How Giphy is using AI to change social communication
This article is part of our “How AI is changing the world” event series, held in San Francisco, New York, and Tel Aviv from June to November 2019, featuring insights by leading scientists and entrepreneurs on how AI will change healthcare, communication, agriculture, travel, and other industries. Check out all 12 talks here.
* * * * *
With services like Giphy, AI is helping with online social interactions. Anthony Johnson, CTO of Giphy, described how challenging something as simple as searching for a GIF can be. Giphy is a service used by Facebook, Twitter, Slack, Instagram, SnapChat, and other companies that provides GIF images users can add to their posts.
Interaction with Giphy is simple: You type a few words describing the topic of interest and receive back GIFs that match your interest. According to Anthony, searching using a few terms, is surprisingly difficult. He points out that images and language are deeply entwined and a small number of words is not enough to constrain the search. Moreover, thanks to technology, the breadth of how we communicate has grown exponentially from what it was just a few generations ago.
“Imagine 400 years ago when society was much more basically homogeneous. Everyone had to learn the same poetry, they had a shared literature,” Anthony says. “You could have a lot of ways of sharing things that go beyond a single word. We’ve lost that.”
The internet has globalized communications and created a shared culture that could be a reference model for understanding what people mean when they say things in few words is still developing. Anthony says one challenge now it to create a common language for a global user base that has different experiences and lifestyles.
Giphy is using AI to model both the user’s interests and intention as well as the content creator who uploads GIFs. Machine language programmers at Giphy model relationships between people and between GIFs. The are also using AI for language modeling to help capture user’s intent. All of these models are only possible because of massive amounts of data that are currently available.