Twitter Buys Another Machine Learning Startup

Machine learning continues to make inroads among hyper-scalers who are increasingly using it to train rather then program algorithms. As if the technology needed any more promotion, Twitter announced its acquisition of a London-based company that has developed machine-learning techniques for visual processing.

According to reports, Twitter (NYSE: TWTR ) paid about $150 million for AI startup Magic Pony Technology. Other details of the deal announced Monday (June 20) were not disclosed. The social media company’s stock rose on news of the transaction.

The deal gives Twitter another machine learning tool, this one designed to improve live and streaming video. Twitter previously invested in the Harvard University machine-learning spinout Whetlab last year and image search startup Madbits in July 2014.

Twitter CEO Jack Dorsey noted in a blog post that Magic Pony’s development team would join its Cortex team of data scientists and machine-learning researchers. “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 and video,” Dorsey said. The Magic Pony staff includes eleven PhDs, including co-founders Zehan Wang and CEO Rob Bishop. The team specializes in computer vision, high-performance computing, computational neuroscience as well as machine learning.

Twitter’s latest machine-learning acquisition underscores the growing importance of the AI technology as it finds new automation applications required to process and organize huge datasets. The social media company’s third machine-leaning acquisition over the last two years gives it one of the strongest deep learning research teams as Twitter, Facebook and other social media companies look to augment their services with more video and even virtual reality offerings.

Machine learning and neural network technologies are widely viewed as among the most promising of AI technologies for handling the enormous datasets involved in these new network offerings.

“Our team has researched and developed state-of-the-art machine learning techniques for visual processing that can identify the features of imagery and use that information to process it in new ways,” said Rob Bishop, Magic Pony CEO and co-founder.

The startup’s technology reportedly works by combining neural networks designed to learn with machine learning tools to expand the amount of data in an image. The technique is said to enhance images by adding data from composite images. That capability is seen as important for image and video quality since Twitter users tend to post images from mobile devices. Hence, the machine-learning approach could be used to enhance those images.

Others have sought to leverage machine learning and machine vision tools to get a handle on the online image explosion. For example, image and music licensor Shutterstock Inc. (NYSE: SSTK) launched new search features in April based on machine learning techniques and, specifically, its proprietary “convolutional” neural network technology.

Convolutional neural networks are a machine-learning construct usually comprised of multiple layers that is often followed by more fully connected layers as in a standard multilayer neural network. Convolution nets are made up of neurons with “learnable” weights and biases.

Shutterstock said the first application of neural net technology is a “reverse image search” as an alternative to standard keyword searches.

Recent items:

Machine Vision Used to Wrangle Image Explosion

Unleashing Artificial Intelligence With Human-Assisted Machine Learning