Generative artificial intelligence (AI) systems are trained on large data
corpora to generate new pieces of text, images, videos, and other media. There
is growing concern that such systems may infringe on the copyright interests of
training data contributors. To address the copyright challenges of generative
AI, we propose a framework that compensates copyright owners proportionally to
their contributions to the creation of AI-generated content. The metric for
contributions is quantitatively determined by leveraging the probabilistic
nature of modern generative AI models and using techniques from cooperative
game theory in economics. This framework enables a platform where AI developers
benefit from access to high-quality training data, thus improving model
performance. Meanwhile, copyright owners receive fair compensation, driving the
continued provision of relevant data for generative model training. Experiments
demonstrate that our framework successfully identifies the most relevant data
sources used in artwork generation, ensuring a fair and interpretable
distribution of revenues among copyright owners.

一次训练中的数据 Shapley

Data Shapley in One Training Run

We study the problem of distributing a set of indivisible items among agents
with additive valuations in a $\mathit{fair}$ manner. The fairness notion under
consideration is Envy-freeness up to any item (EFX). Despite significant
efforts by many researchers for several years, the existence of EFX allocations
has not been settled beyond the simple case of two agents. In this paper, we
show constructively that an EFX allocation always exists for three agents.
Furthermore, we falsify the conjecture by Caragiannis et al. by showing an
instance with three agents for which there is a partial EFX allocation (some
items are not allocated) with higher Nash welfare than that of any complete EFX
allocation.

本文研究在具有加性估值的代理之间以公平的方式分配一组不可分割的物品的问题，通过 EFx 公平的概念，我们证明了对于三个代理，总是存在一个 EFx 分配，同时证实了 Caragiannis 等人的猜想，并展示出一个实例，其中对于具有三个代理的部分 EFx 分配（某些物品未分配），其 Nash welfare 比任何完全 EFx 分配的 Nash welfare 要高。