By Alex Kirpatrick, PhD (they/them), Communication Scientist, Center for Sustaining Agriculture and Natural Resources, Washington State University
Artificial intelligence (AI) is filtering your spam, gatekeeping your newsfeed, chatting with you online, and underpinning many of your regular activities. Many vaunt the potential of AI in agriculture to help land-managers adapt to uncertain and extreme weather, increase production through automation, mitigate greenhouse gas production through optimized precision farming and more. Realizing any of this potential depends on addressing potentially negative unseen effects and barriers to adoption, like a lack of transparency in AI systems and science and issues of data integrity. I work with the Center for Sustaining Agriculture and Natural Resources at WSU, and the Institute for Transforming Workforce and Decision Support (AgAID), a federally funded national AI institute (USDA-NIFA), to help address some of these barriers, and help agriculture find purpose for AI.
Emerging realization
I am a student of scientific revolution. Though I’ll admit to blushing a little over the past decade whenever I’d confess that my research focus was how people understand and communicate about AI (bleep bloop, robots, doodadsand the like…). I had witnessed AI in action and, like innumerable others before me, predicted a near-future underpinned by intelligent machines. Yet public salience of AI was still relatively low in 2017 when I arrived in Washington from England to churn out my PhD. There seemed little awareness in the social sciences of the potential impacts of AI on human communication. It was unclear to many, even myself, how this admittedly abstract computer science could impact media systems practically or wobble the staunch columns of journalism. Clarity, and Large Language Models (LLMs) like Chat GPT, would arrive given time.
Nowadays, I increasingly research the nexus of science, society and agriculture. I no longer blush as I talk on AI’s expanding role in society and the agricultural sector. Instead, I feel a familiar momentum building, unstoppable despite an industry’s ambivalence. But again, I am uncertain as to how exactly AI will be incorporated into everyday agricultural practice, what this means for our workforces or whether it can achieve its touted potential to help address the effects of climate change. I’m more certain that engaging with the reality of contemporary AI is essential to our collective understanding of agriculture 4.0 and our individual responses to it.
An adroit technology for a shifting harvest
Climate change affects crop yields, alters pest and disease patterns, and strains water resources. By analyzing and learning from vast datasets, AI models have proven potential to better predict weather patterns, model crop outcomes under various hypothetical conditions, optimize resource allocation, and might eventually minimize the overall environmental footprint of everyday farming practices. AI models have the potential to analyze data from interlinked networks of satellites, drones, and ground sensors, to help farmers understand their fields and orchards in finer detail than ever before. Such monitoring might inform the precise application of water, fertilizers, and pesticides, reducing waste and environmental impact. Type “Agriculture AI” into YouTube nowadays and you will see many examples of AI operating freely in farm environments when housed within robot platforms, exploiting laser detecting ranging systems (LIDAR) to navigate and complete complex tasks autonomously. AI and robotic automation has the added potential to reduce workload for skilled laborers who often work outside, and experience climate driven health and safety threats such as prolonged exposure to heat and poor air quality. Though the threat of worker replacement is at the same time implied.
Mitigation through machine intelligence
AI’s role in climate mitigation may also be transformative. For example, agricultural industry can contribute to reducing atmospheric CO2 levels through enhancing carbon sequestration practices. Machine learning algorithms could facilitate the identification and management of soil types most conducive to carbon storage, guiding the implementation of regenerative farming practices such as cover cropping, reduced tillage and agroforestry. These practices sequester carbon while also improving soil health, biodiversity, and resilience against climate extremes. AI-driven systems can also optimize the energy efficiency of farm operations, from irrigation to harvesting, by predicting the most effective times and methods for these activities. AI’s predictive capabilities may also support the strategic planning of crop rotations and the development of bioenergy crops, contributing to a reduction in greenhouse gas emissions by displacing fossil fuel use.
Embracing AI in Agriculture?
AI offers agriculture an enticing pathway to sustainability and resilience against climate change. Harnessing AI’s power may unlock new efficiencies, protect our resources, and secure our food systems against a warming world’s uncertainties. However, there are risks associated with the diffusion of any novel innovation in agriculture. Mass automation of tasks that humans are currently paid to perform naturally raises questions about mass unemployment, a realistic economic threat to the stability of society overall. AI is also greedy for data, raising privacy concerns and rendering agricultural operations of all sizes vulnerable to cyberattack and manipulation. Unnervingly, even the experts cannot always tell you how AI arrives at certain decisions. For example, generative AI is prone to hallucinating facts, and leaves no identifiable trail from input to decision. This lack of clarity and accountability, or the so-called ‘blackbox’ of AI, is a primary concern among publics, diminishing public trust in AI and intention to use. Even enhanced productivity itself can be seen as a risk. Historically, mechanization and intensive farming practices have harmed environments through habitat destruction and soil degradation. Even training an AI model can leave a huge carbon footprint akin to the carbon emissions of five gas-powered cars during their lifetimes in some cases.
Nevertheless, the potential rewards for our planet, food systems, and communities from AI adoption are vast considering the intensifying negative impacts of climate change on food systems. Ethical applications are dependent on enhancing transparency of AI science and algorithms themselves, promoting public understanding of AI (not dependency), eliminating human biases from code (e.g., structural racism, sexism etc.), safeguarding private data, and preferencing human oversight, control and the accountability of AI developers. A sustainable future might well rest on our ability to collectively manage AI risks and utilize thinking technologies to harvest and process the data from our fields and orchards. The land is talking and AI is listening. But will we listen to AI?
The work discussed in this post is funded by the AI Institute: Agricultural AI for Transforming Workforce and Decision Support (AgAID), supported by NSF and USDA-NIFA by the AI Research Institutes program, under award No. 2021-67021-35344.
Comments