Artificial Intelligence to fight California Wildfires

By Shirin Mirdamadi-Tehrani

A Map of California Wildfires in 2020; Data source: MODIS Active Fire Detections for CONUS (2020), Geospatial Technology and Applications Center, U.S. Forest Service, USDA.

Image from Wikipedia licensed under the Creative Commons Attribution-Share Alike 4.0 International license.

After a record year for wildfires in California, technology may soon change how the state fights fires. According to the California Department of Forestry and Fire Protection, over four million acres in California have been burned by wildfires in 2020.

Now, Chooch AI, a Silicon Valley startup, is looking to artificial intelligence (AI) imaging solutions to prevent more years of record wildfires in the state. Jeffrey Goldsmith, Vice President of Marketing at Chooch, said their technology has the potential to identify and report fires in about 15 minutes.

“In a few minutes you don't have a big task—you have a small fire to put out. In eight hours, you might have ten acres to put out,” Goldsmith said.

The Chooch technology can be configured to multiple camera devices. Through these cameras, the system can directly monitor for smoke and fire—otherwise invisible to human detection. When the system detects smoke or fire, an image or video is captured and sent as a timely report to a specified contact number or email address.

By the end of November, the company intends to deploy this technology for small-scale testing, with statewide plans for next year.

However, according to the National Park Service, the majority of California wildfires are caused by humans. In addition to the nature of AI data collection, this reality raises privacy concerns in applying the Chooch technology.

Lourdes Turrecha, Privacy Tech and Law Fellow at Santa Clara University School of Law, said that privacy is a complicated but integral topic.

According to Turrecha, privacy concerns are inherent to AI, which must be acknowledged and addressed throughout the implementation process.

She said there are several key considerations, such as data accuracy, provision of notice, and long-term data retention.

Goldsmith said Chooch has also considered privacy questions in applying their AI technology to wildfire monitoring and detection. He said he sees Chooch not as a data collector, but as an “alert distributor.”

“Chooch will only detect things based on the AI models it has loaded onto it. We don't do facial recognition… so there is no privacy concern about people or license plate detection. If all we are distributing is smoke and fire detection the cameras can’t detect anything else. So it really is about the AI models that you deploy that allow Chooch to understand anything in the world,” Goldsmith said.

Turrecha noted that a major pillar of privacy is transparency—interpreted as user notice and consent, and she underscored the importance of weighing public interest against individual rights.

“If you have a legal basis for collecting [data] and you’ve made the assessment that this legitimate interest trumps any individual rights, there are exceptions to our rights,” Turrecha said. Here, the public interest is wide-scale wildfire monitoring and prevention.

Accordingly, the California Department of Forestry and Fire Protection (CalFire) is a proponent of integrating new technologies, such as Chooch, in their wildfire solutions. Geoff Marshall, CalFire’s Chief of Predictive Services, said that CalFire has already installed a state-wide network of cameras to monitor wildfires. However, AI capabilities have not yet been incorporated.

In 2019, California Governor Gavin Newson signed an executive order, titled Request for Innovative Ideas, to encourage innovative collaborations and problem solving in addressing the state’s wildfire crisis. With this directive, CalFire has since vetted predictive wildfire technology for wide-scale use, according to Marshall. Thus, while CalFire has not yet interfaced with Chooch, the stage has been set for AI to become part of California’s wildfire solution.

(Editor's Note: This article was originally published in the October 2020 [Volume 51, Issue 1] edition of The Advocate.)