Skip to Main Content U.S. Department of Energy
Deep Science Agile Initiative

News and Highlights

Teaching AI to Identify Clouds

artificial intelligence (AI) system

Co-PIs Donna Flynn and Erol Cromwell's (PCSD) work on developing an artificial intelligence (AI) system was recently featured in the Wall Street Journal, which reaches over 2.2 million readers. The system is being designed to distinguish clouds from other atmospheric constituents in lidar imagery.

To "teach" the lidar-interpreting neural network at the Lab, Donna has been labeling lidar images of clouds to allow Erol to train an AI model to interpret the images as she does. Using her extensive knowledge of lidar measurements of the atmosphere, Donna labels the roughly two million pixels in each lidar image as cloud, smoke, or dust. If successful, this data can produce information that helps climate and weather forecast models improve their ability to assess predictions of cloud formation and identification of atmospheric features. On a test set of images, Donna and Erol's model achieved a 30.89% increase in accuracy over the best-performing previous algorithm.

The article summarized the state of AI and how it's impacting numerous business sectors in the U.S. and addressed the potential of AI to both eliminate and create jobs. The future promises a need for many new AI "teachers" and managers. Donna and Erol's work is one example of this need, which will grow as industries employ AI to accomplish a myriad of tasks.

Deep Learning