Creative Adversarial Networks: GANs that make art

Originally published at:


I was playing with the not-a-kitten and not-a-human GANs, and I really wanted to add my own photos. Very frustrating that’s not an option.

Then I saw a comment in a thread that it takes networks milliseconds to generate images, but hours or days to train on new data. Yea, I don’t want it that bad. But I do really want hot babes in bikinis with Francis Bacon faces.


The underlying theory is that art evolves “through small alterations to a known style that produce a new one”

Well, that may be one way a new style of art is created, but any reading of art history tells us this isn’t a sufficient description. There are radical transformations and breaks in art styles and definitions - e.g. Baroness Elsa von Freytag- Loringhoven’s “Fountain” (wrongly attributed to Duchamp).
The problem with these particular GAN images is that they’re art like applying impressionist Photoshop filters to random images is art. It’s not that art can’t be produced that way, but it isn’t, necessarily. If there’s art here, it’s the software itself. Which wasn’t created by the so-called artists, nor was it created to be art not is it even being presented as a readymade.

1 Like




1 Like

Just when my Pentium IV was taking a tempestuous like to visitors’ guests, laying in spiced rum lemonade drinks and drinking only cream sherry, painting in gesso using the ends of cast off 20oz. soda bottles, and making rent in NYC, the ‘please print’ visual GAN pulls this kind of thing. Why, if its therapist wasn’t in CRAN this would be upsetting.

1 Like

I may not know art, but I know “meh.”

This topic was automatically closed after 5 days. New replies are no longer allowed.