The Story of Deepfakes

January 21, 2021

After years of continual public misinformation campaigns, it's only a matter of time before our social media timelines are littered with deepfakes, but in a sense, deepfakes are already here. The idea of deepfakes is far more consequential than the actual existence of deepfakes.

At the simplest level, deepfakes are synthetically generated media that alters a person's likeness or creates a new person out of strings of binary. They can be audio, video, or perhaps even text depending on your definition.

What distinguishes a deepfake from classic Hollywood visual effects is that deepfakes can eventually be made cheaply, quickly, and at scale by anyone via machine learning.

Democratized production aside, deepfakes carry added significance. Deepfakes embody a mythology of perfection.

Despite only living at the fringes of their arrival, our society has adopted a collective narrative that deepfakes are indistinguishable from reality.

With modern technology, this idea of perfect deepfakes is still a fiction.

Most creations require significant manual visual effects editing after the machine learning algorithms have done their work. Even with the extra editing, most retain obvious tells exposing them as counterfeit.

But what matters is not the reality of our tech, but the story.

If you believe deepfakes can be perfect, then any video, regardless of quality, could be a deepfake.

A video capturing a political scandal or police misconduct? How can you determine its authenticity, and in a post-truth society, do determinations of authenticity matter at all?


© 2020 Connor Daly