Is AI Coming for Your Grow?

Artificial intelligence, as described by Musk, is among the greatest threats to humanity — conjuring images of robot rule and Terminator-style hellscapes in the minds of those who hang on his every word. Others however, believe his warnings to be hyperbolic, and that the dangers he outlines will continue to be the province of science fiction for the foreseeable future. Despite the fact that I too work with a data science company focused on the development of AI-based products, I tend to side with Musk.

However, what the audience for these cautions often lack, is some context around grand AI ambitions that are as yet unrealized, versus those being developed by building upon existing frameworks which result in the innovative tools we are all beginning to use every day.

Artificial General Intelligence, or AGI, has been the decades long goal of AI researchers who are striving to replicate the human brain in silicon — or at least create one capable of being indistinguishable from a biological counterpart in terms of its higher functions, such as learning independently how to perform tasks, and fluent communication. And while today AI systems may indeed be able to pass the famous ‘Turing Test’, which measures whether a bot can successfully pass itself off as a human in text-based dialogue, there is still a very long way to go before these systems could be reasonably described as actually approaching general human cognitive capability.

Certainly the development of an AGI would have profound implications for society, but that it could rapidly lead to Artificial Super Intelligence — or actual machine consciousness — is the fear that keeps researchers at night… and dystopian science fiction authors in business. This stage of AI development in which machine intelligence actually eclipses that of humanity presents a nightmarish Pandora’s Box of uncertain outcomes. Will it evolve beyond human understanding and control? Will it dispense with morals and values it deems archaic or irrelevant to its interests, and set itself in opposition to its creators? Will humanity become essentially enslaved due to its dependency on machines which have invaded every aspect of our existence? The possibilities are endless, and the dangers equally so.

On the other hand, Artificial Narrow Intelligence is firmly rooted in today’s reality. This technology is designed to perform specific tasks with a high level of expertise, but has limited ability to generalize or adapt to new situations beyond that specific task or domain. Personal voice assistants such as Siri, face detection in your camera, predictive text, recommendation algorithms on Netflix or Amazon etc., are all powered by this level of AI. These machine learning systems can be very effective in performing functions which provide significant benefits to society, such as enhancing transportation safety, improving medical diagnoses, and automating tedious or dangerous tasks.

The company I am with, Fermata, was born of a desire to address these last two. Coming from a mathematics and biotech background, our CEO, Valeria Kogan PhD, had an epiphany that the same computer vision based AI techniques she was employing in the medical field, could also be leveraged in the agricultural sector in support of the early detection of pests & disease.

Growers forfeit billions of dollars around the globe every year through crop loss due to these issues. Further, the time, energy, and other crop inputs associated with the loss are also wasted, which is clearly a negative from both a sustainability and profit perspective. As the ability to diagnose a pathogen or discover an infestation at its earliest stages allows for decreased pesticide use, improves quality, and dramatically reduces crop loss, our development of Croptimus, which accomplishes this task, seemed to us a natural marriage of AI and horticulture.

Beyond this, scouting is a bit of a soul-crushing endeavour at times — defined by the monotony of manually searching for tiny signs of problems on plant after plant. We felt if we could help IPM teams focus their energy more on mitigation and less on discovery, the grow would benefit, and workers would be happier. AI, for its own part, never gets tired and is actually one of the few products that improves over time. So it made total sense to us that this tedious and exhausting labour would be better suited to automation, freeing up workers to attend to more high-value tasks such as mitigation. Why spend time finding problems when you could be fixing them instead?

Of course, there are a myriad of other companies developing machine learning systems for horticulture which aim to solve a wide variety of agricultural problems. AI-powered yield prediction for example, enables growers to ensure they are going to be able to live up their contractual obligations and take steps to mitigate shortfalls or surpluses in advance, while also maintaining consistent results across cycles. These products may utilize multiple data sources — such as those to be found in environmental control and irrigation systems — to measure conditions in the greenhouse and compare these with yields over several years to establish relationships between them. Companies such as Canada’s Ecoation on the other hand use AI-powered computer vision systems mounted on robots which gather data about the crop to provide accurate forecasts, enabling growers to ensure they are meeting demand and organizing their labour efficiently at harvest.

Autonomous crop steering has also been a goal of nascent AI companies and well-established industry players alike. Dutch giant Priva for example, has been developing its Plantonomy platform for over a decade. This system automatically determines what plants actually need by way of following their natural biorhythms and then subsequently steering transpiration to achieve an optimal balance. It further takes over 10,000 adjustable parameters within their climate control system and reduces this number to 6, dramatically improving the both the learning curve and the time needed to be spent in front of the climate computer — a clear win for growers.

Further, breeders and crop researchers are now employing AI algorithms to analyze plant genomes and other data sources to identify traits that are desirable for specific crops. This information can be used to develop new varieties of plants that are more resistant to disease, have higher yields, or are better suited to distinct growing conditions. While developing new crop varieties traditionally involves years of studying plants’ performance in the field, AI-powered genomic selection can predict these outcomes before they actually grow, saving enormous amounts of time and money. As climate change inevitably alters the outdoor growing environment, these advancements are a key factor in protecting our food supply as well as medicinal crops.

AI also is being utilized in support of fully autonomous grows which depend on both automatic environmental control and robotic systems which attend to trimming, harvest, IPM, and packaging. While these systems are relatively costly at present, and some aspects of the operation still require human input, ultimately this will become an ideal solution for the production of produce within city centres as the technology improves and economies of scale serve to reduce associated cost. Iron Ox, an American AI and robotics firm, is an excellent example of the application of data science and mechanical wizardry to the problems faced by the industry in terms of sustainability and the declining numbers of skilled workers in the horticulture sector.

Given 1.19 gigatons of greenhouse gasses are released annually in the production of wasted produce alone, any technology which can mitigate these losses is both good for the bottom line as well as the environment — and right now, that’s AI. At Fermata, we strive to make growers lives easier and the companies they work for more profitable through the application of data science. That early pest and disease detection also facilitates a substantial reduction in pesticide use and crop loss is key to our vision for helping to create a more sustainable industry in which we all can be responsible stewards of the environment.

While today’s ChatGPT and DALL-E2 have dazzled users with their ability to generate absolutely convincing textual content and imagery, over the next few years we’ll all be shocked at how rapid the advancement of this technology becomes and the new tools it empowers. So, is AI coming for your grow? You bet it is — and that’s definitely a good thing.

Originally published in 'Maximum Yield' Magazine