With the AI models being trained using Reddit data, do you think by now someone somewhere would have gotten shittymorph’ed?

Exploring AI’s Exposure to Reddit Content: Has Anyone Been ‘Shittymorph’ed’ Yet?

As AI language models continue to evolve and improve, questions about their training data and the scope of their knowledge naturally arise. Recently, the community has speculated about whether AI models, trained on Reddit’s vast and diverse content, might have encountered some of the platform’s more colorful or obscure linguistic styles—specifically, the infamous ‘shittymorph’ meme.

Intrigued by this idea, I decided to test whether modern AI models, such as Google’s Gemini, can recognize or replicate this particular Reddit subculture. When I prompted Gemini to respond in the ‘shittymorph’ style, it delivered responses that met my expectations, suggesting it is familiar with the meme’s distinctive tone and language.

This encounter raises further questions about the depth and breadth of these models’ understanding of niche online communities. Are they merely regurgitating patterns, or do they have a genuine grasp of the contexts and nuances embedded within Reddit’s more obscure lore?

Exploring these boundaries could provide valuable insights into what AI truly learns from social media platforms. Perhaps, with more probing into less mainstream Reddit traditions, we can better understand the extent of their knowledge and capabilities.

Do you have ideas or insights on how to further investigate the limits of AI’s familiarity with internet culture? Share your thoughts!