Exploring AI Knowledge Through Reddit’s Hidden Layers
In recent discussions about AI training datasets, a recurring question has emerged: given that many AI models are trained on Reddit data, is it inevitable that some of these models have encountered the infamous “Shittymorph” style content? Has someone, somewhere along the line, been “Shittymorphed” by these algorithms?
This curiosity led me to conduct a simple test. I prompted the AI model Gemini to respond in the distinctive “Shittymorph” style—an intentionally exaggerated, chaotic tone associated with certain Reddit subcultures. The result? It did not disappoint. The AI successfully adopted the tone, demonstrating that it’s familiar with the style and, by extension, the underlying Reddit language patterns.
This experiment raises an intriguing possibility: by exploring more niche and obscure corners of Reddit, could we better understand the scope of what AI models have absorbed? Perhaps these hidden subcultures and specialized memes can serve as a kind of litmus test, revealing the depth of an AI’s familiarity with various online communities.
If you have ideas or suggestions on how to probe these hidden layers further, I’d love to hear them. Unlocking the extent of AI knowledge embedded in the depths of Reddit could open up fascinating avenues for research—both for understanding AI and for preserving the nuances of online culture.
Leave a Reply