How AI Can Ethically Fight World Hunger

In 2021, 828 million people were affected by hunger, an increase from 2020. Hunger is still a growing problem across the globe but recent developments in AI technology could change that. AI is revolutionizing how we farm at every step of the process and it has the potential to massively increase efficiency – but if this new technology isn’t handled ethically, it could end up doing more damage than good.

In this episode, we’re going to talk about some of the ways AI has changed farming methods and what ethical questions this raises, how ethical business decisions can incorporate AI, what the future could look like if we use AI responsibly for farming, and what might happen if we don’t talk about ethics.

Changes to the farming process

There are many ways AI is improving the process and efficiency of growing crops, and we’ll go over a few examples to give you an idea of what these developments look like.

One example is targeted herbicide application. AI can identify the difference between crop plants and weeds and use this information to spray herbicides only on the weeds instead of across entire fields. This can cut down on herbicide use by almost 59%. In the long term, it can replace the need for genetically modified cash crops, reduce runoff damaging the local environment, and cut costs for farmers.

AI can also be used for monitoring field conditions: AI drones and satellites can collect data on soil conditions, including the amount of water in the soil. They can also identify if crops are showing signs of disease, withering, or other issues, and monitor stages of growth. This data can save a lot of time for farmers so they can focus on solving problems instead of looking for them. It can increase the efficiency of yields – generating more food in local communities.

While these new strategies have a lot of potential, they are also creating some ethical concerns around data privacy and data ownership. In order for AI to monitor crop conditions and do targeted herbicide application, there has to be a lot of data collected. Cameras, satellite images, and other sensors may be needed to measure and interact with the farm, and it might not be clear who owns this data.

AI companies need a lot of data to train the algorithms on, so they’re highly incentivized to claim ownership over the data collected while an AI-powered tool is farming. However, they’re taking images and other information on someone’s private property, and it’s not always clear where the boundaries are for data protection. In order to use the AI software, farmers may have to agree to give their data to the company developing the AI, if the terms and conditions require it. However, in some places this might be illegal.

There are a few other considerations to keep in mind with data privacy, especially what might happen if things don’t go according to plan. An AI might accidentally capture private information if a camera is directed the wrong way. Data transmissions might be compromised, allowing bad actors to steal data. In cases like this, who is responsible for the damage? The companies developing the AI tech? Or is it considered just another risk of using technology?

AI-informed business decisions

Next, we’re going to talk about how AI for data analysis and predictions is changing how people make informed business decisions for farming.

One way AI is changing how farmers plan is through creating localized, detailed weather and climate predictions. AI can create more accurate weather predictions by analyzing massive amounts of data, which in turn means farmers can better prepare for adverse weather events and even plan their crops around predictions of droughts or floods.

This can reduce the impact of natural disasters on vulnerable populations and make it easier for farmers to adjust their business strategies to account for changes in climate before they happen. The more time farmers have to prepare, the better – they can invest in different crops and equipment, plan new mitigation strategies, and develop new tech before it’s needed.

AI can also take things a step further and combine climate and weather predictions, farming data, and other insights into models that predict which crops will perform well for a given farm. AI can make predictions about the demand and best timing for certain crops so farmers can make more educated choices about what to grow and when.

While all these changes sound great at first, like many AI applications, there are some serious ethical questions to be worked through. One of these problems is data bias. Data bias occurs when the information used to train an AI doesn’t capture important details that affect the outcomes of what the AI predicts. When an AI is trained on an incomplete data set, it may make poor decisions that could have some severe impacts.

For example, data bias may cause AI systems to prioritize high-yield crops without taking into account the local ecosystem and economy. If it doesn’t have enough data on a particular location, it may recommend commercially popular crops over local ones, even if local crops would have higher yields and support the local economy better. It also might not identify local pests and problems if the AI wasn’t trained on data that would recognize it.

