Our planet has got itself a serious problem. There are predictions that by 2050 there will be nearly 10 billion people in the world and the planet must support them. Rising concerns about global warming and water shortage has made scientists think of how to avoid the disastrous consequences of not cutting the consumption of goods.
There have been many books written, and films made, on how AI has destroyed the human race in the future. In this case, though, we may be seeing it as a way to save the planet. Imagine tractors that detect crops and remove dead ones with AI-powered apps that can tell exactly which disease has killed it.
Deep learning is a powerful technique that enables machines to learn how to recognize certain patterns, instead of being told what to specifically look for. The developers can feed millions of images of unhealthy plant leaves and program the algorithm to recognize them. From this, the machine will learn how the diseased leaves look and determine the health of the new leaves on its own. Biologist David Hughes and epidemiologist Marcel Salathé, who developed the PlantVillage app, experimented with this method on 14 crops infected by 26 diseases. They fed a computer more than 50,000 images, and when the machine memorized them, the program could correctly identify 99.35% of the newly uploaded images.
The project is still at a testing stage and accuracy is only around 30-40%. Hughes and Salathé haven’t given up though and hope this AI-method will power the app, that allows farmers to upload images of their sick plants for experts to diagnose. This is because farmers often misdiagnose their plants, thinking it is an infection when in fact, the plants could simply be lacking essential vitamins, heat or water. By attempting to heal the sick ones with the unnecessary use of chemicals, farmers simply create another problem.
While developing countries are hungry for agricultural knowledge, the developed world is using millions of tons of pesticides and herbicides where it could have been avoided. A company called Blue River Technology came up with the solution and introduced its LettuceBot which looks like a typical tractor but in reality, is a machine-learning powered equipment. The bot can roll through a field and photograph up to 5,000 young plants every minute, using algorithms to identify plants as sprouts, weeds or lettuce. With an accuracy of a quarter inch, the bot sprays each weed on the fly. If it spots a lettuce that is not growing right, it will spray it too, to help create a more uniform and healthy crop. Considering that the only alternative is the traditional approach of spraying herbicides on everything, the method could be revolutionary. According to Blue River Technology representative Ben Chostner, such a method can reduce the use of chemicals to ‘up to 90%.’
This kind of innovation within the agricultural sector can help farmers to avoid negative consequences, improve the quality of the crops produced and increase the yield of each farmer. Although small increases for individual farmers may not do much alone, when added up they can create significant boosts in crop yield.