Researchers Highlight Dangers of AI in Agriculture
Though AI is commonly mentioned in the context of business efficiency and profitability, a quiet revolution is underway in a critical sector that affects all of us: farming.
According to a new report published earlier in the week, global agriculture is poised to benefit from the deployment of AI in agriculture.
AI for crop management
For instance, AI in agriculture can improve crop management and agricultural productivity through rapid diagnosis of plant disease and efficient application of agrochemicals, among others.
In the report titled “Responsible artificial intelligence in agriculture requires a systemic understanding of risks and externalities”, researchers called out the need to review risks around unintended social-ecological consequences of machine learning models, and potential safety and security concerns.
According to the researchers, concerns around the relatively new field of AI are typically focused on bias, inequality, and privacy. When it comes to global agriculture, however, a different set of considerations emerge.
One revolves around the risks associated with deployment at scale, say the researchers. They argue that as AI becomes indispensable in agriculture and culminates in an increasing reliance on common ML platforms such as TensorFlow and PyTorch, this creates a potential point of failure that an attacker can exploit.
“[Farmers] will bring substantial croplands, pastures, and hayfields under the influence of a few common ML platforms, consequently creating centralized points of failure, where deliberate attacks could cause disproportionate harm.”
In addition, non-malicious mistakes might also be vastly amplified, taking place at machine speed faster than humans can respond – with disastrous consequences.
“If monocultures—where a single genotype of a plant species is cultivated on extensive lands—are irrigated, fertilized, and inspected by the same suites of algorithms, a model error or poorly calibrated sensors may lead to excessive fertilization and soil microbiome degradation, at the risk of large-scale crop yield failures.”
The researchers conceded that the widespread deployment of AI in agriculture is both valuable and expected. To avoid the pitfalls, though, necessitates careful navigation of the road ahead by implementing comprehensive risk assessments and anticipatory governance protocols.
Finally, they concluded with a warning that the risk landscape discussed in their paper is similarly applicable to agricultural systems that provide non-food products, such as the production of fiber, fuels, pulp, oils, rubber, and plastic.
Image credit: iStockphoto/valio84sl
