- Data access and quality is a limiting factor for the development of future AI applications, said Deere CTO Jahmy Hindman, speaking at CES 2022 on Jan. 5. Data labeling is still mainly a manual process, which sets the pace for developing advanced models in the enterprise.
- To respond to the limitations, the company uses automated data labeling, which lets Deere speed up the process of building future AI models, Hindman said.
- Enterprise leaders looking to accelerate their AI strategies must make sure to "set up data architectures with the future in mind and in a structured way, so that making the data useful doesn't take a lot of effort," Hindman said.
Deere, a top manufacturer of equipment for the agriculture industry and maker of the John Deere brand, plans to use AI-based technology to revolutionize agriculture. The most recent result of the company's AI strategy is its autonomous tractor, presented during CES.
Available to farmers later this year, the company's new offering combines the 8R tractor with a GPS guidance system, self-driving capabilities and stereo cameras for obstacle detection. At the center of its development is visual data.
Visual data helps the autonomous tractor system detect objects around the tractor for navigation, while detecting the health of specific crops the unit drives through, or the conditions of the field behind it, said Hindman.
Just one in five enterprises are considered mature adopters, or AI leaders, according to a report from Cognizant. The consulting company determined the importance of AI for business processes and the number of areas where AI is deployed for the report.
To reach mature AI applications, Deere starts with a data architecture that enables future use. Then, data validation and auditing lets the company advance products under development, as it pursues more complex AI-based applications.
To future-proof data collected for AI applications, Hindman recommends companies collect it "at the highest resolution that you possibly can [and] at the highest frequency that you possibly can, because you don't know what the future use cases for it are."