Why I'm Uncharacteristically Excited for Laser Weeders
Arm AIs with lasers, it will work great actually
I see a lot of shiny tech promised to farmers: it’s unveiled with excitement and trumpets, it’s piloted with some beta test farms, and then it’s quietly ‘put out to pasture’. The “it seemed like a good idea” graveyard has a lot of really interesting ideas. More often than not, when somebody shares an ag tech startup plan with me, I nod and acknowledge what’s cool or promising about their idea, but I have to finish the sentence with “but this won’t scale the way you want it to scale.” Sometimes I add “margins in ag are too thin to let you make the return on investment you’re hoping for.” Often, I say “the market is too fragmented.”
But laser weeders—I’m here for them. Laser weeders are going to succeed, and they’re going to meaningfully improve agriculture in so doing. Here’s why.
Solves a Problem That Hasn’t Been Solved Before
Speaking from a perennial specialty crop perspective: perennial weeds are a thorn in a farmer’s side. Herbicides can be selective, i.e., they only kill grasses or only kill broadleaf plants. But when you have a weed that is very similar to the crop you want to sell, there is often not an herbicide that can kill the weed without killing its valuable cousin. In annual crops, which are planted anew each year, a farmer can use tillage or a broad-spectrum herbicide before planting the desired seeds, in order to start with a clean slate. In perennial crops (pecans, wine grapes), a new planting takes 5+ years to establish before producing a harvest. A farmer will only make a new planting to add a new variety, so the perennial farmer might only get a clean slate once every 80 years. For the perennial farmer, the control of weeds that are cousins of the desired crop is basically unsolved. The only uneconomical answer is to pull these weeds by hand. Given labor costs, there is no economical answer.1
Until: the laser weeder! Even plants which are similar enough to be vulnerable to the same herbicides have some differences. Height, color, emergence time, lack of seeds. The machine vision capabilities of a laser weeder can identify the unwanted weed, and ZAP it. The weed burns up and dies.
The cost of a laser weeder is not currently economical for many farmers, but the cost of laser weeders is coming down, while the cost of labor will only increase.
Solves a Problem EVERYBODY Has
A major problem with tech companies attempting to enter agricultural markets is that they often design products (both hard and soft) that reach huge markets without needing customization. While ag may look monolithic to outsiders, each crop mix and each geography create nuances that make scaling really difficult.
Farmers are a tricky market—products for them don’t exist off the shelf. So farmers improvise and garage-engineer their own solutions. They are used to working with tools that might be quirky, but were often tailor-made with exactly their problem in mind. Consequently if a company identifies a general need and offers a general solution, farmers may chafe at the lack of specificity. Sometimes the lack of specificity actually makes the product unusable in regions that weren’t part of the beta test.
But laser weeders don’t face this problem. Yes, they will need different track or carriage systems for some crop mixes. And of course they’ll need appropriate ‘weed vs crop’ training data. But those customizations can be handled at low cost without jeopardizing the entire product.
Crucially, everyone in ag deals with weeds that they can’t control with herbicides. Specialty crop farmers don’t get access to many herbicides. Organic farmers have to kill weeds by breaking up the soil (tillage), and this leads to soil erosion with the next rainfall. Conventional row-crop farmers might have herbicides, but they often face weeds that have evolved herbicide resistance2. Perennial farmers have weeds that will never go away.
Finally, tech has produced a product that every farmer will want to own, regardless of crop or geography.
Finds Extra Hours in a Day
A common greeting in farm country is: “How’re you doing?” “Not bad, just need more hours in a day” . Laser weeders use active sensing—emitting visible light or NIR so their onboard machine vision cameras can “see” weeds regardless of ambient lighting conditions. They can work all day and all night.
It’s likely at current costs, that laser weeders will be timeshared among nearby farms. Until costs come down, there are many more acres that want to be laser weeded, than there are laser weeders—so every hour of productivity when the ground’s not frozen will help get laser weeders on more acres, amortizing the costs.
Frees Up Humans For Higher Productivity Work
My first job, as soon as I could walk an acre without getting distracted, was “I’ll give you a nickel for every maple sapling—here’s a pillowcase to fill, get out there.”
I can say firsthand: hand weeding is not a good use of human talent. It is punishing to the body. It is uninteresting for the intellect. There is no artisanal benefit that humans can confer to this task. Hand weeding is precisely the kind of work that we should be entrusting to robots. Human hours and human lives are precious, and they should be devoted to higher callings than weeding.
Tech That Enables Laser Weeders
So many technologies have had to mature to bring laser weeders into the realm of the possible. I’m not an expert in any of these realms, but I do want to give a few shoutouts to the giants on whose shoulders we’re standing.
Active sensing and machine vision. The low cost, high resolution cameras that enable them.
AI-powered image classification for identifying what’s friend and what’s foe. “Let a robot see a weed and kill the weed” has been the subject of lots of pilot projects over the years. The clunkiest early attempts had separate ‘notice a weed’ and ‘kill the weed’ functions, and required a stop at a computer to give the kill-bot the information collected by the notice-bot. As machine vision improved, and using the heroic half-measure of fuzzy logic for faster image categorization, the notice-bot and kill-bot were merged. Fuzzy logic was a pre-AI attempt to convince computers to make human-like judgements, and it worked well enough for proof-of-concept. (Training classification for a new crop was fodder for an entire PhD dissertation.) But AI image classification works well enough and fast enough for machine vision to scale—both in model training, and when recognizing weeds in the field.
Battery life improvements. I know of some crops where the weight of the laser weeder will be the hurdle to overcome, so I hope battery life and battery weight see continued improvement!
Sub-inch RTK GPS accuracy, enabling precise mapping and easy wayfinding for robots.
Lasers themselves almost go without saying. The earliest anti-weed robots prototypes set their sights low: simply doing targeted, weed-only herbicide applications. I also saw prototypes attempting flamethrower—but fire is both less precise than lasers, and less lethal.
I’m so grateful for all of these domains. I’m glad they have attracted smart and dedicated talent, churning out improvements year after year, so that ag can be one of many industries to benefit.
Bring on the Future
It’s true that some mobility customization will be needed to bring laser weeders to every crop. And costs will need to come down before there will be a chicken in every pot and a laser weeder in every garage. But efforts toward both of these are already under way, and I expect them to succeed.
I am thrilled to be living on the cusp of an era when weeds will be killed without without tillage that causes soil erosion, without legally embattled herbicides, and without wasting human health and time. I hope that you will soon get an opportunity to see the stem of a weed that’s plagued you since childhood, engulfed in flame and reduced to ash—while the crop you love grows less than an inch away, unscathed. It’s beautiful, and it’s coming soon to a farm near you.
Even under the farmer’s common misconception “well my labor is free!”—hand weeding is not a fun job. I was fine with it when I was 6. But as you get taller, hand weeding hurts more. And as you get smarter, there are so many more useful things you could do with your time.
Barber, Tom; Bob Scott; Jason K. Norsworthy; Tommy Butts. Distribution and Management of Herbicide-Resistant Palmer Amaranth in Arkansas. FSA 2188, University of Arkansas System Division of Agriculture, Research & Extension, Sept. 2025. https://www.uaex.uada.edu/publications/PDF/FSA2188.pdf further information: https://www.uaex.uada.edu/farm-ranch/pest-management/weed/field-crops.aspx