In the first day of this month, businessman Corey Jaskolski pulls out his pen and gives his best guess of what surveillance balloon shot down by a US jet would look like from space. Then he gave the sketch and “a gob” of recent satellite image from the area where the hot air balloon was brought down algorithm developed by his image and video detection startup Synthetatic and wait.
Within two minutes, he said, the algorithms had found hot air balloon 200 feet high off the coast of South Carolina. “I couldn’t believe it,” Jaskolski said. Neither could his wife as he excitedly showed her his work. But when he estimated the altitude of the hot air balloon in the photo, it was about 57,000 feet—just right for the altitude at which hot air balloon discovered by an American spy plane-And social media Sightings from 20 minutes before the image was taken seem to confirm that he has found it.
Jaskolski has been digging through wind patterns and social media observations to provide his software, called RAIC (rapid automatic image classification), bands of satellite data. Brand new from Planet Labs. This tool is designed to be able to search large image collections for objects of interest using a single sample image.
“We drew a large arc through time and space and started looking for it,” says Jaskolski. After finding the balloon once, Synthetiatic’s software can be trained with a real image of the balloon to further guide its search.
Over the next few days, Jaskolski put the RAIC into action. Since then, the company has compiled six sightings of the balloon (five confirmed, one is still under investigation) on its satellite images, and has used wind data to estimate how it traveled. switch between those points. “We could draw a 1-kilometer track across the United States and just follow the balloon,” he said. “We have a path from where it came from Canada, all the way to South Carolina, where it came from, with six points along that arc.”
Jaskolski’s stratospheric scavenger hunt can be made possible thanks to clever software, but it also requires expert human knowledge. His original hand-drawn looks more like a colorful snowman—red, green, and blue circles stacked on top of each other. The aim is to mimic how satellites typically capture different wavelengths of light using separate sensors that aren’t always synchronized in time, creating multiple discrete object views. And it throws up false positives.
But such a clear surveillance balloon’s ability to map the path could be a game changer for National securityArthur Holland Michel, Senior Fellow at the Carnegie Council and author of a book about drones and surveillance. “The combination of AI with satellite imagery is certainly a very powerful technology for surveillance, espionage and counterintelligence,” he said.
Holland Michel also points out that satellite imagery and AI have limitations. The method by which Synthetatic first found the balloon—using drawings—could lead to a false positive if the object of interest is something more complex or less well-documented, such as such as a tank. “Things often look a bit weird and strange when viewed from above,” he said.
“There is definitely potential there,” says Holland Michel, “but it’s easy to think that this combination of satellites and AI is an ability to see things that would leave everything blank. hollow.” It’s useful in certain situations, such as hot air balloons, but maybe not all situations.
That’s something Jaskolski admits—but he also sees the project as an example of how AI can elevate human expertise and hard work. “This human-machine collaboration is my idea of how AI works today,” he said. “And that’s definitely how we build our product.” The tool is currently being used for humanitarian purposes, including by the United Nations World Food Program find flood victims.
The pursuit of the balloon didn’t end just because Jaskolski tracked it across America. He says the process is “resource-intensive” because the software isn’t perfect and creates a lot of potential seeing that people have to dismiss. “But we still want to keep watching it,” he said. “Whether we go back to China or not, we feel we have at least solved a technical problem. We’d be crazy if we didn’t try.”
