SONOMA COUNTY, Calif. (KGO) -- Memories of recent disasters are still smoldering as the Bay Area waits on this year's drought-fueled fire season. But in Sonoma County, responders are hoping an extra set of eyes can help spot trouble before it erupts -- with the help of artificial intelligence.
"Yes, we're the only ones I believe beta testing the AI technology, says KT McNulty of the REDCOM dispatch center, which covers Sonoma County."
The dispatch center's walls are covered with monitors tapped into an existing network of remote fire cameras. But now, that system is being backed up by artificial intelligence software.
"When the AI detects a column of smoke an alarm is going off on that giant media wall, the screen turns red and it's very apparent, come look at the smoke column we've detected," McNulty explains.
The software was developed by a South Korean company called Alchera, which uses similar AI technology in Asia, to monitor everything from airports to major industrial systems and utilities. In this case, they taught the algorithm to quickly recognize smoke.
"What we're looking to do is catch fires early and often before 911," says Alchera's Robert Grey.
They worked to layer the software over the existing Alert Wildfire network, a powerful fire spotting consortium made up of utilities like PG&E, state and local agencies, and others who have cameras in a position to spot fires. Alert Wildfire members often build the cameras and help place them. The network has not only grown to include more than 800-cameras but it's also launched an uncounted army of professional and amateur fire spotters who monitor them on the network's website. Graham Kent is the director and is closely watching the AI project.
"I'm agnostic,I don't care if it's a citizen, please don't stop calling in. If it's AI great," he says.
Kent says the gold standard for reporting speed has always been 911, in part because so many people in California are motivated to join the fight against wildfires. But early feedback on the AI system is promising. Back at REDCOM, KT McNulty says they recently logged roughly 60 alerts over a one-week period since the AI officially went online last month, including one that beat the 911 report by a significant margin.
"The AI camera was able to pick up the smoke 10 minutes before her 911 call. So we had a 10-minute jump on a valid fire, says McNulty.
Officials in Sonoma County used part of a federal grant to install the AI system. If it's successful in Sonoma, the company is hoping to add more counties around the Bay Area and statewide. Alert Wildfire is also rapidly expanding its own network after a decade of growth. Graham Kent, who is a professor of engineering and geological sciences at the University of Nevada, Reno, says they're hoping to reach 1,000 cameras in the near future. Supporters of the project are hoping the combined power of 911, camera crowdsourcing, and AI will all lace together in time for one of the most concerning fire seasons the Bay Area has faced in some time.
To learn more about Alert Wildfire and get a look at their camera system: http://www.alertwildfire.org/