Will artificial intelligence revolutionize farms as much as the rest of the world? 

By Scott Gabbard, Purdue On The Farm 

There has been a lot of discussion during the last year about artificial intelligence (AI) and how it will revolutionize the world, it’s power consumption and the sheer processing power it requires. This was very apparent when I went past a Facebook (Meta) processor “farm” while touring real farms in Idaho last month. 

I had already started writing this column when I got a nice write-up in my Inbox from Todd Jantzen. It was a good read and some of the legal themes and limitations he discussed are very relevant to the agricultural sciences, as well. 

Talking about AI to a farm audience means you have to clarify whether you’re talking about your herd breeding program or a large language model, often with a couple of chuckles from the audience. In the research world, one has to clarify whether it is a large language (General) model or a smaller (Narrow) model used for data assimilation or machine learning. 

Everyone reading has experienced Narrow AI. It could be an engine’s computers adapting to a driver or Siri accommodating the user’s request better over time with repeated usage. Today’s topic will be the General AI that has been in the news since ChatGPT entered our lexicon. 

You don’t have to have too long of a memory to remember when Google was going to change the world. It has. At the end of the day, the person on the other end of the search screen still needs to have some working knowledge of what is going on. 

While insect or disease management for corn and soybeans can be found with Google, applying what Alabama does to Indiana can be helpful but it can also get the farmer in a lot of hot water, too. 

The same can be said of the new AI products being offered now on multiple platforms. They also have limitations. Knowing their limitations and how to work with them is important if one is to use them effectively. 

Value and limitations 

While several Purdue units and researchers use a multitude of platforms of large language models or generative AI, the two that are available to all employees are Microsoft’s CoPilot.ai and Perplexity.ai. Considering the capability of these models; faculty, staff and students are to adhere to Purdue policies of intellectual property and proper attribution. 

While the previous paragraph sounds like something that has nothing to do with farming. It does. Never giving a model personal information is one thing. Verifying whether the model is “learning” from your thoughts is another (a big concern for researchers). Now, lets discuss the value and the limitations. 

I asked Microsoft CoPilot the same question three times this question: Explain to me what tar spot on corn is, how it spreads and what can be done to manage it? 

I received three different answers with varying levels of detail. All three gave me answers on how it spread with varying levels of detail. None of them gave me fungicide. All three iterations emphasized cool temperatures and high humidity, but only one mentioned frequent rainfall and two mentioned dense plant populations. 

Two of them had six management strategies, one of them had five. The one with five omitted optimizing plant density. Pretty good, two of them suggested reaching out to ag extension or an agronomist for more information. 

Perplexity.ai was better in this use case. It was able to the give growth stages for scouting, management and even suggested a few fungicides. Better yet, it also cited its sources. 

It cited Bayer, Michigan State, a hybrid company I didn’t recognize and many more. It cited Purdue first and often, including the Pest and Crop Newsletter. Click on the links and it pops up the original articles for further reading. It’s a great way to bone-up on information that is out there and readily available. 

Better organize our thoughts 

In all, this new technology will change how we find information. It will help us better organize our thoughts and provide insights that may have been missed. In the end, boots on the ground still matter. 

Human interaction still is of value. Just like prior technological breakthroughs, we’ll just accomplish more in a shorter period of time. If you’d like to learn what your children or grandchildren are learning about AI in 4-H, just type in this shortened link: extension.purdue.edu/4-H/volunteer/resources-and-development/stem-volunteer-resources.html 

This summer, we’ll be using AI to stitch hundreds of photographs, rewrite some e-mails and even see if it identifies something we did not notice. It’s nothing to be afraid of, just another tool in the toolbox. 

Funny to think about it, but some of you might be using AI while reading this column as your equipment plants the field. It’s certainly not 1939 anymore (the year “The Wizard of Oz” was released).