3 reasons Twitter just bought machine-learning startup Magic Pony
Twitter has made no secret of its interest in machine learning in recent years, and on Monday the company put its money where its mouth is once again by purchasing London startup Magic Pony Technology, which has focused on visual processing.
“Magic Pony’s technology — based on research by the team to create algorithms that can understand the features of imagery — will be used to enhance our strength in live [streaming] and video and opens up a whole lot of exciting creative possibilities for Twitter,” Twitter cofounder and CEO Jack Dorsey wrote in a blog post announcing the news.
The startup’s team includes 11 Ph.Ds with expertise across computer vision, machine learning, high-performance computing, and computational neuroscience, Dorsey said. They’ll join Twitter’s Cortex group, made up of engineers, data scientists, and machine-learning researchers.
Terms of the deal were not disclosed.
The acquisition follows several related purchases by the social media giant, including Madbits in 2014 and Whetlab last year.
This time, Twitter stands to benefit in three ways.
1. Better videos
Video quality is surely the most obvious reason Twitter undertook the move.
“Though video and live video are becoming increasingly critical parts of social and sharing sites, image quality often leaves much to be desired,” said Charles King, principal analyst with Pund-IT. “That’s a problem that Magic Pony’s team has tackled successfully by developing algorithms that automatically sharpen and fix low-resolution and blurred images.”
The deal should help Twitter enhance the quality of videos shared on its site, thus improving user experience and enjoyment. That’s a big deal for a site that’s been struggling to attract and retain loyal users.
2. A sharper focus on events
Indexing and categorizing unstructured data like pictures and videos isn’t easy, but it’s critical to making sure they can be found and promoted. That’s especially important for real-time news, which is Twitter’s specialty, said Rob Enderle, principal analyst with Enderle Group.
“With this they can take an event like the Orlando shooting and collect, near instantly, all the user-created video and pictures into a single stream and provide a level of coverage a network would envy,” Enderle said. “You could literally watch the stories emerge through the virtual eyes of the folks there.”
Combined with commentary, that content could essentially form a news service, either from Twitter or through another business that pays Twitter for the technology, Enderle said.
Machine learning “could become one of their best paths to profitability, because the result could be both compelling and a huge ad revenue generator,” Enderle said.
Of course, “I’d anticipate a huge spike in spending before they get the benefit of the result,” he added. “That could be problematic for investors.”
3. Deeper analytics
Machine learning and neural networks are becoming “a very hot part of the overall big data and analytics market space,” and with investments like this one, Twitter can “move beyond being a data source of human comments” and push further into analytics, said Nik Rouda, a senior analyst with Enterprise Strategy Group.
For example, it could use the new capabilities to start offering intelligence services to analyze the Twitter streams of businesses that don’t yet have machine-learning capabilities of their own, Rouda said.
It could also move beyond just hashtags and text analytics to find trending topics in the images and videos that get tweeted, he added.
Twitter could use the technology to identify specific people, locations and things in photos and video, and to profile users for marketing ads or driving more engagement, he suggested.
Finally, it could tap machine learning to identify copyrighted materials — “even those that have been modified in memes,” he said.
“This helps Twitter continue to evolve beyond just short text comments into some potentially more valuable areas,” Rouda said.
Source: InfoWorld Big Data