Another example has to do with the size and structure of farms. If an AI is trained on data that is mostly from large scale commercial farms, small farms using it to predict what crops to plant and other business decisions might end up bankrupt.

The ethical dilemma here is how data bias is dealt with. To reduce data bias, AI training sets can be tailored to include data for as broad a set of conditions as possible: big and small farms, farms that are commercial growers, farms sustaining the local area, and more. However, getting a data set this large runs into the problems with data privacy we talked about earlier. Are farmers ethically obligated to provide training data to reduce data bias? Should companies have to pay for this data?

Companies developing AI tools can also set up algorithms to check and clean the training data of bad data points, false measurements, and other errors that might affect the AI’s decisions. This is another important strategy for reducing data bias, but if the decision to do this is left up to the developer, they may not prioritize the time and money to build these systems. Should companies training AI be regulated to ensure these safeguards are in place, or would this hamper technology development?

Giving some thought to what the AI should prioritize is also important. When making decisions, the tool may have to choose an option that favors one goal at the expense of another, and what decision it makes has to do with what the people developing it have trained it to prioritize. For example, it might prioritize growing the maximum amount of food while sacrificing the integrity of the local environment. This is another problem where regulations might be an answer, but people might not know how the AI is making decisions or what data it has been trained on.

Transparency around how an AI is trained, the data it is trained on, and what it prioritizes can help people make decisions about whether the tool will make good recommendations for their farms. However, too much transparency about training data starts running into the data privacy issues again. Finding the right balance won’t be easy, but the more we look at each of these issues and dive deeper into them, the more we can make informed decisions about how to use AI farming tools in a way that helps communities grow food sustainably – which can help end world hunger.

Intensifying economic divides

We’ve talked about some of the ways AI can make farming more efficient and help farmers make more informed decisions when running their businesses, both of which can put a serious dent in food insecurity across the globe. But what happens if things go wrong?

AI tools in farming have the potential to drastically increase gaps in inequality, damage local economies, and concentrate wealth in a few large corporations. In the long run, this can set up massive, systemic failures that could send shockwaves through the international food chain.

Getting access to AI predictions, tools, and equipment might be complicated – if the AI tools making predictions are owned by companies, small farmers might have to pay significant fees to have access to the AI models predicting which crops would be the best choice. The more local a farm, the less likely they’ll be able to afford access to this information. This means larger farms that already have the income to invest in new technology could avoid problems like droughts, decreases in demand, shifts in climate, and others, while smaller farms are more likely to be unable to compete.

This is even more critical given the increase in weather variability and severe weather events due to climate change. Farming has gotten riskier, and smaller farms may already be struggling to ride the waves. If local farms can’t sustain themselves, they may have to sell their farms to companies that might not be based locally – and they often have different priorities than small farms supporting a community. They may switch to exporting cash crops instead of growing a variety of food for a local community, damaging local economies in favor of larger profits elsewhere.

On a large scale, centralizing farming like this makes food production more dependent on fewer suppliers. This leads to less diversity in farming strategies and in the supply chain, lower food quality and variety, and more volatility in prices.

It also increases the risks associated with some potential crises: less diversity in crops means a disease could wipe out entire strains of crops, like what happened with the most commonly grown banana in the 1960s. That type of banana is now extinct, and was replaced with a less flavorful variety. If other varieties of bananas weren’t grown at the time, there wouldn’t have been anything to replace the original banana with.

Conclusion

We’ve discussed some of the ways AI is changing how we farm, how these changes are raising questions about data privacy and ownership, data bias, and where things could go wrong. AI has the potential to empower communities struggling with food insecurity by revolutionizing farming – but we may need to figure out how to use AI farming tech ethically for this to work.

There are many ways to approach ethical AI farming: subsidies to help small farms gain access to equipment, better AI training data, independent checks for data bias, involving small farms in AI tech development, and more. What do you think? What questions should we prioritize answering? What do you think ethical farming with AI looks like? Do you think AI tools could solve world hunger?