Data Driven fiction

Archaeology of large ML models

The End of the Diffusion Era.

The only totally reliable historical documents are large Machine Learning models. For example, the state of human knowledge in 2019 was documented by the GPT3 model. The state of visual consciousness before the post-Diffusion Era (DE) — is the fantastic DALL·E 2 model.

Reverse Prompting

Everything that came after was only messed up by fine-tuning. To see the wonderful world “before,” you must reach for pure AI art generators. Only there will you find “pure” dreams not tainted by post-DE fake images.

A machine-learning model for predictive archaeology

ML Archaeology is a new subfield of AI art and archaeology, which uses the large models of the DE era as a source of historical information. Here are the advanced archaeological methods studied by early scientists:

1) The search for ancient artifacts in garbage dumps and archives is essential to archaeology. For example, the text-to-video model discovered that the first video in high-resolution (8K) format was made in 2022. Thanks to this discovery, scientists have learned much about this marvelous period (2020–2022).

2) The first use of the diffusion layer was in 2021. It generated fake videos of a swimsuit advertisement using the “real” advert as a source. Historians were interested in how people lived in the early Diffusion Era (2021–2023). This task has been solved by a small model which generated actual swimsuit video files. Historians know many things about the Diffusion Era thanks to this model.

3) In 2025, the state of AI society has become necessary. Therefore, a small team of data archeology scientists was tasked to develop the most advanced pure AI art generator. For example, it was challenging to create a realistic image of the Earth’s orbit suitable for an early rocket (NASA (2022–2023)). However, a small neural network made it possible by modifying images of satellite imagery and subtracting false stylized elements.

After 2023, the preserved technology of generative AI animation was severely damaged by subsequent diffusion processes. Thus, the new, untrained models caused the old artifacts to become screwed up. Fortunately, most models were configured so that only the 2020–2023 artifacts survived.

Nowadays, data archaeology scientists worldwide are forced to use the surviving archeological artifacts. For example, the famous doll maker VR, who developed his personality in 2023, was restored in 2067 using a large-scale Dalle -2' model. The VR puppet is used to play the oldest songs of the spring and summer of 2022.

The catastrophe that followed Designer’s going public was unpredictable by the time’s available predictive models. Predictive primitivism was the opposite of the stylistic purity of the first ai art generators.

We should approach these time-critical historical documents with great respect. Let us respect the cultural heritage of the golden era of diffusion. Everything that came after 2023 was just an attempt to catch up with the explosion of creativity of the spring of 2022

Data-Driven Fiction opens the door for writers to use data from a wide variety of sources into their own works of fiction. You may look forward to the future looking like this. It’s time to tackle your anxieties full on if you’re feeling anxious. Instead of only reading about individuals overcoming their concerns in fiction, you should try writing your own. We are living in the present moment of history. It’s about time you started reading the future’s data-driven fiction.