Artificial intelligence and possibilities of its application in modern agriculture

P. Lykhovyd,R. Vozhehova,L. Hranovska,V. Ushkarenko,,
DOI: https://doi.org/10.31473/2305-5987-2023-2-33(47)-6
2023-01-01
Technical and technological aspects of development and testing of new machinery and technologies for agriculture of Ukraine
Abstract:The article is devoted to the study of the possibilities of using artificial intelligence for agricultural production from the point of view of scientific theory and solving practical problems. The purpose of the study was to investigate the possibility of using a modern artificial intelligence system «Agri1», developed specifically for the needs of agriculture, to solve theoretical and practical problems related to agro-industrial production, namely, in three such fields as soil science, agrotechnology, and land reclamation. Methods. Questions of a scientific-theoretical and practical nature from the specified branches of agriculture were submitted to the «Agri1» artificial intelligence system according to the developed methodological plan, which provided 10 questions for each group of tasks, namely: 10 scientific-theoretical questions and 10 practical questions on soil science, agrotechnology, and land reclamation. Thus, a total of 60 questions were put into the system. The quality of the answers was evaluated by the percentage of correct and complete answers, considering the need to ask additional clarifying questions. Results. As a result of the investigation of the capacity of artificial intelligence system «Agri1» it was established that the percentage of correct answers averaged 81.67, 83.33, and 88.33% for the questions from soil science, cultivation technologies, and agricultural melioration, respectively. Higher accuracy and quality of answers were obtained for practical questions, while the «Agri1» system gave less correct and qualitative answers to theoretical requests (86.67% vs. 82.22% for practical and theoretical requests respectively). Besides, a significantly larger number of additional requests were made to the «Agri1» system for solving the tasks, which are atypical for international agronomic science and practice. The gross output of correct answers was 84.44%, which is sufficient for the recommendation of «Agri1» application in scientific and practical purposes. Setting an unlimited number of additional requests and internal self-learning of the system allows to suggest that under certain circumstances it is possible to achieve 95% or even higher accuracy and correctness of the answers. Conclusions. The «Agri1» artificial intelligence system is a reliable and high-quality tool to obtain scientific and theoretical information and practical advice on soil science, agrotechnology and land reclamation, which can be used as an additional source of information for the research and production purposes. The disadvantages of the system are the English interface, a limited number of questions in the free version, and low information about the non-typical for the EU countries and the USA technological issues. Key words: agrotechnology, soil science, information technology, land reclamation, artificial intelligence.
What problem does this paper attempt to address?