use of the Agent-Based Models (ABM) and Multi-Agent Simulation (MAS) method in studying consumer behavior and markets gives one the potential to cope with the dynamic changes and complexities in the real-world business environment. By creating a large number of heterogeneous consumer agents in an artificial market, multi-agent simulation (MAS) can be used to exhibit a market dynamic phenomenon originating from the individual behavior of heterogeneous consumers and their interactions in the real-world complexe market.

The strategy is to create a virtual market by using a large number of artificial consumer agents to simulate the emerging market dynamics. The algorithm controls these agents' buying behaviors. The validation of the effectiveness of the agent-based purchase provides the potential to explore and predict dynamic market variations, e.g. fluctuations in market shares, the influence of economic gouvernement policies, technomogical innovations, and market's responses to expected or unexpected events.