Getting your hands in the dirt is essential for gardening. A recent article about dirt reported, “Experts say that regular contact with healthy soil has physical and psychological benefits for adults as well as children.”

We continue to advocate direct involvement with installing and maintaining plants as an enjoyable and beneficial activity. Still, we cannot ignore the technology that surrounds us.

Three areas of technology are currently prominent:

• Energy production (solar radiation, wave motion, geologic heat)

• Robotic operation

• Artificial intelligence

Each technology relates to gardening in at least small ways. For example, energy production occurs in the garden with solar cells that activate path lights or fountains. Commercial farms and plant nurseries use robots extensively, while home gardeners have small-scale robots: irrigation controllers, lawn mowers.

Artificial intelligence, in contrast, has numerous applications for home gardening.

Three of our recent columns featured garden-related artificial intelligence tools. The first column in this series focused on prompt engineering, which directs AI to address a particular topic of interest. The second and third columns focused on using AI tools to identify plants, diagnose plant diseases and cultivate specific plants.

To review these columns in the Sentinel’s archive, browse to santacruzsentinel.com, search for “Karwin,” and scroll to “Using AI to nurture garden needs.”

Those columns address important aspects of gardening. AI is rapidly evolving, so those columns will become outdated in a short time.

Today’s column features other garden-related applications of this technology: AI tools to generate landscape designs. Like all of AI, these tools have been evolving and will continue to develop additional features. We focus on the easy uses of the current stage of this development.

We share AI-generated images for two designs as basic examples of this technology.

In the following sections, we explore how these tools relate to garden designers and encourage readers to try this technology for their design projects.

Parking strip design

Our first design relates to a planting area commonly available at residential sites: the parking strip (sometimes called the hell strip). This area can be challenging to design, develop and maintain. We generated a design based on California native plants because they require little maintenance once established.

We directed AI with the following prompt: “Photorealistic landscape design with California native plants for a residential parking strip.”

AI’s response to this simple prompt included several descriptive comments (not shared here) and a small image of an organized design with four kinds of plants:

Low groundcovers (2-8 inches tall)

• Carex pansa (California meadow sedge)

• Arctostaphylos uva-ursi ‘Point Reyes’ (bearberry)

• Achillea millefolium ‘Island Pink’ (island pink yarrow)

Mid-height perennials (8-24 inches tall)

• Salvia spathacea (hummingbird sage)

• Erigeron glaucus (seaside daisy)

• Epilobium canum (California fuchsia)

Structural plants or accent shrubs (2-3 feet max height)

• Artemisia californica ‘Canyon Gray’

• Ceanothus ‘Anchor Bay’

• Manzanita ‘Emerald Carpet’

Seasonal color

• Sisyrinchium bellum (blue-eyed grass)

• Dudleya farinosa (coastal dudleya)

My follow-up prompt yielded a high-resolution image suitable for newspaper publication.

This landscape design exercise, my first use of AI for this purpose, easily produced a plan that a gardener could install as presented or modify. The most likely modification would be to specify the dimensions of the planting space, because AI assumed an area larger than typical parking strips. The design can also be used to develop a larger part of the garden.

A second follow-up prompt added plant tags and re-shaped the planting area to resemble a typical parking strip more closely.

Rose border design

Our second landscape design exercise called for a garden border that features roses. The prompt was: “Create the image of a design for a 10- to 15-foot-long garden border featuring roses in Gertrude Jekyll’s style. Include a plant list with botanical names. Produce the image in the resolution needed for newspaper publication.”

A noteworthy detail in this prompt was to request the style of English artist, horticulturist and garden designer Gertrude Jekyll (1843-1932). To learn about her work, part of which included mingling roses and cottage flowers, browse to en.wikipedia.org/wiki/Gertrude_Jekyll.

When a gardener uses AI to generate a garden design, he or she can affect the result by specifying a favored designer’s style or a generic style, such as formal, natural, Japanese or any other.

In retrospect, the prompt did not specify the bed width or setting and did not request a photorealistic image. AI’s first response provided a diagram of the bed with a plant list and a caption with faulty details.

A follow-up prompt corrected these shortcomings to produce a new image. We specified a 15-foot-wide bed in front of a residence wall and called for a photorealistic image.

Human garden designers

Just as robotics can add efficiency to many forms of physical work, AI can replace a wide variety of cognitive work. This technology could concern cognitive workers who worry about losing their jobs, including garden landscape designers.

Today’s column does not recommend or even suggest that AI technology makes garden designers obsolete. A well-qualified designer brings a wealth of experience, creativity and thoughtful consideration to design tasks and responds to the garden owner’s preferences, priorities and objectives.

This column’s examples of AI landscaping design show the early stages of a process that should include a series of follow-up prompts to “tweak” the design through a series of drafts and eventually achieve the level that the gardener finds acceptable, satisfactory or even great.

A gardener with design skills can use this technology to produce a design that serves as the starting point for plant selection and installation stages that refine the design.

In another approach, the gardener could use AI to develop a preliminary design to discuss with a garden designer, who could bring a fresh perspective and new ideas to the project.

Our point is that AI can be a helpful resource, but not the ultimate solution to garden design.

This week in the garden

Your project for this week can be using AI for a garden landscaping exercise.

We have used chatgpt.com for this purpose, but there are several online resources. ChatGPT requires registration and is free, at least for the present.

The process begins with writing the initial prompt, anticipating that a series of follow-up prompts will likely be needed. A garden design prompt should indicate the size and shape of the area for the design, plant preferences, design style and any details of interest. The environment for the design should be described, especially when conditions are noteworthy: shady, wet, dry, windy, high elevation, etc.

Garden design with AI can be an adventure.

Enjoy your garden!

Tom Karwin is a past president of Friends of the UC Santa Cruz Arboretum and the Monterey Bay Iris Society, a past president and lifetime member of the Monterey Bay Area Cactus and Succulent Society and a Lifetime UC Master Gardener (certified 1999-2009). He is now a board member of the Santa Cruz Hostel Society and active with the Pacific Horticultural Society. For garden coaching info and an archive of previous On Gardening columns, visit ongardening.com.