Can AI address Africa’s agricultural trade deficit? AI in agriculture in Africa might help reduce trade deficits and boost output.
Most African nations waste billions of dollars in yearly trade earnings on food imports. There was a deficit of over $3 billion in agricultural trade between 38 nations in 2022, with Algeria, Egypt, Nigeria, Morocco, and Angola ranking highest.
Contrast this with the fact that 23% of GDP in sub-Saharan Africa comes from agriculture, more than 60% of the continent’s population works in the agricultural sector, and 65% of the world’s remaining arable land is on the African continent.
The trade deficit also impacts the fiscal health of nations. Countries are more sensitive to external shocks when they cannot compensate via other exports, such as oil, as is the situation with Nigeria, Angola, and Algeria.
Russia’s termination of the Black Sea grain agreement and embargo of Ukrainian supplies have recently disrupted the cereal value chain, drawing attention to the need for agricultural reform in Africa.
These upheavals occurred in the same year that artificial intelligence (AI) saw extraordinary advancements.
AI in practically any industry, including farming, is considered a potent productivity enabler. But can AI boost Africa’s domestic food supply and reduce trade deficits?
Using AI to increase agricultural output in Africa
Despite AI’s current popularity, at least two years ago, it was regarded as a tool to assist farmers in increasing their revenue. In 2021, Daphney-Stavroula Zois, a professor of electrical and computer engineering at New York State University at Albany, collaborated with the non-profit AGRI-WEB to improve agricultural production prediction models for smallholder farmers in Ghana. Although “there are many other uses of AI,” she puts it.
“Examples include, but are not limited to,” Zois explains, “predicting crop amounts, providing guidance on irrigation, when, where, and what to plant, what price they should sell their products at, and to whom.”
Google included her work as one of thirty sponsored by its “AI for Social Good” program, which funds research to aid groups often left out of AI’s benefits.
A smartphone app for illness diagnostics in Uganda and a model to monitor grassland conditions and forecast local market pricing in Kenya are two more examples of Google-funded AI initiatives in African agriculture.
Artificial intelligence (AI) is being considered by the creators of technology that has previously proven useful to farmers. A software named “Afriscout,” created by the United States-based global development organization Global Communities, offers semi-nomadic pastoralists digitalized maps of traditional grazing sites overlaid with satellite-derived vegetation information to help in migratory decision-making.
Researchers discovered that over two years, pastoralists who had access to Afriscout maps had a net gain in wealth of over $4,600 compared to their counterparts who did not. To “optimize grazing plans for land and herd health,” Afriscout hopes to use AI.
“Sound use of AI can definitely help farmers address both short-term and long-term issues with farming activities,” says Zois. In addition, “it can also guide them through market globalisation,” she adds.
AA’s lack of data and financial resources hampers the adoption of AI.
Some hopeful observers have speculated that in the next years, thanks to advances in AI, a scalable solution may emerge in Africa’s agricultural industry. However, widespread uptake among farmers is still a ways off.
Zois says “access to infrastructure and data quality” are the two biggest obstacles. “State-of-the-art AI algorithms nowadays need access to high-quality data and reliable infrastructure to function… “The cost of collecting and processing such data with the current infrastructure is enormous,” she argues.
The most effective AI models need massive amounts of input data. A model is “trained” first by being exposed to labeled data. For instance, computers may be trained to identify “infected” and “healthy” cassava crops if they are shown enough labeled photos of sick crops. The model is next tested using data that it has never seen before, known as the “test set” or “validation set.”
All of it takes a lot of cash, the kind of capital that is only accessible to businesses with abundant reserves, such as venture capital (VC) money or other types of investment. ChatGPT’s parent company, OpenAI, secured a billion dollars in seed investment from companies like Amazon and Elon Musk in 2015. This allowed the business to hire some of the best minds in artificial intelligence.
Pitchbook, a database of venture capital investments, estimates that in the first quarter of 2023, Silicon Valley businesses received about half of the $11.1bn invested globally in startups using artificial intelligence and machine learning. In contrast, since 2020, African businesses emphasizing artificial intelligence have raised $255 million, with half of it coming from the Tunisian company InstaDeep.
Even though a Kenyan climate tech business, Amini, recently raised $2 million in a pre-seed fundraising round from an EU-based VC, this is still a tiny fraction of the pie compared to the proportion for agriculture-focused companies. Amini built a data aggregation platform for African food and beverage manufacturers to optimize their operations.
Other African firms have used preexisting algorithms to create agricultural solutions.
A startup investment of $200,000 was made in Rural Farmers Hub, an Abuja-based organization that offers advice services to farmers using satellite-based remote images.
Our current approach is trained on multispectral data (satellite images often include many bands of information). “We can select specific bands or combine multiple bands (obtaining a band ratio) depending on what is being monitored; this is also called feature extraction,” explains Isah Abdul-Azeez, a machine learning engineer at Rural Farmers Hub.
Our team has done substantial studies on satellite imaging (at various depths) and microwave connections, although they are still developing their respective products. According to Isah, open-source datasets are his go-to.
Farmingtech, situated in Kenya, likewise uses AI innovation on a shoestring. The business received a $500,000 grant from the international non-profit Heifer International in February and has included an artificial intelligence tool into its “DigiCow” software to boost dairy farmers’ output.
Ronnie Kimani, regional sales manager at DigiCow, says, “AI enables us to understand herd structures in locations, understanding new additions and deaths; this supports future predictions on key things such as expected milk production.”
Some worry that AI may have harmful repercussions, even if firms like DigiCow and Rural Farmers Hub collect enough money to develop their proprietary algorithm.
“There are a number of actors in the system who can exploit AI for nefarious purposes to influence market prices and beyond,” explains Zois.
It would be simple to introduce inaccuracy into AI systems if they were vulnerable to assaults, such as corrupting the data required to train the model. Because of this, predictions will be off, leading to bad advice for farmers and other problems.
There are fundamental problems that can’t be fixed with technology alone, such as the power imbalance between nations.
For instance, economic watchers on the continent have long lamented that African farmers cannot compete with agriculture in Europe and the United States because of government subsidies. One-third of the EU budget goes toward agricultural subsidies, allowing European farmers to sell low-priced goods to African customers.
Those who indirectly damage Africa’s agriculture sector are free to develop AI; this might compound the historical causes of the continent’s severe trade imbalances. “AI can easily learn from human behavior and copy all the bad traits,” adds Zois.
