With its cascade of blush-pink flowers offset by a verdant halo of foliage, it was certainly a plant that stood out. But for RHS botanist Dr Jordan Bilsborrow, when he saw the picture of a weeping begonia it just felt… wrong. Jordan is part of the RHS Horticultural Advice Team, so identifying obscure plants and, on occasion, even hand-drawn illustrations sent in by members, is just another day in the office. But this begonia? This was different.
The image had pinged into his inbox on a typical spring day at RHS Garden Wisley, submitted by a bewildered RHS member. What is this begonia, they asked; their mother would really love one, they added; but they’d been unable to track one down anywhere. “Initially I thought that there’s no way that’s real. It looked too… weird. Too good to be true,” says Jordan. “The leaves weren’t right for a begonia, for a start, and the inflorescence definitely wasn’t right. I showed my colleague, RHS Botanist Saskia Harris, and we agreed.” Could it be possible that this RHS member had somehow stumbled upon a new species? Or was there something altogether more sinister afoot? The pair then approached US begonia expert Mark Tebbitt, who said the inflorescence on the mystery begonia looked more like that of a flowering cherry stuck onto a strawberry plant. One thing was definitive: this was no begonia.
Intrigued, Saskia continued to dig, throwing herself into amateur sleuthing with the single-minded relentlessness of an on-screen detective. She found a Facebook page where dozens of plant-lovers professed their admiration for this new begonia. Further investigation revealed the image had also made it to TikTok and Pinterest, with similar levels of appreciation. Eventually, Jordan came across what must have been the source of the image, and the root of this modern mystery: a company based in the United Arab Emirates claiming to sell packs of these weeping begonia These are fleshy, rounded, underground storage organs, usually sold and planted while dormant. Examples include daffodils, tulips, hyacinths, lilies, onions and garlic. The term is often used to cover other underground storage organs, including corms, tubers and rhizomes.
bulbs for £11. Instantly, alarm bells began to ring.
“Firstly, begonias have tubers, not bulbs,” says Jordan. He scouted around the website, clicking on other cultivars that were equally eyecatching and obscure. “There was a bright blue one; others had flowers that seemed to be dangling from fishing wire. It was not the sort of site you’d normally buy plants from.” Jordan’s instinct proved correct. The too-good-to-be-true begonia was just that. According to fact-checking website Snopes.com it was a scam. But more than that, these images of fabulous new plants, realistic enough to intrigue experts on either side of the Atlantic, had not been produced by human hand at all, but by artificial intelligence (AI). And they’re not the only ones. AI images are everywhere. Unlike computer-generated images (CGI), which are painstakingly made with rendering software, AI pictures are created by a machine capable of mimicking patterns, textures and styles after being trained on data sourced from millions of real images across the internet. More often than not, the AI-generated plants are fantastical species you wouldn’t expect anyone to fall for, such as a plant with flowers resemblant of a kitten’s face, called the Cat’s Eye Dazzle. It’s a particularly silly example, yet it still reportedly sold well on eBay last year. Plant scams, however, such as the one Jordan uncovered, are just one example of how AI is being used.
You’ll most likely have encountered AI chatbots such as ChatGPT, and heard how they can churn out essays in seconds, books in just minutes, and answer people’s queries about everything from flight times to the future of humanity. But AI is also being used as a force for good, for myriad purposes in myriad fields. Not least, believe it or not, in gardening. Soon enough, AI will be having a tangible impact not just on the horticulture industry, but even in domestic gardens, and in countless ways we can hardly imagine. The fact is, AI isn’t a future concern for our gardens any longer – it’s already here.
In the past two years, AI technology has exploded into the mainstream. Thousands of column inches and entire books have been dedicated to a technology that experts predict will be as transformative for humanity – nay, life on Earth – as the Industrial Revolution. And yet, many of us are uncertain as to what AI is, how it works and whether or not we should be concerned about its ubiquity. “AI is essentially a system that’s able to make decisions, learn about whether those decisions are good or bad, and adapt to make better decisions next time,” says Dr Daniel Hulme, CEO of award-winning AI company Satalia, who last year spoke on the subject to the Horticultural Trade Association (HTA).
