Researchers warn that AI models trained on AI-generated data could spiral into unintelligible nonsense, making them less useful and filling the internet with incomprehensible babble. A study published in Nature found that as AI systems rely more heavily on their own output for training, they may create self-damaging feedback loops, leading to model collapse. This issue can be addressed by taking care when designing synthetic data and ensuring models are built to improve over time.

Source: https://www.livescience.com/technology/artificial-intelligence/ai-models-trained-on-ai-generated-data-could-spiral-into-unintelligible-nonsense-scientists-warn