The Silent AI Revolution in Agriculture
India is known for a lot of things. But there’s one that Indians take deep pride in: mangoes. The sweetness and aroma of Indian mangoes is unparalleled. But Brazilian, Peruvian, and Israeli mangoes are more popular than their Indian counterpart.
In 2020, India exported just 0.52% of its nearly 25 million tonnes of fresh mango produce. In comparison, Peru sent out about 46% of its produce, Thailand 24%, Mexico 18% and Brazil 11%. It isn’t because Indian mango varieties are not as popular as the others. There is a great demand for Alphonso, Kesar and Totapuri mangoes. But Indian farmers can’t take their produce to the highest paying customers.
Here’s some food for thought: the export earnings from mango pulp in India are higher than the export earnings from fresh mangoes. Unfortunately, a farmer can sell their produce for a premium if it’s fresh produce that’s being traded. Pulp doesn’t bring in the same revenue as fresh produce. The reason is straightforward. Indian fruit doesn’t make the high water mark of requirements that the US and Europe, two of the most attractive export markets, have set.
“Exporters don’t want to bring fresh mango into Europe, and the main reason is a possibility of finding mango nut weevil,” says the owner of a food processing plant in Southern India. And it is difficult to decipher which fruit is likely to be infected. “That is how the Indian farmer suffers,” the owner explains. Let’s add another layer of complexity to this. The farmer can’t separate premium from non-premium mangoes, this further damages the revenue that can be made from the fruit.
Why are we talking about mangoes? Because mangoes are just the tip of the iceberg. This story is what plagues farmers across the country selling not just mango but everything they grow. Agriculture is important to India, it employs 54.6% of India’s total workforce. Agriculture accounts for around 19% of the gross domestic product (GDP) and accounts for exports worth over $50 billion. This sector drives the rural economy, and there are anywhere between 90 million and 150 million farmers in India. Yet, farming losses post-harvest continue to be alarming, issues of low farm yield (30–50% less than developed nations, often also lower than many developing nations) persist, and farming families remain poor.
Climate change is impacting weather patterns, which have a direct bearing on crop yields, and ultimately affect farmer earnings. All of India’s staple crops — rice, wheat, maize — face the prospect of diminishing yields in the years to come. The government has plans to reduce the impact of unpredictable weather patterns on farmers, but technology can change the trend.
To be sure, efforts have been made to fix some of these problems with the use of remote sensing, geo-referencing and drones, with varying degrees of success. While all of these use some form of artificial intelligence for processing inputs, newer advancements in AI hold a lot more promise. The processing plant owner is among a crop of exporters trying to bring greater use of new technology into the agriculture value chain. “For the mango issue, it is possible to set up an AI system that can look at different crop signatures, and identify three-four critical defects. The system can then develop software to monitor these defects and eliminate the pieces unsuitable for export as fresh fruit,” he says.
The market for AI in agriculture stands at around $1.7 billion and is expected to grow to $4.7 billion by 2028. Studies show that the adoption and application of AI in agriculture have limitations in terms of achieving scale, IoT devices, data annotation, data security and privacy, and technical understanding. But things are set to change, with stakeholders from government, businesses, NGOs, and individuals coming together to figure out effective ways to tackle different agricultural problems.
“Agriculture problems are massively societal in nature… The need for a solution also has to align with the scale of the problem or the complexity of the problem. So, it is absolutely imperative that the solutions we design have to work at speed, they have to work at scale and have to be sustainable.” –Anand Rajan, Mission Leader, Apurva.ai, Chief Platform Advisor, Societal Thinking at Ekstep Foundation.
At the first People+AI campsite, put together by EkStep on April 1, Krishnan Pallassana, Managing Director, India Programmes and Lead for South Asia
at Digital Green, addressed a session on understanding the ways AI can solve problems in agriculture. “To overcome those uncertainties (and) unpredictable variables that they are facing daily, farmers need advisories that will help them enhance their productivity… Despite all the facilities that are available right now, a majority of the farmers are poor,” he said.
Ongoing projects
There are some pilot projects that are experimenting with solving some of these problems at scale. There are over 2,300 agritech startups in India, solving for different aspects of agriculture. With generative AI, the underlying technology behind the hugely popular ChatGPT, newer possibilities have opened up.
For instance, by training large language models (LLMs) in local languages, farmers can have better access to weather advisories that will help save crops. Then there is the possibility of using generative AI to ensure farmers get access to the most appropriate kind of credit.
But, arguably, in order to make a meaningful difference in the life of farmers, the government is the biggest and most important stakeholder. And it is firing on all fronts in supporting industry and agriculturalists. Samuel Praveen Kumar, Joint Secretary, Ministry of Agriculture & Farmers Welfare, detailed some efforts the ministry has been undertaking, including those carried out in partnership with the private sector.
The AgriStack
The AgriStack is the first large-scale digital scale intervention planned for the agriculture sector in India. Conceptualised as India Digital Ecosystem Architecture (IDEA) in 2021, the project aims to solve many common problems faced by farmers through the extensive use of technology.
Source: IDEA concept paper 2021
The government is working on building the backend. The end goal is for every farmer to have a unique identifier, on the lines of Aadhaar, and have an interconnected set of databases that help provide optimum services. With LLMs trained on government data, the possibilities are huge. Farmers, for example, can get personalised advisories in their local language, understand the schemes they are eligible for, and train themselves in newer agri technologies.
Development Innovation Lab
Nobel laureate Professor Michael Kremer, a development economist, has established a lab called Development Innovation Lab, based in Chicago. “They are offering services to several developing countries for testing some innovations they are trying to evaluate,” explained Kumar.
