Adoption is facing a severe bottleneck. Generic data, slow deployment, and socio-economic conditions of farmers.
Dr. Ajay Kulkarni et al 24. Ph.D. in Computational Sciences and Informatics from George Mason University, Postdoctoral Associate in the Commonwealth Cyber Initiative. Yingjie Wang. Bradley Department of Electrical and Computer Engineering. Dr. Munisamy Gopinath. Distinguished Professor of Agricultural Marketing. Daniel Sobien. Research Associate in the Intelligent Systems Division of the Virginia Tech National Security Institute. Abdul Rahman. Deloitte and Touche LLP. Feras Batarseh. Commonwealth Cyber Initiative, Bradley Department of Electrical and Computer Engineering, Virginia Tech, Department of Biological Systems Engineering, Virginia Tech. “A Review of Cybersecurity Incidents in the Food and Agriculture Sector.” 03-12-24. arXiv preprint, arXiv:2403.08036. gautam
Data - Data plays a crucial role in developing AI solutions, and it is a significant bottleneck in getting labeled high-quality data to train AI models [213]. Recent surveys published by Ferrang et al. [214] and Alwis et al. [215] listed different datasets that can be used for different smart farming applications. However, they are not specially obtained from agriculture or smart farms. It is essential to represent the characteristics of agriculture in the data to develop a robust AI solution. This creates a need to obtain high-quality domain-specific datasets. One of the possible solutions to make diverse, high-quality data available is using agriculture testbeds. Recently, Batarseh et al. [216] and Oberascher et al. [217] reported soil testbeds that can help to produce agriculture-specific network and sensor data. This approach may help researchers to tackle data quality issues and provide robust AI solutions for agriculture.
Deployment – Deployment of AI solutions in real settings is another challenge observed in the agriculture domain. Most security solutions, such as malware, DoS/DDoS detection systems, and intrusion detection for UAVs, are not deployed. These security solutions are crucial for detecting and mitigating cybersecurity threats to farms. The deployment of AI solutions can also uncover the scalability and stability issues that may occur while using these solutions. Additionally, issues such as data and model drifts can occur with the deployment of models, which also needs attention to use AI as a decision-support tool.
Adoption - The adoption of AI in everyday decision-making for farmers is challenges because it depends on different factors such as personal, social, cultural, and economic [218, 219]. Prokopy et al. [220] noted that the education level of farmers, capital, income, farm size, access to information, and environmental awareness could also affect adopting new technology. In addition, the missing link between AI and AI assurance could increase the risk of using AI at the farm, affecting farmers’ trust, confidence, and safety. Thus, it is essential to incorporate AI assurance along with the deployment of AI applications and involve farmers in essential steps, as shown in FCAI framework. Further, providing informative sessions on explaining the need for AI, its use, and real-case scenarios could also help increase AI adoption in agriculture.