Brief Thought Experiment about Machine Vision Robotic Weed Control

I think about novel ways to control weeds in farm fields a lot. It is one of those things that I will daydream about when I think about the future of farming. If you are a futurist imagining what agriculture is going to look like 10,20,50 years from now you will probably envision teams of small robots moving autonomously through the fields with complex machine vision systems identifying weeds and then killing them with some targeted means.

So here is my thought experiment.

1- everyone knows that over the past 3 decades various weeds have evolved to be resistant to glyphosate. We have basically caused the genetic makeup of the plant to change over time.

2- does it follow that with weeding systems based on machine vision would over time become less effective because the weeds would change the way they look- phenotype? Eventually looking more and more like the cash crop in the field, so as to make it impossible for the machine vision to distinguish between them at different stages in their growth cycles?

3-would 2 happen faster than 1, slower than 1, or not at all

Thanks for reading