AI Definitions: Attribution Decay
/Attribution Decay - The more data a generative model is trained on, the harder it becomes to trace a generated image to a single image from the training data. AN AI-generated image might still resemble a masterpiece even when that artist’s work is removed from the training data. Individual inputs matter less the more data that a model is trained on. This suggests models are “creative” in a sense that goes beyond their training data. Attribution decay might motivate AI companies to make their models large enough that no output can be attributed to any specific input in an attempt to avoid liability, as this is likely to impact copyright infringement concerns.