The Indian government has identified four key areas on which it is working with DIL: climate resilient seeds, weather forecasting advisories, digital agriculture like AgriStack and Krishi-Decision Support System, and soil health management. “In these areas, digital tools can actually help in decision-making and also enhance the way we do agriculture,” he added.
MoU with Digital Green
In February 2023, the Ministry of Agriculture partnered with Digital Green to empower and strengthen the agricultural extension system. With this, about two lakh extension workers reaching about 13.5 crore farmers in the country would be humanly impossible. This partnership will leverage technology to bring relevant content to farmers, in order to enable them to do more with available information.
One objective is to develop a digital portal which will host audio and video content and advisories. For instance, developing content around millets, how to grow them, increase productivity and make them mainstream are among the current focus areas. The second is to have an enabling infrastructure for stakeholders across the board. “We’ll also have a capacity building program built initially around 40,000- 50,000 frontline workers and then probably upscale it to other parts of the country, starting with about 10 states and 170 districts,” Kumar added.
Work done with Apurva.ai
The government is also working with Apurva.ai, an open innovation platform that trains on underlying LLMs and uses AI to solve societal problems. The ministry has identified a few areas for further work with Apurva.ai:
- Integrated plant health management is where the government wants to ingest all the content that is available from various sources and strengthen the information available to farmers. It has rich content from 102 Indian Agricultural Research Institutes (ICAR), 731 Krishi Vigyan Kendras, and extension workers setup of state and central governments.
- Farmers’ innovation is another key area. For instance, the quality of onions grown in Nasik is much better than the quality of onions in Karnataka. With similar climatic conditions and facilities available, farmers are keen to learn practices that will improve their yields, and it is imperative to connect them to those who are finding ways to improve their yields. The government has identified about 2,000 such innovative farmers through the Agricultural Technology Management Agency (ATMA).
- Using Apurva.ai for in-house meetings to train it in language. The “brain” of the system will evolve and develop by learning through conversations. The expectation is to find a lot of insights and pointers for deciding further actions.
- Empowering farmer-producer organisations (FPOs) to increase the size of landholdings for farmers. A system like Apurva.ai can help provide an assessment of the on-ground impact of various government schemes.
Future-ready
With this context in mind, the group of people gathered to discuss the impact of AI on agriculture got thinking. Many have used technology to make a real impact on farmers’ lives. For instance, Naman Bajpai, the founder of Krishi Connect, a social platform for farmers, said they used chaos theory and big data to predict the movement of locusts for the Odisha government during the locust attack of 2020, which prevented extensive crop damage.
Further discussions were split across groups of people. These groups identified problem areas, possible solutions, and challenges faced towards realising those problems. The one area most participants had near-unanimous views on was personalising information for farmer needs. Hyper-focused, customised advisories, building trusted sources and making it all available in a language the farmer understands will be critical components for making a meaningful change through AI.
“Agriculture is one science where there is a lot of indigenous knowledge. And this knowledge has been empirically generated as well. So when you are creating a scientific wealth of knowledge, especially in light of climate change, you need a blend of scientific knowledge along with local and indigenous knowledge. It is important to package the solution in a socially, culturally and politically appropriate way.” - Ram Kiran Dhulipala, Senior Scientist at International Livestock Research Institute (ILRI).
One group, whose ideas were presented by K Yatish Rajawat, the CEO and founder of Centre for Innovation in Public Policy, discussed the need to establish trust in existing sources of information and use those to build customised advisories for farmers. One more area where AI can be beneficial for farmers is real-time information that has the potential for a tremendous impact on the decision of choosing which crop to plant.
Another group outlined the challenges in building trusted data sources from enormous amounts of input data that exist across different databases. “There is a data network where agricultural schemas are defined so that data can be flowing in a consumable, digestible format from trusted sources. For this AI to work well, the trusted source angle is supercritical,” said Parth Lawate, co-founder of Tekdi Technologies, presenting on behalf of his discussion group.
A farmer registry, with data attributes about farmers, will help make it hyper-local.
One proposed solution towards enabling this from the group was a curated AI farm advisory for the crop the farmer wants to grow, and the region they want to grow in. District-level or broad advisories aren’t equally helpful for all farmers.
Many more ideas were presented. Some conceptual, some rooted in current implementations. What was clear was that this is not a problem that can be solved in isolation. As with all other tracks at the Campsite, this one was also geared towards building a community that is going to keep the work of making a meaningful impact on agriculture using AI.
“This has to be done with communities,” said Anand Rajan, mission leader, Apurva.ai and chief platform advisor for Societal Thinking at Ekstep Foundation, who was moderating the session. “One is the community of experts coming together, the community of farmers and the ecosystem itself. No one solution can be deployed by one organisation or one team. But you will have to bring the larger ecosystem because the problem is not waiting for us to solve, it is mutating. We will have to really work at scale to solve this.”
While the focus is on building solutions and complementary capabilities with all stakeholders, it is also important to remember the importance of co-development, or engaging with the intended beneficiaries at an early stage of the development cycle, said Ram Kiran Dhulipala, Senior Scientist at International Livestock Research Institute (ILRI).
“The main focus of the entire ministry is how to de-risk agriculture, how to make agriculture more predictive so that we are able to kind of buffer them (farmers) from the vagaries of weather and unforeseen climate changes. And also to ensure that they get an assured price for their produce…In the next one or two years, you will see this transformation actually happening on the ground.” Samuel Praveen Kumar, Joint Secretary, Ministry of Agriculture & Farmers Welfare
Helping agriculture adapt to modern realities is a large problem, but it is also one that will impact the lives of every single Indian. It’s a discussion worth continuing, a community worth building.