AI in agriculture calls for imaginative policy-making

Ignoring technological changes in agrarian relations would be ill-advised

Studies on the industrial impact of artificial intelligence have so far focused on work and production in the services sector, such as programming or healthcare. There is little focus on manufacturing, but even less focus on its potential impact on agriculture – even though about 27% of all workers in the world engage in agriculture.

An even larger number – almost half of the Earth’s population – lives in households that are linked to agrifood systems. This is unevenly distributed: only about 3% of all employment in high-income countries is agricultural, while the same figure stands at 25% in China, 43% in India and 75% in Chad, for instance.

Employment in agriculture is highly heterogeneous. Landholding sizes differ, as do farmers’ connections to the land, whether self owned or leased. Farmers can produce for subsistence or for local, national or international markets and can have varying levels of productivity and profitability.

Eroding effects on working conditions and employment

Historically, the mechanisation of agriculture has resulted in dramatic changes in employment patterns globally. But we should understand AI as a continuation not only of the mechanisation of agriculture, but also the digitalisation of agricultural production and markets. In recent years, digitalisation in the form of data analytics, sensors and drone imagery has led to fears about the consolidation of agrarian landholdings under the control of a small number of people and corporations. These fears are based on increasing evidence of technology firms using farmers’ own data to outcompete them and even to acquire land.

The use of AI could hasten this process of consolidation, leading not only to mass unemployment, but also the degradation of working conditions for those still engaged in agriculture. The latter results from a change in land ownership patterns, because agricultural workers who do not own the land they work on end up being exploited more than landowning farmers.

Agricultural markets can be similarly affected. Many millions of people are engaged in agricultural marketing activities, including transport, storage, sorting and food processing. This sector has seen a flurry of investment in some markets with a view to ‘eliminating middlemen’ and controlling large parts of the food supply.

Food distribution and security

However, ‘AI in agriculture’ is a phrase that evokes smart robots darting amid rows of crops. It’s clear that progress in robotics is not a precondition for AI to transform agricultural production and marketing. At present, the capabilities of AI provide an informational advantage and are likely to consolidate agricultural production into large landholdings with landless workers and agricultural revenue being redistributed upwards.

AI is likely to displace not only peasant farmers, but also people who work to bring agricultural commodities to consumers. The development of agricultural datasets and advancements in imaging technology could be more important determinants of the effect of AI in agriculture.

The displacement of hundreds of millions of people, especially in developing countries, would be less disastrous if there were jobs available outside agriculture. But push migration of this kind does not create its own jobs and AI automation is beginning to affect urban jobs first. This is in sharp contrast to earlier waves of mechanisation in agriculture. The industrial revolution pulled people into urban centres with new jobs. The use of AI in agriculture might worsen an already dire situation in urban areas.

Even more worryingly, the use of AI in agriculture has implications for food security. There are both economic and technological motivations behind what this implies. Agrarian economies are held together today through a delicate balance of input subsidies, state procurement, market rules and food distribution programmes. While parts of this arrangement are decidedly unjust and exploitative, the system maintains a minimum level of food security. More inquiries need to be carried out into the potential effect of AI on this sectoral balance and on food distribution, in particular.

International chain of events

Any effects of AI in agriculture are likely to be global. If production is automated at a very large scale using one or a few AI models, a single error or misjudgement can affect food availability in entire nations.

AI is a centralising force in agriculture and centralisation harbours vulnerability. Its widespread use in agriculture, even if it is limited to advanced countries, can change terms of trade with developing countries, bringing down returns to labour in developing countries.

Industrial policy has returned to advanced economies, but under the guise of national security. It would be fitting for the spectre of AI in agriculture to shift focus from national security to preventing mass unemployment, wanton consolidation of production, food insecurity and a global race to the bottom towards all these effects. This should be accompanied by the deep involvement of farmers’ organisations in setting agricultural policy, imaginative expansion of government procurement of agricultural commodities and new models of collective data and AI ownership.

The economic perspective of a sizeable number of people primarily from developing countries is missed because it is neither a direct existential risk nor principally a socio-political issue of bias and discrimination. But changes in agrarian relations topple governments, transform social relations and change the course of history even today. The world would be ill-advised to ignore them.

Jai Vipra is an independent AI policy researcher.

This article was published in the Autumn 2023 edition of the Bulletin.






Join Today

Connect with our membership team

Scroll to Top