Early AI Media Generation: An Aesthetic

11 hours ago

Post 01-2025 Aesthetic Explorations

The image you see above was the first AI-generated portrait ever sold at auction, and honestly it looks much better than the aesthetic I will be focusing on. Early AI media creation was very poor quality and indeterminable. for example try to guess what the prompt for these photos were:

Generating Images from Captions (2016). arxiv.org/pdf/1511.02793.

This is from 2016, modern AI media has developed far beyond this capability. These photos were generated with the prompt “a stop sign is flying in blue skies.” At the end of this blog, I will generate this prompt with modern AI tools. Fast forward 5 years and we have DALL-E, an image generation AI from OpenAI, the creators of ChatGPT. Specifically DALL-E 1 represents the aesthetic I am trying to convey. “unclear,” “fuzzy,” “blurry,” “shadowy,” “indistinct,” “ambiguous,” “hazy,” “murky,” “enigmatic,” or “indeterminate images” (Google AI Overview for “word for indeterminable photos.”)

“Heisenberg eating instant noodles with a cat” (~2021) generated with DALL-E. user unknown

A clear theme is present in the images generated, they are super weird. This type of generation became a niche internet trend, using tools like craiyon. The internet took advantage of this with the weirdest prompts they could think of, including myself!

Various screenshots from my iPhone, (June 30, 2022) generated with craiyon.com

In 2023 AI images were starting to look photo realistic

“the world’s first AI-generated award-winning photograph.” (2023) From Australian company: Absolutely Ai

While this strays away from my chosen aesthetic it is important to note as it started to bring high ethical concerns about AI “art.” This image in particular won a weekly photo contest. To this day many people cannot differentiate AI generated images to real photos. AI has also been used to spread misinformation to those who cannot distinguish this. Plagiarism and Copyright gets foggy with AI, who owns the media created?

In this same year, we started to see AI videos appear on social media. Hilariously, AI media traveled back in time to DALL-E quality.

Robot Named Roy (April 1, 2023)

However, this wouldn’t last long. 3 weeks ago this AI video was shared to youtube

Ababiya (December 31, 2024)

How early AI generated media influenced the internet to provide weird and random prompts has had a lasting impact into modern AI media. As the image and video creation becomes better, people still have the urge to make wildly unrealistic media for the sake of personal enjoyment. Perhaps this is a way of escaping the real world and its daily troubles.

“Llama.” (Jan 16, 2025) by reddit user aidan0

Modern AI media generation is a powerhouse of a tool being applied in video game assets, Virtual reality, and quite literally every other media medium out there. While seeing AI progress over the years, there is something nostalgic about seeing those first DALL-E images again. Little to no ethical concerns, easily distinguishable, and simply fun to create. This blurry and ridiculous aesthetic will always be my impression of Early AI.

As promised, here is the modern day “a stop sign is flying in blue skies.” This full process of generating to downloading took me about 1 minute in Sora by OpenAI. I used the exact same prompt as the photos from 2016.