Introduction: Forget the scarecrow, the future of agriculture lies in the hands of intelligent machines.
Humanity has posed a challenge to itself – while the population explosion continues, the Earth will not become bigger. When the Earth welcomes its 10 billionth person in 2050, it will still have to use the same amount of land to feed people, and with the continuous reduction of water sources caused by global warming, whether humans can eat enough in the future will be a big problem.
Perhaps there is a divine will in the dark, and artificial intelligence will make a timely appearance at this time. True intelligent robots and machine learning algorithms will bring about a green revolution, a revolution that will feed humanity with the limited land on this planet. You can imagine satellites that automatically detect signs of drought, tractors that can automatically identify and kill sick plants, and AI phones that can automatically notify farmers of crop diseases.
Forget the scarecrow, the future of agriculture lies in the hands of intelligent machines.
Digital Agriculture Expert
Deep learning is a powerful algorithm where programmers no longer explicitly tell the computer the target, but train it to recognize several patterns. You can provide the computer with photos of plant leaves indicating health and disease. From these photos, it can learn how to distinguish between healthy and diseased leaves, and then determine whether other leaves are healthy.
This is exactly what biologist David Hughes and epidemiologist Marcel Salath é did with 14 crops infected with 26 different diseases. They provided a computer with over 50000 photos, and through self-learning, the computer achieved an accuracy rate of 99.35% in judging whether plants were healthy from the photos.
However, these photos have been artificially processed and have a uniform brightness and background, making computer judgment much easier. The computer’s accuracy in identifying a photo of a sick plant found online is only 30% to 40%.
Although the result is not good, it is not bad either. Hughes and Selasser hope to apply this AI to their mobile application Plant Village. At present, farmers around the world can take photos of their sick crops and upload them to the Plant Village forum to determine the disease they are infected with. However, currently it is human experts who are helping to make the judgment. To enhance the intelligence of the application, it will learn by continuously providing photos of sick plants. We will provide it with a variety of photos, such as different photography techniques, seasons, locations, or other attributes, “said Seraser.” This algorithm will learn to distinguish on its own
Their goal is not just to eliminate infected plants, as plants can get sick for various reasons. Most of the diseases that affect plant growth are physiological stresses, such as calcium or magnesium deficiency, high soil salinity, or excessively hot weather, “Hughes said.” Farmers often believe that these are bacterial or fungal diseases. Misdiagnosis can lead to farmers wasting money and time on pesticides. In the future, their AI will help farmers quickly and accurately identify the root cause of problems.
However, the follow-up work after identifying the problem needs to be done by someone. Although an application can help identify problems, only human experts can find solutions based on specific climate, soil, and seasons. The United Nations Food and Agriculture Organization considers this technology to be an “effective tool” for crop management, but expert opinions are the only conclusive evidence. Fazil Dusunceli, a plant pathologist at the Food and Agriculture Organization of the United Nations, said that such electronic judgment results are good, but the “final management of agricultural pests” needs to be decided together with local experts.
Agricultural machinery trainer
When developing countries urgently need agricultural knowledge, developed countries are being overwhelmed by pesticides. In the United States, 310 million pounds of pesticides are dumped on corn, soybeans, and cotton each year alone. This is a way of “praying after spraying”, which is more like carpet bombing compared to sniper style pest control.
A company called Blue River Technology is likely to have found a solution to this problem, at least solving the lettuce issue. Its Letter Bot looks like a regular tractor, but in fact it is a deep learning based machine.
Blue River claims that the Lettuce Bot can take photos of vegetable seedlings at a speed of 5000 plants per minute while driving through a field, and use algorithms and machine vision to determine whether each plant is lettuce or weed. If you find its speed unimaginable, “this is still far from the limits of machine learning and machine vision,” said Jeremy Howard, founder of deep learning algorithm company Enlitic. He also stated that the image processing chip can recognize a photo in just 0.02 seconds.
The accuracy of the Lettese Bot can reach 1/4 inch, which means it can accurately identify each weed and spray them with herbicides during operation. If it determines that a lettuce seedling is not growing healthily, it will also spray it (usually, farmers plant seedlings 5 times higher than expected, so they can sacrifice many seedlings). If two vegetable seedlings are found to grow too close (unhealthy), this machine will not mistake them for one large plant, but will spray them separately.
After reading this description and comparing it with the current method of spraying the entire farmland, is it very cost-effective?
It’s like saying, if a few people in San Francisco are infected with the virus, your only solution is to give each person an antibiotic, “said Ben Chostner of Blue River.” People will be cured, but the cost of cure is too expensive, and the potential of antibiotics has not been fully exploited
Through the Lettese Bot, Costner stated that farmers can reduce pesticide use by 90%. At present, this machine has been used in the market: fields supplying 10% of lettuce in the United States are using Blue River’s products.
Blue River Technology’s Lettese Bot can take photos and process vegetable seedlings at a speed of 5000 plants per minute
The reason why the Lettese Bot is powerful is that it amplifies the machine’s inherent advantage of precision through machine learning. Robots cannot run or manipulate objects like humans, but they are meticulous and can become perfect agricultural snipers.
The AI in the sky
On an orbit 400 miles above our heads, NASA’s Landsat satellite utilizes electromagnetic waves beyond the visible spectrum to provide people with incredible Earth surface exploration data. These data are like heavenly books to humans, but for machine learning algorithms, they are not at all difficult.
And this is an invaluable treasure for agricultural monitoring. Especially in developing countries, when governments and banks severely lack data to help them decide which farmers need loans or assistance, these data will have a great impact. Take a drought in India as an example, the degree of water shortage varies between regions, and even within the same region, some farmers have relatively more means to obtain water.
So, a startup company called Harvesting is using machine learning to analyze massive satellite data, with the intention of helping government departments and banks allocate resources more effectively. Our goal is to use this technology to classify farmers and villages so that banks or governments can distribute money to the right people, “said Ruchit Garg, CEO of Harvesting.” A human analyst can handle 10 to 15 different variables, while machine learning algorithms can handle over 2000 variables, which is a difference of two orders of magnitude.
The climate change brought about by global warming means that governments are facing the challenge of how to allocate resources reasonably. Traditional agriculture in India has always been very regular. I learned how to farm and distinguish seasons from my father and grandfather, “Gage said,” but due to climate change, their experience is no longer useful
We are now facing a new world order. Farmers can choose to be eliminated or choose to learn more intelligent agricultural methods – more data, more AI, more pesticide spraying robots.