“AI is really good at making things more efficient; it could remove the frictions of the horticulture industry. For example, an RHS scientist might take many months to crunch through test data, but with the assistance of AI they could do it in a fraction of the time. Does that just incrementally improve the industry, or does it unlock new opportunities?” Needless to say, for Daniel and his colleagues, it is very much the latter.On the domestic front, smart gardening technology that fuses AI and robotics in lawncare and irrigation is already available for the deeper pocketed gardeners among us. Watering can now be automated with AI-backed systems ranging in price from under £100 to more than £1,000. The technology that forms part of these systems can track the weather and adjust schedules to suit conditions, and is controllable by the owner, via smartphone, from anywhere. Install a few soil moisture sensors and it can bring data together in seconds and adjust irrigation accordingly, saving untold litres of water. But even these sophisticated garden helpers are relatively low-tech, compared to what AI is capable of.
“Generative AI, like ChatGPT, is really good at knowing things about the world,” says Daniel. “It can identify a plant, the health of it, its history and what you can do to improve it. The next iteration of AI will be able to act on your behalf via a connected, smart environment that’s making decisions for you. It will be able to improve the health of a plant by analysing the soil it sits in and be able to then change the To grow well, plants need a wide range of nutrients in various amounts, depending on the individual plant and its stage of growth. The three key plant nutrients usually derived from soil are nitrogen, phosphorus and potassium, while carbon, oxygen and hydrogen are absorbed from the air. Other vital soil nutrients include magnesium, calcium and sulphur. Most garden soils contain enough nutrients to promote healthy plant growth.
nutrients and watering plans to maximise effectiveness. This sort of AI could open up gardening to many more people who don’t have the knowledge or time to learn or maintain their space.”
These aren’t pie-in-the-sky concepts from a science-fiction story. At the RHS, in one of the most pioneering projects in its history, scientists are developing an AI capable of identifying plants and their most useful characteristics. This bespoke AI will inform the newly announced RHS Plants for Purpose project, an ambitious scheme with the University of Nottingham that will see as many of the UK’s cultivated plants as possible labelled with the environmental benefits they provide. You may be familiar with the RHS Plants for Pollinators badges that already grace many plant labels. Well, through Plants for Purpose, badges will be created for additional categories including drought tolerance, flood mitigation, pollution capture and many more besides. But first, a huge amount of research data must be collated and combed through. This could be an unimaginably time-intensive process, representing the combined work of botanists, physiologists and entomologists. Enter AI. With the ability to sift through 94,000 specimens in the RHS Herbarium and identify plant traits in a matter of minutes, AI is being used at the RHS today to help gardeners identify the right plants for the right places – with the right purpose – tomorrow.
As with so many changing trends and developments over the decades, to see what all of this might appear like in real terms, you need only look to the RHS Chelsea Flower Show where, later this year, these ideas are set to be explored in a first-of-its-kind Show Garden. Award-winning designer Tom Massey has partnered with architect Je Ahn of Studio Weave, and Microsoft and Accenture-founded AI specialists Avanade, to show how AI could be used to help the plight of urban trees, which are suffering globally with the challenges they face due to climate change, harsh city environments and lack of aftercare.
“On The Avanade Intelligent Garden, we’ll use sensors to monitor individual tree health and use an AI tool to feed back information in real time, painting an accurate picture of how that tree is faring in its environment,” says Tom. “We also plan to create digital twins of some of the trees, to predict how they might perform in say 50 or 100 years.” The digital trees will be accessible online, allowing engagement and interaction from visitors to the show and the general public.
When the garden was first announced, it was met with both intrigue and incredulity from a number of well-known garden designers and horticulturists, some of whom publicly questioned the need to bring AI into gardening, and the value of giving this platform to a technology that they fear could put jobs, including Tom’s own, at risk. It begs the question: Why on earth would he do it?
