Ending crop pests, disease with neural evolution
If you raise crops like corn, wheat or soybeans anywhere in the world, you probably know well the growing threat that herbicide-resistant weeds pose to future productivity. The ag chemical industry has gotten beyond proficient in knocking down weeds, but Mother Nature has caught on, and weeds resistant to one chemical in particular — glyphosate — are creating serious concerns for farmers.
Glyphosate — often better known by its common trade name, Roundup — has been a staple weed control chemical in crop production since its commercial rollout in 1974. But, a combination of factors like misuse and gradual plant adaptation, has shown the chemical is starting to lose its efficacy in some cropping systems. And, at a time when crop production is in high demand but market prices have been lackluster at best, a lot of farmers simply can’t afford to have a key piece of their agronomic puzzle cease to fit where it needs to go.
But, weed control may not always be about whether your latest Roundup application will knock down the weeds in your fields. Recently, a video game designer and engineer created an artificial intelligence (AI) system that essentially learned how to play Super Mario Brothers. After playing for a few hours, the deep-learning MarI/O used neural evolution to in effect learn how to play the game. It’s a simple example that shows deep learning can have a profound effect on the speed and effectiveness of artificial intelligence.
It’s just a video game and it’s just a simple AI system, so the learning can’t be that “deep,” can it? Think about it, though. In a video game like that, the player is forced to make decisions. To move forward and in reverse, jump, shoot fireballs, and avoid “bad guys” and obstacles. Versus a simple on/off command or something similar, this array of choices and potential actions means MarI/O had to learn a lot more than a few simple commands in its 24 straight hours of gameplay. In the context of artificial intelligence development, it’s like going from a Ford Model T to the International Space Station in just a couple years.
There’s now a way to apply that kind of deep learning — on an exponentially more complex scale — to living things. Deep Genomics applies AI technology to the genomes of living things. Applying neural evolution and deep learning to the human genome, for example, can ultimately yield “on-demand” medications that specifically target the genetics of a disease, for example, in the specific context of a living thing’s genetic makeup. Think a vaccine that, when the potential building blocks for a disease are detected in the genetic makeup of a living thing, automatically assembles the treatment mechanism within the body, preventing the disease from ever becoming symptomatic, or killing it altogether. Or, one step further: Think a vaccine that predictively treats a disease before it even materializes based on genetic markers the AI has “learned” through neural evolution.
Now, back to glyphosate resistance and crop weed and pest control. One day, an AI mechanism — delivered by a spray or other agronomic in-field system or bred into the seed itself — can make chemicals like Roundup obsolete. A crop that can either intrinsically or with a single crop input application both detect the disease and eradicate it before you, as the farmer, even have to take a single action. It could eventually yield a crop that doesn’t need a single crop protection application throughout its life.
It will take time, and it will be costly to the producer, at least up-front. But ultimately, once the curve of neural evolution is established so it keeps up with the adaptation of the plants it targets, this type of technology can free up the farmer to direct his or her energy and assets more specifically at production. Defensive corn hybrids will become a thing of the past and the future will be one in which all new seed technology will have yield as the single goal since artificial intelligence will take care of all the things for which you need Roundup — and a stable of agrochemicals — today.