Fabiola Larios Artificial Intelligence Artist

This piece extends the history of self-portraits as a category of image. The subject as a public image is a self-representation that has morphed from 17th-century painted portraits, all the way to 21st-century accessible selfies on the internet and social media. A self-portrait which was once an image created by painters for those with economic reach is now an everyday feature in smartphone cameras. The work, “Internet Humans” reflects on the current techno-political environment surrounding data-sensitivity and surveillance capitalism through the historical context of appropriation in contemporary art. The scraped dataset for the machine learning model was gathered through Instagram hashtags, public profile pictures on Facebook, and voluntary personal relationships. Although the dataset is not public, the issues of data sensitivity and privacy in regards to machine learning are raised, leaving one asking, are these companies using AI with my photos? We treat our information with such little care that it is easily available to be used for profit, advertising campaigns, and against us at our expense. Accessible data on the internet is now public property for computational and economic gains.

"Infinite Taco" is a generative taco video interpolation using the StyleGAN2 machine learning model.

The model was trained on 1000 images of traditional Mexican Tacos. The audio is from an authentic Mexican Tianguis (Market).

Artificial Selfie Queen deifies the internet culture surrounding Paris Hilton. The work reflects Paris’s social media iconicity. As the inventor of the Selfie, she has been reproduced massively with over 177,000,000 search results. The generated artificial model is achievable because of Hilton’s profound reach in beauty culture and portrait iconography. The machine learning network further creates uncanny selfies of Hilton’s life: the one she shows to us and the one the internet gets through other sources. The piece encrypts her legacy into an internet object, an NFT, and now an artificially generated persona.

800 images of emo selfies circa 2008 were used to train a machine learning model based on a custom pre-trained StyleGAN2 model entitled “SelfieGAN”. By using the custom .pkl file trained on selfies, we were able to maintain the selfie angle while also ensuring a variety in the color of the skin. The custom .pkl file also helped keep the age of the generated humans between 15 and 35 years old. The music was generated using JukeBox by openAI to train a neural network with the audio and lyrics of “Note to Self” by From First to Last. emoGAN is a project that demonstrates not only AI's ability to learn the facial features that make us human, but also AI's ability to learn the styles, fashions, and visual cues to signal that we are a part of a community, subculture, or tribe.

"AI Internet" is a video piece visualizing the nostalgic dial-up tone of pre-wifi landlines with the Machine Learning audio-reactive BigGAN model. The work generates liquid computers that flow to the signal of an internet connection, a moving image between state-of-the-art technology, and programmed obsolescence.

Video piece visualizing liquid phones that flow to the nostalgic Nokia ringtone. The work highlights the economics of planned obsolescence with technological innovations.