“What really appealed to me about integrating AI into a garden at RHS Chelsea was to explore how technology and sustainable landscape design could exist together, in a way that could be beneficial and helpful, particularly in the face of the climate emergency,” says Tom. “There are all sorts of ways I could think of AI being used negatively, but it can be used positively too, with human agency. People are understandably worried about job security and pay, and AI’s potential to devalue their skills or replace parts of their job. But in my view, humans have a unique connection to nature that cannot be replaced by artificial intelligence, and I do not believe that gardeners or garden designers are at immediate risk. A digital AI doesn’t have a physical body – it simply cannot replace the physical interaction and human intuition that we have with gardens and landscapes. That’s also not what we’re trying to do.”He continues: “What we’re aiming to create is a powerful analytical tool that aids custodians of urban trees, by giving them insights into things that aren’t necessarily visible to the human eye, such as flow of Sap is a fluid that circulates through a plant's vascular system, in a similar way to blood moving through our veins. Phloem sap carries the sugars produced in the leaves by photosynthesis down to roots and other storage organs, as well as carrying minerals and hormones. Xylem sap is watery and transports nutrients absorbed by the roots upwards to the rest of the plant. Some plants leak sap from wounds or pruning cuts, and this is known as bleeding.
sap up the trunk, or rate of growth. It would be impractical too for humans to constantly monitor trees, and it would also require a huge level of skill, knowledge and expertise to receive and interpret that information. Imagine a tool that could network all urban trees, creating a digital ‘wood wide web’ allowing trees to communicate with their custodians, whether that is a group of school children with a newly-planted tree in a playground, a group of residents looking after trees on their street, or a large-scale developer planting thousands of trees in urban developments.” He adds: “In those circumstances, having AI doing that monitoring, and alerting those people to problems, could be really powerful and could allow urban trees to live for many more years, providing all the benefits and ecosystem services that they bring.”
You may be relieved to hear that Tom and co-designer Je Ahn did not turn to AI to design the garden for them. But the idea that they might do so is not as out-there as one might assume. Already today there are a number of increasingly sophisticated AI-based apps available online that make garden design accessible for all. OK, so these are yet to reach a level of sophistication to rival top designers like Tom, but already they are offering everything from garden layout plans to pest management solutions, plant selection advice to landscape visualisation – all from as little as £10 per month.
As any artist or designer will tell you, of course, what a computer – crunching, compiling and extrapolating existing data sets – cannot do, is match the imagination, the creative DNA, of a human being. Or can it? Courtesy of US startup company Delphi, anyone, including garden designers, can now turn to AI to create a digital A clone is a genetically identical copy of a plant, produced by asexual or vegetative reproduction. Clones have the same genetic makeup as the parent plant.
clone of themselves. After being trained on a person’s existing notes, documents, emails, videos, books, voicemails and more, this clone can learn to think like them, make decisions like them, even sound like them. If a designer like Tom Massey or one of his RHS gold medal-winning colleagues were to use it, this digital proxy could offer customers advice on garden design in their style, be available 24 hours a day and even provide ongoing guidance once the human designer has left this mortal coil. That’s right. A garden designer could achieve immortality, albeit in digital form.
Tom however, is unconcerned for his livelihood. “People want that human interaction,” he says. “If you told a client that instead of a real live designer they’d be talking to an AI simulation of that person, I don’t think they’d have the same level of excitement or engagement. Building a garden is often quite an emotional process for people – a lot want and need the human interaction to guide them through the process, and like to stand in the physical space to imagine it together. I’d be sceptical about the degree of nuance an AI doppelgänger would give. That said, I could see how AI could really help to design a complex planting scheme that is resilient, adaptable, suited to a specific site now and in the future, using vast amounts of data to inform decisions that would be too much for a human designer to process. That’s something I’d be excited to explore.”
Taken at its worst, AI can seem to be a threat: an infinitely smart digital brain that can perform tasks quicker and arguably better than we can. Understandably, there are many, across every industry on the planet, who are wary of this great computerised workforce, standing in wait. However, there are many instances where AI is not replacing a human being, but performing a task that no person – or even team of people – could. And perhaps surprisingly, some of the most revolutionary applications of AI are being used not in labs or dingy basements, but out in the wild.
In what can only be described as the greatest lonely-hearts story in the history of tech, AI is currently being used to search for a mate for one of the rarest plants on Earth, Encephalartos woodii. This rare cycad is thought to be extinct in the wild, surviving only through A clone is a genetically identical copy of a plant, produced by asexual or vegetative reproduction. Clones have the same genetic makeup as the parent plant.
clones of a single male specimen in private and botanical collections. No female plant has ever been found. But thanks to drones, multispectral imaging and AI, areas of oNgoye Forest in South Africa, where the male tree was discovered in 1895, are now being scanned for a mate.
