Carbon Robotics launches AI model to identify weeds in real time

Seattle-based Carbon Robotics announced a new AI model called the Large Plant Model (LPM) on Monday. The LPM allows farmers to target new weeds without needing to retrain the company’s autonomous weed-killing robots. Previously, retraining the machines to recognize a new weed species took about 24 hours. The new model eliminates this bottleneck, allowing the robots to identify and target new weeds instantly.

The LPM is trained on more than 150 million photos and data points. This dataset was collected by the company’s fleet of LaserWeeder robots operating across more than 100 farms in 15 countries. Mikesell stated that the training set now contains over 150 million labeled plants, which allows the system to determine a plant’s species and structure even if the machine has never encountered that specific plant before. The model now powers Carbon AI, the AI system that serves as the software brain inside these autonomous weed-killing robots, distinguishing it from the physical LaserWeeder hardware.

Paul Mikesell, the founder and CEO of Carbon Robotics, noted that the update allows farmers to manage their fields dynamically. “The farmer can live in real time and say, ‘Hey, this is a new weed. I want you to kill this,’ and that was something that had never been done before,” Mikesell said. Under the new system, which reaches existing machines via a software update, farmers can select photos collected by the robot’s user interface to specify which plants to target and which to protect, bypassing the need for manual data labeling or retraining.

Mikesell, who previously worked at Uber and Meta, began developing the model shortly after the company started shipping its first machines. The company’s technology relies on neural networks—a type of AI architecture.

Key details on the company’s history and capitalization include:

• Founding and shipping: Carbon Robotics was founded in 2018 and began shipping its first machines in 2022.
• Venture backing: The company has raised more than $185 million in venture capital.
• Key investors: Backers include Nvidia NVentures, Bond, and Anthos Capital.

Why it matters

The introduction of the Large Plant Model eliminates the previous 24-hour retraining bottleneck for autonomous farm equipment. By allowing machines to adapt to new weed species in real time, the update shifts agricultural robotics away from static, pre-labeled datasets toward dynamic, on-field decision-making.