Worth noting that it's often worth using this tech to kill the crop too...
Most plants must be planted with the right spacing to maximize yield. Plant too close and the plants will overcrowd eachother and block eachothers light and yield will decrease.
However, not every plant gets the best start in life. Some may grow slowly or not at all, have roots hit a rock, or have a host of other issues.
By deliberately planting the seeds 'too close', but then deliberately killing the ones who don't get off to a great start, you can actually get an even higher crop yield.
As soon as lasers on farm machinery are common, this can be done at scale.
Toxic herbicides? Do they mean Roundup? It's rendered inert by contact with water. None remains in the soil, as I understand it.
I believe they're using hyperbole to market their robot. But with field applications costing on the order of $10 per acre, it has to be a very cheap robot to run, to compete with that.
> It's rendered inert by contact with water. None remains in the soil, as I understand it.
There is a growing body of evidence that the main ingredient in roundup (glyphosate) is pretty harmless, but that the other non-active ingredients have a pretty substantial environmental and health risk, especially at high doses.
I don't normally ask for references but for this one I would prefer to see them as I have run across plenty of studies on nih.gov on endocrine disruption from glyphosates and articles about lobbyists and roundup. One can find some of these studies with a quick google search. [1]
The Small Robot Company in the UK, whose "Head of Intelligence" worked at DeepMind for 7 years, is another such company working on this (and other) problem(s).
Their solution has three robots named Tom, Dick, and Harry. Tom maps the field, assessing soil conditions and weeds. Dick zaps weeds. And Harry is the haulier of nutrients for crops.
> Weeds are getting harder to kill as herbicide-resistant varieties survive
Surely it won't take long for camera-resistant weeds to evolve? They will learn to avoid detection via camoflage and mimicry. Still, I guess it's a better problem to have than a chemical arms race.
That's already happened - Rye is an example of Vavilovian mimicry[1][2], a crop that evolved to be hard for humans to distinguish from weeds. Thankfully, the selection pressure is also to be tasty and useful, as burning a field to prevent it's weeds from spreading remains a viable solution.
I bet we can update the laser targeting firmware faster than natural selection can update weeds. I would also guess we can take some inspiration from hummingbirds and use the plant's IR signature as part of the targeting selection criteria, too
Most plants must be planted with the right spacing to maximize yield. Plant too close and the plants will overcrowd eachother and block eachothers light and yield will decrease.
However, not every plant gets the best start in life. Some may grow slowly or not at all, have roots hit a rock, or have a host of other issues.
By deliberately planting the seeds 'too close', but then deliberately killing the ones who don't get off to a great start, you can actually get an even higher crop yield.
As soon as lasers on farm machinery are common, this can be done at scale.