GANS and the Future of American Fiction

GANS and the Future of American Fiction

Generative adversarial networks (GANS) can create human images almost indecipherable from actual images. Below I detail how creative writers and students of media studies can benefit from this technology.

James McAdams | September 7, 2020

What are GANS? 

GANS (or generative adversarial networks) denote computerized processes that combine thousands of random pixels to create a uniquely new human face. You can find an infinite repository here, at This Person Does Not Exist.  You can read about the technology on this page; I won’t attempt to parse it. Instead, I’ll give some examples of “the people who do not exist”:

GANS of little girl
A GANS-generated image of a young girl.
GANS image of young woman with daughter
GANS-generated image of young woman.
GANS image of middle-aged Hispanic women
GANS-generated image of woman.

 

 

 

 

 

 

 

 

 

I call these pictures GANS, although GANS (“generative adversarial networks”) refers solely to the process of generating them. But GANS make sense to me, because it differentiates them from PhotoShop. In Photoshop, you take a real image and alter it. There is no real image here, or rather there are billions of “real” pixel recombined into something new. Think Terminator-2  or Her. Think of “The Singularity” here. It’s creepy, but there are uses for authors. 

Three Ways GANS Help us Write Fiction

They provide visual aides.

As I’ve written about a lot in this blog, my novel consists of a large number of characters in a FL rehab. In some cases, these characters are hazy. GANS provide a great answer here. I can recycle through pictures for hours and then a magical click occurs, and I will “recognize” my character. The third image on the right above, picturing a middle-aged woman of Hispanic descent, automatically registered as Abdaliz. Abdaliz is one of the older staff in my novel, a kindly, charitable, exhausted woman who takes in addicts as her children. She “looks” exactly like that character. 

They let us use show the reader pictures of our characters without getting sued.

As literature becomes more digital and multi-media, more and more books are embracing graphics. We no longer just have text to work with. However, most pictures are protected from us by copyright or fair use legislation. In my case, in which I’m writing a prison memoir based on real people, HIPAA (or whatever) would never let me get away with it. But there isn’t protection for GANS, because they a) either don’t exist; or b) have no legal rights if they do. 

They provide a visual shorthand for characters.

As humans, we think we can infer things from their appearance. In my literature classes, I often provide students with 5 GANS and ask them to write a page describing that person without conferring with others. It’s amazing how often we’re all “certain” about a GAN: that the woman in yoga pants lives with many cats; that a man with a grey mustache and smart eyes is an engineer; that the woman in the middle above is a painter who emigrated from Eastern Europe. She was a real assignment and 12/20 of us indicated she was some form of artist. The level of consensus is astonishing. 

Faces have inspired writers since our ancestors wrote on caves. Many writers I know work in coffee cafes or libraries, explaining, like Tolstoy, that the faces motivate them. Like them, I’ve often found people on social media or online dating sites in whom I “recognize” a fictional character. I meet their real-world Doppelganger. GANS are certainly creepy, but ethically I feel less guilty using them, since there is no real person to betray. As David Bowie said, “Face the Change.” 

 

 

Leave a Reply

Become a Creative Pinellas Supporter