What is and is not a weed that needs to be removed from the field is determined by the eyes of the farmer – and now, especially, by a new AI model from Carbon Robotics.

based in Seattle Carbon Roboticsmaker of LaserWeeder — a robotic fleet that uses lasers to kill weeds — announced a new AI model, the Large Plant Model (LPM), on Monday. This model can instantly identify plant species and allows farmers to target new weeds without having to retrain robots.

LPM was trained on more than 150 million photos and data points collected by the company’s machines on more than 100 farms in 15 countries where the robots currently operate. The model now powers Carbon AI, the AI ​​system that serves as the brains inside the company’s autonomous weed-killing robots.

Paul Mikesell, the founder and CEO of Carbon Robotics, told TechCrunch that before LPM, every time a new type of weed appeared in a field – or even the same type of weed in different soil or with a slightly different appearance – the company had to create new data labels to retrain its machines to recognize the plant.

This process took about 24 hours each time, Mikesell said. Now, LPM can learn a new weed instantly, even if it has never been seen before.

“The farmer can go live in real time and say, ‘Hey, this is a new weed. I want you to kill it,’ and that’s something that’s never been done before,” Mikesell said. “There is no new labeling or retraining because the Large Plant Model understands, at a deeper level, what it is looking at and the type of plant.”

Mikesell said the company, which was founded in 2018, began developing this model shortly after it began shipping the first machines in 2022. Mikesell has years of experience building these types of neural networks from previous roles at Uber and working on Meta’s Oculus virtual reality headset.

This new model will reach the company’s current system through a software update. From there, farmers can tell the machine what to kill and what to protect by selecting photos collected by the machine in the robot’s user interface.

Carbon Robotics raises more than $185 million in venture capital from backers including Nvidia NVentures, Bond, and Anthos Capital, among others. Now, the company will look to continue refining the model as the machines continue to feed new LPM data.

“We have over 150 million labeled plants now in our training set,” Mikesell said. “We have enough data now that we need to be able to look at any picture and decide what kind of plant it is, what species it is, what it’s related to, what its structure is, without having seen that particular plant before, because we have a lot of data going into the neural net.