Understanding the Legal Issues for AI in Agriculture

FBN recently unveiled “Norm,” an artificial intelligence (AI) advisor for FBN members. Norm may be the first dedicated artificial intelligence platform for farmers, but it will not be the last. AI is definitely on the rise in agriculture and elsewhere. This post explores questions about AI’s use in general and for agriculture.

What is AI technology? 

First, some basics. AI stands for “artificial intelligence.”  According to Dr. Anastasia Lauterbach, contributor to The Law of Artificial Intelligence and Smart Machines, AI should be thought of as “narrow” AI or “general” AI. Narrow AI is focused on solving a particular task.  When we talk about “machine learning” (ML) that is generally what we are talking about. ML involves a computer using vast amounts of data to make a decision—but not just any decision but to continue to make better and better decisions. ML allows the computer to learn from its past decisions.   

General AI is what we have been talking about more recently. According to Dr. Lauterbach, general AI is similar but seeks to mirror the behavior and capabilities of a human to solve problems. What we are seeing now with ChatGPT and other technologies is “generative” AI, which is a type of general AI that can generate new content that never existed.  Generative AI like ChatGPT uses information from vast amounts of data that is publicly available, creating original content in response to inquiries from users. 

How might farmers use AI? 

If you think about how much of farming involves looking at vast amounts of data to make informed decisions for the future, you can start to get some ideas about how AI might be helpful in increasing plant and animal production. What variety of corn should I plant this year?  No human agronomist could analyze every variety and determine what might be best for a given field given the soil profile, weather predictions, pest predictions, etc. But AI, with the right training data, could do that.

Likewise, we are seeing narrow AI already used in applications like see-and-spray technologies. ML trained equipment learns to spot weeds and differentiate those from crops, so that applicators only have to use pesticides on the weed and not broadcast broadly.

What might go wrong with AI? 

Putting aside the Hollywood doomsday predictions of AI becoming so intelligent it decides to destroy humanity for the good of the planet, there are other more immediate and realistic concerns. AI is only as good as the data that trains it. If the training data is corrupt or skewed by a company to increase shareholder value, such decisions could create problems. Imagine a seed company figures out what data something like ChatGPT is using to make farming decisions and that company starts flooding the internet with false reports about its seed—data what we (as humans) never find through search engines but AI zeros in on the data.  That seed company could skew the AI results to favor its products. Just as companies use tactics to game search engine optimization (SEO), we are going to see marketing departments start to try to game AI systems to skew recommendations in their favor.

Remember, too, that AI has to make mistakes in order to learn what is right. This means mistakes will be made on the road to the future.

What are some of the legal implications for AI? 

Privacy is a big concern. Imagine an AI tool that uses farmers’ ag data to make decisions. Farmers using the AI tool get crop recommendations, based upon the field data they input and that data from other farmers using the platform. It may not be immediately evident to the farmer that their data is essentially known to all other farmers who use the AI platform.

Will AI respect the rights of ownership of data, copyrights, and other forms of intellectual property (IP)? Currently, only humans or companies can legally create or own intellectual property. What happens when AI uses proprietary information to create new, derivative content? Who will own the resulting IP?

Is it fraudulent for companies to attempt to fool AI into making decisions that may not be based upon accurate data but instead based upon false data published to skew AI results in the company’s favor? I don’t know of any laws that address this scenario.

Finally, there is a concern that people will make decisions based upon AI generated false information. There are many examples of “deepfakes” on the web—these are videos of politicians or celebrities saying things they never actually said, but AI is able to generate a convincing fake.

Conclusion

There are concerns, but there are enormous positive implications for the use of AI in agriculture and elsewhere. AI can be much better at analyzing vast amounts of data in an unbiased fashion than humans. AI can see patterns and problems that we cannot. Like all new technologies, the good will bring some bad that will have to be navigated along the way.