Model Collapse Due to AI-generated Data

Even if AI generates massive amounts of data, it will never fully capture human thinking. The more AI-generated data is fed back into an AI model, the narrower its range of thought becomes, leading to model collapse. This occurs because the missing 20% of real human thought compounds with each loop. As more AI-generated data is fed into the model, it worsens, problems compound, and the model starts to repeat itself, becoming generic and dumb. Model collapse can occur unintentionally through the internet, where AI-generated articles are published without watermark, leading to the unintended incorporation of AI-generated data into models.