AI for IGCs?

I recently asked a popular AI chatbot to write a biography of The Garden Lady. It started out on the right track, saying that The Garden Lady was C.L. Fornari, and that I was the author of several books about gardening. But things quickly went awry as it proceeded to write that my love of plants was developed by growing up on the shores of Cape Cod. Actually, I grew up in Colorado, California, Indiana and Wisconsin, only moving to the Massachusetts coast as an adult. Among other inaccuracies, it went on to say I had a degree in natural science from a university I’d never heard of. Clearly, artificial intelligence has no qualms about making stuff up.

In the green industry, AI is a topic of conversation at trade shows and in webinars, podcasts and articles. There are apps and websites available that use AI to answer questions about gardening and plants, undoubtedly with many more to come. In some ways, this technology can be useful for IGCs and our customers. You might ask one of these programs to generate an article for your company’s newsletter or blog. A photo generative program could be used to create illustrations for a presentation or website. But in these early days, we should take what the chatbots and digital art generators produce with a large dose of caution.

You may have read about the Toronto mayoral candidate whose staff used AI to create photos for his campaign platform. The illustrations appeared to be photos of real people around the city of Toronto, but they were exposed as being artificially generated because one of those images contained a woman who had three arms. What’s amazing is that none of the campaign staff caught this mistake before it was made public.

Such embarrassing errors aside, what is the harm of using a chatbot or photo generator to produce content for our garden centers? I would argue that at this point, the problem is that AI isn’t accurate enough to be trustworthy. In our business, we know that all gardening is regional, and until we can have faith that a chatbot will know the plants and garden problems in our particular area, the information produced is likely to be inaccurate or misleading.

For example, in my part of the country, we had a “polar vortex” freeze last February. The temperatures plunged overnight, turning what had been a mild winter into a sub-zero event. When our customers began asking about their dead or damaged roses, lavender and heather, we knew just what had caused the problem. The chatbots I consulted could list the various reasons those plants might have died, but at this point they were unable to factor in the real cause: sub-zero weather on the night of Feb. 4. The bots didn’t know that for some plants, those sudden temperature extremes are more damaging than weather that gets steadily colder into the winter.

Your experienced IGC staff will know if the summer has been unusually cloudy and cool, leading to a bad crop of tomatoes. They know when the typical insects show up, so when a customer comes in with a lacy leaf, they can not only offer an accurate diagnosis of the problem but may also know that the culprit has come and gone and that no treatment is necessary. If you’re in the business for more than a few years, you know what is currently in flower in the area, and you’re able to tell your customers if what they’re seeing is normal or problematic.

Inaccurate, misleading or just plain wacky garden information has been passed over the garden fence for years. We have often laughed or moaned about people thinking that ants help peony buds to open, or that Epsom salt is a magic cure-all. Now, that backyard fence has turned into a world wide web where inaccuracies live forever, joined by silly garden hacks and fake plant scams. From jars of banana peel water to seeds for rainbow roses, people can go down multiple misinformation rabbit holes that don’t positively serve them or the natural world.

Perhaps someday, AI will be the counterbalance to such fads and fabrications. Maybe in the future, a chatbot will tell people that an organic, balanced fertilizer is more beneficial than a stinking jar of banana water. Maybe AI will be able to know the latest plant scams and warn people not to waste their money. And it could be aware of the weather in any given region and how that affects the garden. But until these programs get a lot smarter, it’s up to us to continue to give our customers accurate, local information. Use digital generative software for illustrations, but look at them with a critical eye. Employ a chatbot to get you started on an article or paragraph for a newsletter, but have someone who is very familiar with the topic proofread the results.

I’ve tested four different AI software programs that are designed to answer questions about plants and gardens. I’ve found that about half of the time, they get the basic facts right, but the remaining 50% of answers range from slightly off to mostly wrong. In all these cases, I was told that I could go back and correct the answers so that the program would learn and improve. I’m sure that this is true, but I’d rather spend my time educating the public, not the bots.

AI and digital image generators have great potential moving forward, but right now we need a good amount of HI — Human Intelligence — to make sure we’re giving our customers accurate and useful information.