Source of Endogenous Hallucination Problem and Path of Copyright Regulation in AIGC Large Model

As the engine of new quality productive forces, artificial intelligence is driving profound changes in the field of work creation. However, the output result of AIGC large model is influenced by the provider’s value concept, manual screening content and training input data, and its output content is often accompanied by the emergence of endogenous hallucination problems such as virtuality and bias. In order to solve the inherent hallucination of AIGC grand model and ensure the balance of interests between AIGC grand model and human authors, we should realize the main tone of value rationality leading the development of AIGC grand model and integrate the whole process of AIGC data training with copyright incentive mechanism at the level of theoretical amendment. At the same time, in terms of concrete measures, firstly, in order to solve the problem of algorithmic bias, the internal reasoning logic can be supervised through the output result of the work; Secondly, the burden of proof should be borne by the AIGC provider in the lawsuit to prove the compliance of its own hallucination-eliminating behavior; Finally, adhere to the guidance of competition policy to achieve the harmonious coexistence between market competition and science and technology.