When it comes to artificial intelligence reference is made to a varied world of technologies, useful for solving a variety of problems and united by the same principle: the development of algorithms able to imitate human thought and learn as new input is provided.
The best known example is certainly ChatGPTthe tool developed by OpenAI and launched in November 2022. However, the platforms available today are multiple and cover different sectors.
Read also Ten ideas for using ChatGPT on the farm
But what are the implications concrete examples of the use of artificial intelligence in agriculture? Exist Three large fields of application: the first concerns the provision of technical assistancethe second is related to the image recognition and finally there is thebig data analysis. These are three macro applications which are not really distinct and which should be integrated in an optimal software, but which are useful for categorizing the different approaches to the use of artificial intelligence in agriculture.
Read also Artificial intelligence calls, agriculture responds
Decision support, an agronomist always at hand
Farmers Business Network is an American company that has developed a platform to support farmers where information can be exchanged, agrochemicals and insurance policies purchased, and thanks to which it is possible to implement the principles of precision agriculture. It was launched a few months ago Norma chatbot based onChatGPT algorithm but trained on qualified sources relating to the agronomic world (proprietary data, scientific papers, company documentsUSDAthe United States Department of Agriculture, etc.).
This chatbot should have as objective that of support farmers in managing their fields. In principle he should be able to answer very technical questions, such as: which nitrogen fertilizer is best suited to fertilize durum wheat on calcareous soils with high temperatures? Or, how many diffusers should I put in the vineyard for sexual confusion against the vine moth? But even more generic: define a sustainable defense strategy for an organically managed apple orchard.
We tried this tool, as well as French Agri1 and German AgriGPT and the level of detail of the answers provided is absolutely not (for now) sufficient. However, it is about experiments in the beginning, which are based on limited datasets and are therefore able to provide generic answers, but not very useful for the farmer. But you can bet that, within a few years, they will have achieved a level of specificity such that they are commonly used. The fundamental thing, at this point, is to have good starting data.
Norm’s interface
(Photo source: Taken from the page dedicated to Norm)
Another interesting aspect of the LLMLarge Language Model, this is the acronym that identifies language-generating artificial intelligence tools, is to provide a simple interface to complex software. An example? Today, many farmers find it difficult to interact with Dss, Decision Support Systems, such as those developed for crop protection, as they are difficult to interpret. In the future the operator will be able to have a virtual assistant to simply ask “how is my vineyard today?”. The system will give a simple answer and possibly recommend which treatment to carry out in which plot. In short, it will be like having a agronomist always at hand.
But the aspect that perhaps intrigues farmers the most and makes them feel less replaced by a machine is the assistance in carrying out the paperwork. Between the Common Agricultural Policy (CAP), organic certifications and SQNPI, regional production regulations, specifications of large-scale retail trade (large-scale distribution) and more, farmers spend a lot of time behind bureaucracy. They hate and fear her. L’artificial intelligence in the future you will be able to automatically fill out the single application, but also suggest a cultivation plan or answer any doubts about the thousands of rules, exemptions and exceptions surrounding the CAP.
Artificial intelligence to support data analysis
The other great strength of artificial intelligence is that of makinganalysis of huge amounts of dataof various formats and in continuous evolution (the so-called big data) in order to identify meaningful correlations. This ability is already exploited today in the world of research, for example for genetic improvement or the development of new agrochemicals.
In fact, several large multinationals use artificial intelligence to create digital twins of plants and pathogens in order to make them interact with millions of virtual molecules with theobjective to identify those that could be interesting for defense purposes. In the past, years and millions of dollars were spent on laboratories test the molecules to identify those that are suitable for the development of a new product, today this entire phase is done on a virtual level, with a huge reduction in costs and a speeding up of processes.
But also the breeding is taking advantage of the possibilities opened up by artificial intelligence. Today there are automated systems for phenotyping of plants capable of recording and analyzing thousands of characters. But artificial intelligence is also used for decode the genome of plants and was fundamental for mapping that of common wheat, which has a genetic heritage five times that of humans, but largely composed of repetitions.
The farmer certainly has less awareness of these positives of artificial intelligence, but these are applications that are the basis of the tools that he will increasingly use in the field.
Image recognition: from insects to tractors
The human body is a perfect machine, the result of millions of years of evolution. Our eyes are able to perceive the light reflected by objects, send electrical impulses to our brain which transforms them into images. We are then able to distinguish objects and of interact with the world that surrounds us. All with one speed surprisingly high, not yet matched by technology. Artificial intelligence is learning to do something similar through so-called algorithms image recognition.
But what applications can the ability of a software to recognize the arrangement of objects in the real world? Infinite. An example concerns the smart traps. A camera is placed above the sticky panel of the trap which photographs the captured insects. The image is processed by a specific algorithm that counts the number of catches and is able to identify the species to which they belong. The collected data is then passed to analysis algorithms of data that identify population curves and defense strategies.
An example of smart traps
(Photo source: Taken from “Insect pest monitoring with camera-equipped traps: strengths and limitations”)
Another example is the autonomous driving. If at a technological level today a tractor is able to drive itself, its interaction with the outside world is less accurate. In fact, a tractor can follow a trajectory thanks to his GPS, but if he has to avoid an unexpected obstacle or advance on uncharacterized terrain he is in difficulty. In the future, artificial intelligence will enable a truly autonomous drivingable to identify obstacles or potential sources of danger and make decisions on how to manage them.
An interesting application of artificial intelligence in the service of image recognition concerns precision weeding. The startup Carbon Robotics has developed one laser weeder capable of distinguishing weeds in the field from the main crop and devitalizing them with a targeted laser beam. What a hundred years ago weeders did by hand, with their legs immersed in water, is now done by a laser beam and software.
The Carbon Robotics system can distinguish weeds alone
(Photo source: Carbon Robotics)
Is the era of farmers over?
Could anyone imagine a dystopian future, in which farmers will no longer be needed and machines will autonomously cultivate the fields and take care of cows and pigs. For now it’s just science fiction, but artificial intelligence and, in general, technology will be indispensable in the coming years to manage the epochal revolution that is going through the sector.
Within a few years in Europe the number of agricultural holdings will drastically reduce and the Used Agricultural Surface (UAA) will increase rapidly. There will be fewer companies, but larger and more specialized and they will have to produce more and more food while dealing with the effects of climate change, with increasingly complex regulations and a greater demand for sustainability.
These are challenges that perhaps man would not be able to manage alone. With technology probably yes. But this must be understood as one tool at the service of the farmerwhich maintains its centrality in the countryside, rather than being replaced by machines that imitate human thoughts.
Read also Artificial intelligence and its surroundings
