Hardcore gardeners sometimes, when no one else is listening, talk quietly to their precious flowers. But at next year’s Chelsea Flower Show, visitors will be encouraged to chat with its first AI-powered garden.
Garden designer Tom Massey has teamed up with Microsoft to create the Avanade “smart” garden. Sensors in the ground are paired with an AI trained on plant data and gardening advice from the Royal Horticultural Society, meaning visitors can ask the garden: “How are you?”
Massey said: “I could respond: I need a little more water, maybe I could use a haircut.”
In addition to making gardening easier and more fun, Massey said it could have broader applications. Wireless sensors hidden in the soil measure moisture, nutrient levels, acidity and alkalinity. These will be connected to a computer located in a pavilion in the garden, which people can consult to see how the garden is doing.
“It’s supposed to be interactive,” Massey said. “It’s supposed to showcase this kind of really new emerging technology and also save resources.”
He said simple sensor technology could be implemented on large properties, so instead of having automatic irrigation systems that turn on on a timely basis, plants are only watered when soil moisture drops below a certain level.
“We are wasting so much water and London is at risk of running out of water in the coming years,” he said. “Imagine how much water you could save in a large development if you only used water when necessary. Irrigation systems are turned on at regular times to water whether it is necessary or not.”
The garden itself would not have a futuristic design, he said. “We wanted it to feel green, organic and earthy.” It will feature climate-resistant mushrooms and trees, which will be monitored by an artificial intelligence assistant, and many peaceful green plants.
The RHS has adapted its system to be as energy efficient as possible. Jon Simpson, the company’s IT director, said: “With AI, you wonder if it consumes more energy when using data centres, so yes, it can be harmful, but the question is how it is used. What we are doing is training more traditional AI models with the RHS database, which is basically called a small language model, which is much more efficient and effective.”