Artificial Intelligence in Agriculture: An overview

An early definition of what Artificial Intelligence comes from Computer Science: “Artificial Intelligence is concerned with the study of computers that do things that, at the moment, people do better”¹. On the other hand, a more comprehensive definition is given by ² who define Artificial Intelligence as: “The area concerned with building artificial artifacts that exhibit intelligent behavior.” The whole question goes through the premise of what is considered intelligent. Defining intelligence is not trivial. Depending on the approach, the meaning can be completely different. For example, in a philosophical context, it means human intellectuality; in a psychological context, it is linked to the ability to solve problems; in the rational approach, it means “thinking” correctly; and in an emotional context, it relates to dealing with feelings.

In general, the definition of artificial intelligence goes through the interrelation of four dimensions that are defined by ³: 1-think like humans; 2- act like humans; 3-think rationally; and 4- act rationally. In our agricultural context, I prefer to define intelligence according to dimensions 3 and 4, remembering that a rational system does “what is right” based on what “it knows” ³. In this choice, I limit myself to the rational component due to two intrinsic characteristics of all agroecosystems: a) they are extremely complex systems; b) they are systems where the knowledge representation must contain a very high dynamic component. That is, they are very challenging systems in Cartesian understanding, so if we include the human subjective component, the solution of models for an AI is complicated and perhaps unattainable in our current stage of scientific-technological development ⁴.

Thus, a proposed definition of Artificial Intelligence for Agriculture (AIA) is a sub-area of knowledge that intends to rationally emulate conditional behaviors in the face of unstructured and dynamic agricultural environments. Emulating conditional behaviors rationally means the extensive and intensive use of mathematical logic to model situations and environments where conditional situation of parameters is the key and definitive or constant information and data do not exist (or exist for a short time). Unstructured and dynamic environments refer to three parameters: the on-the-go agricultural environment constituted by an off-road substrate changeable in time and space, agricultural objects composed by living beings (plants and animals) of the fragile and fractal spectrum and by interfaces and interrelated influences between different agricultural systems and environments. All these components follow a conceptual structure modelled by mathematical chaos, which is greatly influenced by initial conditions.

In this conceptual context of an AIA ⁵ the artificial intelligence research area can be divided into seven major areas: 1-Knowledge Representation (KR); 2-Understanding natural language; 3-Learning; 4- Planning and problem solving; 5- Inference; 6-Search; and 7- Computer vision. Of these seven areas, the only one that is not (yet) widely applied in agriculture is natural language understanding, and all others are significant in the context of my definition of AIA.

Knowledge Representation (KR) deals with how the AI represents the world, represents itself, and its related tasks. This area focuses mainly on the types and forms of data structures, from a terrain map to a database of agricultural pests. Learning is the algorithms and system architectures that enable an AI to learn through various modes, such as observing an event, data mining, guessing alternatives right and wrong, and finally, having data structures annotated by a human. Planning and problem-solving are the core of agricultural robotics, and this AI deals with the ability to plan actions necessary to accomplish a goal or solve a problem. Inference in an AI makes it possible to obtain answers from incomplete information. It is the touchstone in systems that need to make decisions autonomously, such as diverting from an obstacle in a crop. Search in terms of AI means efficiently examining a KR and from problem-solving and inferences to find the correct answer or make the right decision. Furthermore, finally, the computational vision is about automatic signal sensing and image processing systems that allow AI to improve actions or increase data structures and KRs.

All these seven areas are not isolated when thinking about an IA; in many situations, mainly in agricultural situations, there is a recurring entanglement between each of these areas depending on the relevant objectives. An example is given by ⁶ which explains that a precision agriculture system can generate a vast amount of data that configure big data from sensors (data structure – KR), then the use of search techniques mine (Search) data that through AI techniques extract knowledge (ML – learning). This knowledge can be used to validate hypotheses (Inference) that guide a computer vision system based on digital cameras (Computer Vision) that obtain images. These images can be used according to ⁷ for a disease detection system through segmentation techniques coupled with Deep Learning (DL) techniques (Image Processing-Computer Vision) that will enable the planning (Planning) of a application phytosanitary by a drone-sprayer (problem-solving).           

Conclusion

AI is one of the fulcrums of Agriculture 4.0 and is at the forefront of the so-called digitization of agriculture ⁸. Together with robotics ⁹ and connectivity ¹⁰, they will raise the world agricultural standard to a scale never seen before. Transforming all farms, from small to large, to the new state-of-the-art farming paradigm: all will be Smart Farms ¹¹ As a humble and naive recommendation, I suggest that everyone who works in agriculture: start studying AI.

References

1. Rich, E. Artificial Intelligence-; McGraw-Hill: New York, 1983; ISBN 978-0070522633.
2. Cozman, F.G.; Plonski, G.A.; Neri, H. Artificial Intelligence: Advances and Trends; IEA-USP: São Paulo, 2021; ISBN 978-65-87773-13-1.
3. Russell, S.J.; Norvig, P.N. Artificial Intelligence : A Modern Approach ; Pearson series in artificial intelligence; Pearson: New York, 2020; ISBN 0134610997.
4. Albiero, D. Robots and AI: Illusions and Social Dilemmas. 2022, doi:10.1007/978-3-030-95790-2.
   CrossRef
5. Murphy, R.R. Introduction to AI Robotics-; Bradford Books: New York, 2019; ISBN 978-0262038485.
6. Singh, P.; Kaur, A. A Systematic Review of Artificial Intelligence in Agriculture. Deep Learning for Sustainable Agriculture 2022, 57–80, doi:10.1016/B978-0-323-85214-2.00011-2.
   CrossRef
7. Redhu, N.S.; Thakur, Z.; Yashveer, S.; Mor, P. Artificial Intelligence: A Way Forward for Agricultural Sciences. Bioinformatics in Agriculture 2022, 641–668, doi:10.1016/B978-0-323-89778-5.00007-6.
   CrossRef
8. Albiero, D.; Paulo, R.L. de; Félix Junior, J.C.; Santos, J. da S.G.; Melo, R.P. Agriculture 4.0: A Terminological Introduction. Revista Ciência Agronômica 2021, 51, 1–8, doi:10.5935/1806-6690.20200083.
   CrossRef
9. Albiero, D. Agricultural Robotics: A Promising Challenge. Current Agriculture Research Journal 2019, 7, 01–03, doi:10.12944/CARJ.7.1.01.
   CrossRef
10. Albiero, D.; Pontin Garcia, A.; Kiyoshi Umezu, C.; Leme de Paulo, R. Swarm Robots in Mechanized Agricultural Operations: A Review about Challenges for Research. Comput Electron Agric 2022, 193, 106608, doi:10.1016/J.COMPAG.2021.106608.
    CrossRef
11. Knierim, A.; Kernecker, M.; Erdle, K.; Kraus, T.; Borges, F.; Wurbs, A. Smart Farming Technology Innovations – Insights and Reflections from the German Smart-AKIS Hub. NJAS – Wageningen Journal of Life Sciences 2019, 90–91, 100314, doi:10.1016/J.NJAS.2019.100314.
    CrossRef