1 Comment

Summary:

While using machine learning over large data sets to serve up ads inside social networks isn’t new, there’s an era emerging where social network data can be used to help people solve important problems.

David Gutelius Jive Software Structure Data 2013
photo: Albert Chau

Your Facebook wall might be filled with photos of Lil Bub and Grumpy Cat, but underneath the hood, social networks actually do a lot of work. And there’s a growing class of social networks being used by companies and organizations that are tapping into machine learning to solve problems, explained Jeffrey Davitz, CEO of social data startup Solariat, at GigaOM’s Structure:Data event in New York on Wednesday.

These types of networks, called active networks, crunch piles of user data and use artificial intelligence to augment human tasks and goals. In contrast to other AI systems where humans augment algorithms, active networks use machine learning to augment human decisions. “It’s the flip of Watson,” said Davtiz.

Both Davitz, and fellow panelist David Gutelius, the Chief Social Scientist at Jive Software, previously worked on active networks developed for military applications through the CALO project (Cognitive Assistant that Learns and Organizes), which was one of the largest AI projects in history backed by DARPA. The teams built cyber assistants as part of military social networks that could use data on social interactions to give recommendations, like areas to avoid in a conflict zone, or more effective strategies.

Active networks could be used for other applications, too, like fending off cyber security threats, and making working groups within a company work more effectively together. The “collective capability” of the digital assistant and humans is new, and the industry is just getting started, said Gutelius. Apple’s Siri was a spinoff of the CALO effort.

The challenge of designing and building these types of networks is figuring out where the machine learning agent leaves off and the human social networks takes up, said Gutelius. The key to the design is making the technology recede into the background, and make it unobtrusive, said Gutelius, who says he spent six months trying to get a particular social network interface for military offers work.

While using machine learning over large data sets to serve up ads inside social networks isn’t new, Gutelius and Davitz see an era where social network data can be used to help people and solve important problems.

Check out the rest of our Structure:Data 2013 coverage here, and a video embed of the session follows below:


A transcription of the video follows on the next page

You’re subscribed! If you like, you can update your settings

page of 2

You're subscribed! If you like, you can update your settings

Related stories

  1. > In contrast to other AI systems where humans augment algorithms, active networks use machine learning to augment human decisions

    And the best and most effective AI systems use a blend of both methods

Comments have been disabled for this post