Is there a potential in adopting Artificial Intelligence in food and agriculture sector, and can it transform food systems and with what impact?

case study

Is there a potential in adopting Artificial Intelligence in food and agriculture sector, and can it transform food systems and with what impact?

By Thembani Malapela

In brief

  • AI and Machine Learning technologies could transform the agri-food systems and help ending hunger.
  • UN-Wide System is adopting the Artificial Intelligence in its programmes.
  • The Food and Agriculture Organization of the United Nations (FAO) is adopting and experimenting with AI and Machine Learning across the agriculture domain.
  • FAO is also aware of the impact of AI and is now a signatory to the "Rome Call for AI Ethics"


In 2017, l wrote a blog on whether Artificial Intelligence (AI) can improve agricultural productivity and at that time, the question was befitting, as this technology was still not yet widely appreciated. However, three years have passed, and the use and adoption of AI has grown in many sectors. Additionally, Covid-19 has forced governments, companies and individuals to rely heavily on digital technologies.

Attention is increasing on the potential of AI and Machine Learning (ML) to the Food and Agriculture sector, especially on transforming food systems and application to different value-chains. Recently, the International Finance Corporation (IFC) published a report on Artificial Intelligence in Emerging markets, and regarding agribusiness, it noted that, “Artificial intelligence can spur progress toward meeting Sustainable Development Goal #2—to end hunger, achieve food security, improve nutrition, and promote sustainable agriculture” (p.72). The increased access to faster internet connection(s), investment in connectivity infrastructure and wider availability of smart phones (and other handheld devises) provides a rock bed for the uptake of digital technologies and especially AI amongst farmers even in developing countries. (Read example on M-Shwari Case Study)

Meanwhile, across the UN system-wide, the adoption of AI is well documented by the International Telecommunications Union (ITU) in a report that highlighted the use cases of AI amongst UN Agencies. This year the “AI for Social Good Summit” is pencilled for 21-30 September 2020. The summit is organised by the ITU and the XPRIZE Foundation, in partnership with UN sister agencies, Switzerland, and ACM. It offers a platform for discussing various AI related issues and application.

In authoring this piece, I have no intention to be an authority in this technology nor in agri-food systems; however, from a knowledge sharing perspective, l would like to disseminate some initiatives ( known to me at this time) on the use of AI and Machine Learning in food and agriculture by the Food and Agriculture Organization of the United Nations* . The blog will conclude with some issues arising in the adoption AI. 

FAO on Artificial Intelligence

Some snippets of publication of issues related to AI in FAO started appearing from 2019, initially with the launch of a mobile app to monitor fall armyworm. The availability of farm data increased and paved the way to develop and deploy AI in agriculture. The trend has also been spurred by major ag input companies, equipment manufacturers, and service providers who produced products and services that either consumed or diffused farm level data (Rakestraw & Acharya, 2017). Furthermore, the development of AI algorithms and AI applications for the value chains have placed the private sector as a trendsetter in the digital agriculture ecosystem.

FAO China, in 2019 hosted a dialogue in partnership with the Chinese Academy of Agricultural Planning and Engineering (AAPE) themed the “2019 Dialogue on the Application of AI in Agriculture”. The dialogue gathered around 60 representatives from government, institutions, academia and private sector in Beijing (FAO China). That meeting offered insights from public sector, and founders of leading enterprises in the industry, and participants shared their practical experiences, cutting-edge technologies and concerns with AI adoption.

Recent examples on AI and Machine Learning in FAO

Some of the activities and projects where FAO has adopted the AI and Machine Learning in agriculture are listed below. This by no means an exhaustive list but representative of the few l managed to glean during the brief research and rightfully so have specific project leads within the organization.

  • Fish Species Identification.

Fish species identification is classification and categorizing fish, by fisheries taxonomic experts, based on external morphological features, including body shape, pattern of colors, scale size and count, number and relative position of fins, number and type of fin rays, or various relative measurements of body parts (Strauss and Bond,1990).

While a number of tools and guides (including web resources such as Fish Base and the Catalog of Fishes) were developed by FAO to support taxonomic experts, latest technologies have provided additional ease to this process.FAO has explored the use of Artificial Intelligence in Fish Species Identification by using Google Cloud AutoML. Cloud AutoML is a suite of machine learning products that enables developers with limited machine learning expertise to train high-quality models specific to their business needs. It relies on Google’s state-of-the-art transfer learning and neural architecture search technology. After testing different approaches, the FAO CSI team concluded using Google AutoML performed better and was quicker to develop that building the models ourselves even when starting with massively pre-trained models. The team discovered that good quality data was needed and a good mark-up workflow.

