John Deere acquires Spark AI

John Deere, a global leader in agricultural equipment, today announced that it has acquired Spark AI. Spark AI's "human-in-the-loop" technology enables AI models to be improved through human feedback. John Deere was itself a Spark AI customer.

John Deere is a U.S. manufacturer of agricultural machinery, particularly tractors and combines. The company was founded in 1837 by John Deere and is headquartered in Illinois, USA. With 2020 sales of $35 billion and approximately 73,000 employees worldwide, John Deere is one of the largest manufacturers of agricultural machinery.

Spark AI CEO Michael Kohen shared the acquisition name on his LinkedIn profile. Willy Pell also informed about the acquisition on his LinkedIn profile. Willy Pell was Vice President of Blue River Technology. A company that was also acquired by John Deere in 2017.

In challenging environments such as a farm or open terrain with dust, rain and snow, it can be difficult for machine learning to make safe decisions. That's why John Deere's autonomous tractors can automatically transmit images and meta-information to Spark AI. The goal is to solve difficult details of a scene to make decisions quickly. Human decisions and results from a properitarian decision system are combined here. Spark AI returns the solutions to the autonomous tractor and it combines the solution with existing knowledge about its environment. According to Spark AI, the entire process takes a few seconds and takes place several thousand times a day.

The Spark AI acquisition is part of John Deere's efforts to modernize agriculture by integrating AI and other technologies. In the past, John Deere already acquired several companies including numerous patents and other intellectual property rights from Light, a company that specialized in depth sensing and camera-based perception for autonomous vehicles, so John Deere also acquired Light's Clarity platform to integrate it into its tractors.

Acquired Light said on its corporate website at the time that "human vision" is the key component that enables machines to accurately and safely recognize their surroundings, and John Deere seems to be taking a purely visual approach to autonomy as well. Its new 8R autonomous tractor has six pairs of stereo cameras and does not use LiDAR, and the approach of relying solely on cameras and not using additional radar and LiDAR sensors is also being taken by Tesla. However, the U.S. regulatory authority takes a critical view of this. Tesla's Autopilot "Full Self-Driving Beta" continues to make numerous mistakes. John Deere's tractors drive much slower in comparison and only need to be able to see about 20 meters ahead. This could be one reason why they decided against LIDAR systems.