Structure-Aware Diversity Pursuit as an AI Safety Strategy against Homogenization
For AI safety researchers, this paper introduces a conceptual framework to address homogenization, but it is purely positional with no empirical validation.
The paper argues that homogenization (loss of diversity in generative AI outputs) is a key AI safety concern and proposes xeno-reproduction, formalized as structure-aware diversity pursuit for auto-regressive LLMs, as a mitigation strategy. No concrete results are provided.
Generative AI models reproduce the biases in the training data and can further amplify them through mode collapse. We refer to the resulting harmful loss of diversity as homogenization. Our position is that homogenization should be a primary concern in AI safety. We introduce xeno-reproduction as the strategy that mitigates homogenization. For auto-regressive LLMs, we formalize xeno-reproduction as a structure-aware diversity pursuit. Our contribution is foundational, intended to open an essential line of research and invite collaboration to advance diversity.