Now Playing

How Artificial Intelligence Could Help Fight—Or Even Prevent—Another Wine Country Wildfire

Meet the disaster algorithms.

One Concern’s disaster-management system.


The North Bay fires weren’t just a natural disaster, they were an information disaster too. Residents alerted their neighbors of the impending blaze by knocking on each others doors, and local governments faced criticism for not activating public safety alerts soon enough. When the next wildfire strikes, will public safety officials be able to find a better way? The co-author of a study on wildfire-predicting artificial intelligence, Mike Flannigan, says that with enough information, absolutely yes: artificial intelligence could help fight—or even predict—the next catastrophe. “That’s the beauty of it,” the University of Alberta researcher says. “It’ll get better and better as we have more data coming into the system.”

Here are some ways that artificial intelligence is already being used to fight fires.

Wildfire-predicting systems like the one currently used by Cal Fire rely on the same spotty meteorological models as the ones you have in your iPhone. However Flannigan’s new AI system, proposed in a research paper by two universities and the Canadian Forest Service, crunches a complex set of metrics for a more accurate forecast. Better predictions would mean more time to rally firefighters and evacuate victims, and would be more powerful still, researchers say, if used in tandem with existing forecasting AI. And when disaster does strike, Palo-Alto based One Concern’s disaster-management system can identify the hardest-hit and highest-need areas within 15 minutes. The two-year-old startup adopted by San Francisco, Los Angeles, and Woodside, uses AI to run disaster simulations to test a city’s capacity for emergency response.

Instead of monitoring Twitter by hand for real-time information on people affected by disasters, AI products like Ushahidi and TweetTracker enable first responders and relief agencies to navigate and process content on those networks—for instance, grouping requests for goods with people offering to help. At the same time, dispatch centers have begun implementing AI systems like RescueNet Dispatch to anticipate calls before they come in, alleviating pressure during an emergency. One study found that a related diagnostic AI led to a 22 percent reduction in dispatch delays.

First responders from the Menlo Park Fire District fly Intel’s Falcon 8+ non-autonomous drones in search and rescue operations. Intel confirms that it’ll soon integrate AI technology into such drones, allowing them to do their duties autonomously. In the meantime, Allstate combines AI and drones to process insurance claims. Instead of knocking on doors, claim adjusters pilot drones over damaged homes and snap aerial photos, which the AI uses to evaluate claims in half the time. AI technology could also have a bright future behind from the front lines. A virtual therapist service, Woebot, conducts daily check-ins with patients over Facebook Messenger. The Stanford team behind the app says the system reduced anxiety and depression symptoms after two weeks of daily chats. The app has its limits, of course: “I would also hate to think that anyone would chat to me instead of getting help,” Woebot says on the company site. “I’m only a robot, after all.”


Have feedback? Email us at
Email Kelsey Lannin at
Follow us on Twitter @sanfranmag
Follow Kelsey Lannin on Twitter @KelseyLannin