An AI prediction of the «future».

Four months of my personal diary were used as a prompt for GPT-3 to predict a continuation of it. Using two AI systems, the work combines machine-generated text with machine-generated images. Each generated diary entry was then transformed into a visual scene using VQGAN+CLIP. The «future diary» begins on May 11, 2022, and extends to 2024, two years into the future (now the past).

The installation is a display of an «AI slop» (almost 400 texts, images and gifs) that examines the predictability and lack of originality, effort, and potentially, accuracy of AI as a way to predict the future, and instead only repeating the past and amplifying the biases introduced in the training and prompt. It explores large language models (LLMs) biases and the concept of computer-generated imagination (with now unsurprisingly outdated images).

-> Exhibitions
Installed at PATRIM’22: Futurs incerts, at Museu Can Framis, Barcelona, 2023
Installed at Territorios de cambio, at Fundació Barceló, Palma de Mallorca, 2023
Installed at SENSE TÍTOL ’22, at Facultat de Belles Arts UB, Barcelona, 2022

↓ Download research (Spanish pdf) (diposit.ub)