Artificial Intelligence in Farming

The main use of Artificial Intelligence in farming, as it is in all uses, is data collection and analysis. It is in that analysis where AI tools begin to shine.

How Can AI Help in Farming?

Reports of increased yields, reduced chemical inputs and reduced time are inviting, especially as costs continue to rise and margins are ever squeezed. As well as reduced chemical inputs, reductions can often be made in water usage. This is becoming increasingly important as we’ve seen from the recent summer weather. Coupling AI tools with other methods of water application, such as drip irrigation, can lead to drastically reduced water usage. This improves resilience for the farm but also the wider environment.

Risks of Using AI in Farming

From a risk perspective, however, care needs to be taken when using various AI tools. The systems are not infallible, and they do sometimes get it wrong (AI failure). Hence, human supervision is critical. Although currently limited, court cases have shown that there is an expectation of human supervision. If this is not given it is generally viewed as negligence on the part of the user, i.e. the farmer.

It is also likely that the AI software provider will have robust indemnity clauses and limitations of liability, and therefore any error that results in a loss of crop, for example, will likely be borne by the user. It is a developing service, but insurance could mitigate losses from AI failure, although policies would have to be thoroughly reviewed to ensure sufficient coverage.

Aside from AI failure, another risk is any unintended consequences. What happens if an AI tool overuses pesticides or nitrogen because it is focusing too narrowly on yield growth? This could have serious implications for run off risk and environmental outcomes. Research of AI tools is therefore key to ensure the correct parameters can be set and monitored.

In Conclusion

AI could be incredibly useful to many industries, including agriculture, but care does need to be taken. As with all new technologies, I foresee a cohort of early adopters that will essentially do wide scale testing before tools become more widespread and mainstream.