“I was captivated by the story of the male E. woodii,” says Dr Laura Cinti, the biologist leading the project at the University of Southampton. “From the air, cycads’ palm-like canopies stand out against the surrounding forest. Our AI models are trained to recognise the unique shapes of cycads in aerial imagery, allowing us to distinguish them from other vegetation, even in dense and complex environments, not accessible on foot. Finding a female E. woodii would be a significant breakthrough for conservation, and the methods we’re using open up possibilities for more efficient and accurate conservation efforts.”Closer to home, AI is being deployed to track The variety of living organisms (plants, animals, fungi and microorganisms) in a particular environment. Boosting the biodiversity of your garden has many benefits, including supporting wildlife, improving soil health and reducing the likelihood of pest and disease problems.
biodiversity across multiple ecosystems. The Bat Conservation Trust, for example, is using it to help identify bat species in Britain by running their distinct calls through a computer, which is traditionally a labour-intensive, time-consuming process. Meanwhile, a wealth of biodiversity and environmental data collected from 25 sensors within the Natural History Museum’s five-acre garden is also being analysed using AI to help support urban nature recovery. All of which sounds really rather wholesome.
There is, however, a catch. Ironically, one that could have a devastating impact on the very environment AI is being used to protect. Namely, the eye-watering amount of energy and water required to power the growing number of AI data centres as they crunch through increasingly vast amounts of information. Take, for example, a seemingly simple task, like having a back-and-forth with ChatGPT for a typical 10 to 50 responses. To power that interaction alone requires half a litre of water, used to cool banks of servers that process information, according to researchers from the University of Colorado and the University of Texas. It doesn’t sound like much until you discover the chatbot has more than 180 million users. Just one ChatGPT query uses nearly 10 times as much electricity as a Google search, according to the International Energy Agency. As a result, researchers at Goldman Sachs estimate the amount of overall power data centres consume worldwide will rise from 1–2 per cent to 3–4 per cent by the end of the decade, while associated carbon emissions could more than double by 2030.
To mitigate this impact, big tech companies such as Google and Amazon are purchasing vast amounts of renewable energy to power their datacentres. Late last year, Google announced it was turning to nuclear energy to keep up with demand, building several small nuclear reactors to fuel the needs of their AI-powering processing units. “Energy is a big question,” acknowledges Daniel Hulme, though one for which a solution may already be on the table. “Emerging technologies called neuromorphic computing operate tens, hundreds, maybe even thousands of times more efficiently than generative AIs like ChatGPT. As the technology gets better, it will use fewer resources.”
Annette Giardina, Chief Innovation and Sustainability Officer at Avanade, sponsor of Tom Massey’s RHS Chelsea garden, says: “Sustainability is one of the challenges we face every day because many of the activities we undertake to make the world more sustainable consume the very energy or water we’re trying to save. That’s why at Avanade we talk about responsible AI and, by that, I mean AI which adds benefit, not AI for AI’s sake. It’s not about taking people’s jobs; it’s not about taking the human element out of this – you can’t, and we shouldn’t. It’s about making it easier for people to access data that has meaning; it’s about giving people access to data and insights they wouldn’t necessarily be able to determine without the equipment.”And when it comes to energy and water requirements, not all AI is created equal. “When you ask generative AI – such as ChatGPT – a question, it consumes vast amounts of the internet looking for the answer,” says Annette. “This is what’s called an untrained large language model, and it consumes tremendous amounts of power. For The Avanade ‘Intelligent’ Garden at RHS Chelsea, we’re using traditional approaches that have been around for years. These AI models will be run against what’s known as a small language model, which is highly trained on specific data, so uses less power to bring back information. As it gets more data, it learns and becomes smarter and more efficient.”
The sustainability question is certainly giving pause to companies and governments around the world. But what about the other potential casualty in this emerging world? What about us?