  • 1b.       Species recognition

iSharkFin is an expert system that uses machine learning techniques to identify shark species from shark fin shapes. Aimed at port inspectors, custom agents, fish traders and other users without formal taxonomic training, iSharkFin allows the identification of shark species from a picture of the fin. The iSharkFin takes an interactive process.  Users only need to take a standard photo, select some characteristics of a fin and choose a few points on the fin shape, iSharkFin will automatically analyze the information and tell you the shark specie from which the fin comes (Strauss and Bond,1990)

  • Advanced Land Monitoring

The FAO system for earth observations, data access, processing & analysis for land monitoring (SEPAL) helps countries measure, monitor and report on forests and land use, offering unparalleled access to granular satellite data and computing power, for improved climate change mitigation plans and better-informed land-use policies. It uses advanced Cloud computing, AI and machine learning to provide comprehensive image processing capabilities and enables detection of small-scale changes in forests, such as those associated with illegal or unsustainable timber harvesting. The system provides the capability to countries to develop robust national forest monitoring systems, detecting Forest Degradation and Forest Fires. (Bravi, 2019)

  • The FAO Geospatial Machine Learning pipeline: Extracting objects of interest from remote sensing imagery

Another exploratory work was using ML to extracts objects of interest from remote sensing imagery via a GeoML pipeline. More specifically, the machine was trained to detect palm trees (with some hints from Penn State on optimization). The approach was applied to fish cages and fish drying racks and vessels. A refinement was counting the identified vessels. The GeoML pipeline was deployed on Google ML and standalone for desktop use. See presentation here

  • Management of scarce resources

Artificial Intelligence is used by FAO to make better use of scarce resources like water and energy. Achieving Food Security in the future while using water resources in a sustainable manner is a major challenge for our and next generations. Agriculture is a key water user. A careful monitoring of water productivity in agriculture and exploring opportunities to increase it are required. FAO has developed a publicly accessible near real time database using satellite data that allows monitoring of agricultural water productivity and do advanced water management (Read about WaPOR

  • Plant pest detection

Fall Armyworm is spreading fast across many parts of the world, including sub-Saharan Africa, devastating crops and farmers’ livelihoods (FAO,2018). A mobile phone application called FAMEWS, which uses machine learning and artificial intelligence, offers hope in tackling the pest problem. [2] Farmers can easily detect Fall Army Worm damage by using mobile phones. This augments the human intelligence and serves extra extension capacity. At the moment almost 20 plant pests can be detected already by using a mobile phone. [ Listen to the podcast here]

  • Detection of Agriculture Stress

FAO has developed the Agricultural Stress Index (ASIS), a quick-look indicator for the early identification of agricultural areas probably affected by dry spells, or drought in extreme cases. It monitors agricultural areas with a high likelihood of water stress/drought at global, regional and country level, using satellite technology. Drought affects more people than any other type of natural disaster and is the most damaging to livelihoods, especially in developing countries.

  • Support to investment

Earthmap is a new FAO tool based on Google Earth Engine providing simplified access to complex datasets. It facilitates the understanding of land cover and land use dynamics processes for designing and assessing baselines, monitoring and evaluation. (Read the emerging news item)

Potential brought by these technologies and issues arising

The food and agriculture sector still faces the inherent challenge of feed the ever-growing world’s population. Additionally, over 820 million people go hungry, around 2 billion more lack sufficient micronutrients and another 2.5 billion consume excess calories for their needs.

Through innovation, technology adoption and mechanization the agri-food systems has managed to survive until now, with a growing fear of whether amidst mounting global challenges (such as climate change, political unrests, etc) would the food systems cope with increasing food demands. The adoption of technologies, such as AI and Machine Leaning, could improve agricultural productivity.

There are claims that claims that AI capabilities could someday exceed human capabilities. As a result, there is a growing call that these new technologies should be researched and be produced in a way that they do not interfer with human rights, are environmentally friendly and thus not marginalizing the poor and most vulnerable. In this vein, FAO is one of the signatories for the “Rome Call for AI Ethics"

Rome Call for AI Ethics This is ethical resolution on Artificial Intelligence (AI) that stress the importance of minimizing this technology's risks while exploiting its potential benefits. The Rome Call for AI Ethics refers to the need for a highly sustainable approach, which also includes the use of artificial intelligence in insuring sustainable food systems in the future. (FAO, 2020).


The future of Artificial Intelligence seems to be bright within the food and agriculture sector and with a likehood of transforming agri-food systems if upscaled homogeniously. FAO has adopted this technology in the areas highlighted above and the signatory to the Rome Call for AI Ethics affirms FAOs commitment to adopting sustainable technologies with a consideration to the human rights and rights of the poor and the marginalized.

Disclaimer:  This piece is my opinion piece based on the published reports and news and does not attempt to express the views, or represent the views and position of the Food and Agriculture Organization of the United Nations.

Post your comment

Log in or register to post comments