Considering that AI models are developed using Reddit data, do you believe someone has already experienced being shittymorphed?

Exploring AI’s Exposure to Reddit’s Hidden Depths: Has the Taint of Subculture Reached Machine Learning Models?

As artificial intelligence models are increasingly trained on diverse online data sources, many experts and enthusiasts alike wonder about the extent to which these models have absorbed the quirks, slang, and subcultures found across platforms like Reddit. Given Reddit’s vast and eclectic content, it’s plausible to consider whether AI systems have encountered and internalized some of its more niche or edgy meme cultures—sometimes referred to colloquially as “shittymorph” styles.

Recently, I pondered this very question: has any AI model, during its training, ever been exposed to or influenced by the more colorful, less polished Reddit subcultures? Intrigued, I decided to conduct a small experiment by asking an AI assistant—specifically Google’s Gemini—to generate a response mimicking this particular stylized language. The results were quite revealing and seemed to confirm that these models do have at least some awareness of Reddit’s more obscure jargon.

This raises further questions about the depth of AI’s understanding of internet subcultures. How far into the rabbit hole can we go before the models start mimicking or understanding the nuances of these communities? Could delving into rarer or more esoteric Reddit lore help us gauge the true breadth of a model’s knowledge and contextual awareness?

Exploring these boundaries not only reveals the capabilities of current AI models but also helps us understand their cultural integration and potential biases. If you have ideas or insights into how we might examine AI’s exposure to the hidden corners of Reddit, I’d love to hear your thoughts.