Ganbreeder is a machine learning image generator done by Joel Simon. It let users decide which image as the root and how different the new image should be. It also let users to crossbreed two different images to generate new images.
First of all, the “Make Children” button is very addictive and satisfying. With just one button and one slider, it gives me a sense that I have some kind of control or influence on the images that to be generated. I then tried to crossbreed two existing images to create the image that is in my head. I don’t think the GAN got it close, but it seems it doesn’t matter. By the time I saw the newly generated images, I already forgot the image that was in my head and intrigued by the new ones.
I think the project taps into the desire of constantly seeing new, unexpected, stimulating imagery. Combining that with the pseudo sense of power, that you can somehow steer where the images go, it is a facade that getting harder to see past.