From 2001: A Space Odyssey to I, Robot via The Terminator and The Matrix films, the idea of a sentient computer posing a threat to humankind has been part of our cultural consciousness for as long as computers have been part of our lives. Fine, so those phony begonias hardly herald the start of a cyber conquest, but they do point to something somewhat insidious: it’s one thing to rely on AI to perform otherwise time-intensive or complex tasks, but what if the answers we’re given, the solutions we’re presented with, turn out to be unreliable? Can AI be trusted?
“Yes, the issue around scamming and deep fakes is a very real issue that governments need to regulate,” says Daniel. “We also need to educate people to think more critically and there need to be mechanisms in place to authenticate whether something is real or not – such as, for example, watermarking. It might be that you create an expert plant ‘brain’ that can alert you when an image has been AI-generated.” Experts – including Daniel – believe that, far from being replaced by AI, people still have an important role to play. “We’ll start to realise that having a human in the loop is incredibly valuable.”
The RHS Horticultural Advice Team’s begonia sleuth Jordan agrees. “Any AI is only as good as the information being fed into it,” he says. “People are using apps to identify plants instantly; they no longer have to wait for me to get back to them in a day. But whether they’re getting a correct response or not is another thing. So, the information should be treated as guidance, rather than as an absolute.” RHS experts are constantly checking, correcting and adding to the RHS’ AI-powered ChatBotanist service. This chat function draws on a database of more than 200 years of expert knowledge to help answer gardening-related queries. As well as offering tips for how to care for the plants you have, the RHS Grow app also enables you to identify plants using your smartphone’s camera – an increasingly useful and accurate tool with AI at its core. ChatBotanist is also a more sustainable AI tool for horticulture than ChatGPT, as it only trawls relevant pages rather than the entire internet. Whether we feel comfortable with it or not, there’s no doubt that from plant identification and soil analysis, to garden design, planting lists, everyday maintenance tasks, lab-based botany, field work and more, AI is becoming ever more enmeshed in our horticultural lives. But what does the AI itself make of all this? It seems only fair to ask. So we did. Specifically, how did AI think it might change gardening for the better, and for the worse? And as anyone who has dabbled with ChatGPT can no doubt attest, the lucidity of the answer it gave was impressive, ingenious and disconcerting in equal measure. It said the benefits would certainly include “smarter, more efficient and eco-friendly practices,” but it also pointed towards a loss of more traditional skills and an increasing reliance on technology rather than our own knowledge, built up during years of experience. But it’s concluding statement was this: “Balancing AI’s benefits with thoughtful, mindful gardening practices,” will be key.
It’s a notion that would likely resonate with most gardeners and is one at the core of new Aardman film, Wallace and Gromit: Murder Most Fowl, now nominated for an Oscar in the Best Animated Feature category. In the film, available on BBC iPlayer, Wallace builds Norbot, a ‘smart’ gnome designed to help Gromit in the garden. This latest thing in ‘cutting hedge technology’ is very good at mowing and cutting back and does a mean line in topiary; it even does complete garden makeovers that may or may not be to everyone’s taste.
But, says Wallace and Gromit creator Nick Park, Gromit does things because he loves the activity of actual gardening and the peace it brings. “Wallace’s inventions aren’t bad, they’re just doing unnecessary things and, in this movie, it’s very much about how tech can come between people,” Nick says. “Gromit’s life becoming so automated kind of drives a wedge between him and Wallace. I feel as fixated with tech as the next person, and obviously, AI is making incredibly good advances in many ways,” he adds. “But I think generally, we’re questioning how much it’s just a big distraction for many of us. There’s nothing better than getting on your knees and doing a bit of weeding; it seems good for the soul.”
Just because a computer might contain vast amounts of information about plants and gardening, that doesn’t mean that it will stop us from doing the things we love. “Let’s assume that we do get to the point where an AI is able to garden better than a human, because it has access to all the information and is able to make decisions much more quickly,” says Daniel. “But at what point does it begin to take away the pleasure we get from gardening? You know, AI can play chess better than humans, but there are more people playing chess now than ever before. Right now, AI is an expert in your pocket. Eventually, it can become a decision maker and act in the world. But you’ll still be able to control how much autonomy you want to give the technology. When I talk to people about how AI can ultimately free them up, I’ll often ask what they would do with that spare time. And do you know what the most common response I hear is? ‘I’d spend more time in the garden’.”