What am I worried about here is that ML would eliminate some hidden factors allowing survival of biosystems in its goal to increase yields. Nowadays there might be some inefficiency with side effects allowing survival of required species; with more efficient models wrt profit those effects might disappear, causing hard-to-explain issues. An example from the past was when mountainous national parks banned certain animals from feeding on grass within park boundaries, leading to disappearance of rare flowers that needed trimmed grass surroundings to survive.
The headline is amusingly overly modest. Statistics and science have changed agricultural practices in extreme ways many times. As have simple economics. Machine learning doing the same shouldn't be a surprise, since it should just be seen as a continuation of those practices.
This article is a very light without any concrete examples. (edited typos)
This field is lot more about hardware and software automation, where AI can and already has a big role.
The ML part, frankly, is nothing new. With the significant exception that more people are doing it, including groups of people who would traditionally not do it, such as farmers and technologists. It used to be mostly academics and large companies with research teams.
I wanna see "modern agricultural science" pushing the boundaries of ecology, modelling complex, diverse ecosystems with a goal of cultivating them in economically viable ways!
I don't think "machine learning" is the right tool for the job though.
