实时侦测全球新闻动态的网络应用 Banjo 宣布 融资1亿美元。这玩艺威武，一架堪比 NSA 的巨大的实时监控机器？当然，对象不同，Banjo 监控的是新闻。机器、算法，非人的因素，在未来新闻传播中的影响力，会超越人吗？
4:14 pm ET May 6, 2015
Banjo Raises $100 Million to Detect World Events in Real Time
Banjo started four years ago as a simple app for viewing social-media posts tagged to a location or a popular event, like a Beyoncé concert. With $100 million in new funding, the company aims to build something much bigger.
SoftBankhas invested that amount in Banjo to help the startup build software that can detect earthquake-damaged buildings in Nepal, drive-thru patterns atMcDonald’srestaurants, gas-line explosions in Saudi Arabia, and dozens of other events all over the world in real time, using photos and videos uploaded to social media sites.
The funding is SoftBank’s largest investment in a U.S. tech startup since Nikesh Arora, the longtimeGoogleexecutive, joined the Japanese tech giant last year to lead its venture deals. Most of Arora’s deals have focused on tech firms outside the U.S., including a $627 million investment in India e-commerce startup Snapdeal and a $600 million bet on Chinese taxi-booking service Kuaidi Dache.
Arora was attracted to Banjo’s big vision of a machine that can scan the world’s tweets, Instagram photos, updates to Facebook, Chinese social network Weibo and Russian website Vkontakte along with billions of other messages, images and videos unfolding in real time around the world.
An early version of this software is already being used by large media companies as an alert system for breaking news, such as the shootings at Florida State University in November, and to zoom in and monitor ongoing events, such as the Baltimore riots.
Damien Patton, Banjo’s founder and chief executive, likes to call the program a “crystal ball,” because it gives users a view of the entire globe. His team has overlaid a map of the globe with a grid of 35 billion squares, each one about the size of a football field. Any time an image is publicly shared to the Web and tagged with a specific location, Banjo’s system automatically places it within the grid and keeps track of what time it happened and classifies objects within the image. Patton said his computers currently process about 1 quadrillion computations every 10 seconds, and by the end of the year that will be up to 1 quadrillion per second.
Social media doesn’t provide all the data Banjo needs to create a full picture of some events, so the company is working to add more data sources, such as weather data from various countries around the world and satellite imagery.
Banjo’s software recognizes objects within photos posted publicly on social media sites and attempts to make sense of them.
banjo
“People were capturing the explosion by accident,” he said. “It’s like a giant Tivo – I just hit rewind and it played the entire incident.”
Once the software knows what each square of the grid looks like normally, it can detect when something is amiss. When a gas explosion started a fire in a Manhattan apartment in March, “somebody took a photograph of it before 911 was called,” Patton said. Banjo’s software detected something was new in its image of that particular square, and went to work identifying the shapes inside the new image to determine it was an explosion. The Banjo team then alerted their media partners, who reported the news.
Patton came up with the idea for his crystal ball on the day of the Boston Marathon in April 2013. Initially, onlookers didn’t know a bomb had caused the explosion near the finish line, so in the minutes after the incident he scanned photos that had been collected and tagged to that location on the earlier version of Banjo. Piecing them together and seeing for himself that it had been a bomb, he realized that the technology could be valuable to a range of companies who need to quickly scan the world for imagery in real time, or even go back in time and replay events.
“People were capturing the explosion by accident,” he said. “It’s like a giant Tivo – I just hit rewind and it played the entire incident.”
The technology is already becoming a central part of news gathering toSinclair Broadcast Group, an owner of more than 150 local TV stations across the country and one of Banjo’s first business customers. Rob Weisbord, Sinclair’s chief operating officer, said the company has trained about 3 to 5 employees in each market to use the software.
A Sinclair station in Baltimore showed images from inside the recent riots it collected from Banjo. One of its stations broke news about a recent shooting in New Orleans, 15 minutes before national outlets like CNN had the story, Weisbord said.
Banjo, based in Redwood City, Calif., charges tens of thousands of dollars for annual subscriptions to companies such as Sinclair. Its customers also include advertisers such as Anheuser Busch, which uses the system to find social media images they might place in marketing; and financial services firms, which use Banjo to alert them to world events that impact global markets.
Patton plans to use the funding to build a formidable team of data scientists and engineers who specialize in computer imagery. Banjo has already doubled in size to more than 60 employees in the last five months, and plans to double again by the end of this year.
Banjo has raised a total of more than $120 million from venture-capital investors including Balderton Capital, BlueRun Ventures and VegasTechFund. The company didn’t disclose its valuation or other terms of the new investment.
