Dan Nuñez pulls up a video feed on his computer.
“Can you see that?” he asks, pointing to some rolling, foggy hills in rural Oregon. The footage was captured seven minutes before Nuñez opened up his browser. Barely perceptible on the screen is a rising plume of smoke.
“I would not have caught that,” Nuñez said after hitting replay. “But the AI does.” The artificial intelligence Nuñez is referring to comes from Pano AI, a California startup that makes high-definition cameras to detect, track and manage flames. As the manager of wildfire planning and analytics at Portland General Electric, Nuñez is responsible for doing everything possible, as early as possible, to find and put out fires.
A cottage industry in fire detection technology has taken root, especially in the western US, as fires have become more frequent, larger and financially ruinous.
PGE so far has installed five Pano cameras, with plans to add 17 more across Oregon this summer. The hardware offers minute-by-minute snapshots of the surroundings. According to Sonia Kastner, the startup’s founder and chief executive officer, Pano’s software can distinguish benign clouds from troubling smoke. The artificial intelligence that makes that possible has become the main selling point. For firefighters, government agencies and utilities, spotting wildfires early is critical to limiting the damage.
“Minutes matter here,” said Mark Miller, chief commercial officer for AEM, an environmental technology company owned by Union Park Capital. AEM sells weather stations and panoramic cameras for weather detection, along with satellite data. Like Pano, AEM says it relies on advanced AI that has learned to identify what a wildfire looks like and when it arrives.
False alarms do occur, but less than some might expect. Pano operates a call centre where staff analyse the video footage before alerting local customers or authorities. Kastner said Pano’s 32 active camera systems have spotted actual wildfires with 90% accuracy.
Cameras aren’t particularly novel solutions. Marin County, north of San Francisco, installed a fleet of remote security cameras to replace volunteer fire lookouts in 2014, after a record drought. The county had to swap them out within four years because they couldn’t properly distinguish smoke from fog. Now Marin uses a statewide camera system that multiple utilities finance.
Nuñez believes additional use cases for the devices will emerge. “From a safety and environmental risk perspective, it’s probably one of the most important projects in our portfolio,” he said. PGE has spent US$4.5mil (RM19.80mil) on Pano’s systems so far.
Certainly, utilities and state governments are spending more to combat the problem. PGE has budgeted US$32mil (RM140.86mil) for wildfire operations this year. In 2021, the agency reported incremental costs of US$45mil (RM198.09mil) from wildfire responses, up threefold from the year before.
The sums are miniscule compared with what Pano and other consumer or enterprise tech companies can earn in revenue. (Pano declined to disclose that figure.) Pano’s Kastner used to work on smart-home devices for Google’s Nest. She founded the company in 2019, then recruited other alums from Google, Apple and Tesla.
The startup estimated that California would only need 1,800 of its cameras to have fully equipped statewide detection, if the money arrives. “We’re a team that’s used to shipping millions of products a month,” said Kastner.
Pano also markets its cameras to private businesses, including a timber company in Australia, where the startup will install six cameras this summer. It is working with two resorts in Big Sky, Montana. AEM, according to Miller, has more than 6,500 customers, primarily in government; there are some prospective deals AEM is exploring with insurance companies.
Chris Martinelli, the fire captain for Marin County, has seen waves of demos from companies pitching AI for fire detection. “Some of them are really expensive,” he said. He’s mostly interested in adding the software to the county’s existing cameras. And he’s yet to see any product that helps with detection at night. For that, the AI tech might not offer any advantage. “It’s not as beneficial as just getting a visual,” he said. – Bloomberg