通过概率工程引导生成模型适度幻觉，反而能提升效果。
Engineering of Hallucination in Generative AI: It's not a Bug, it's a Feature
- 用概率工程控制生成过程，主动引入可控幻觉。
- 适度幻觉能让模型输出更符合人类期望的结果。
- 适合想提升生成质量的研究者与开发者参考。
生成式人工智能正以惊人速度融入生活。大型语言模型如ChatGPT可回答问题或撰写文本，大型视觉模型如GAIA-1能根据文字描述生成视频或续写视频。这些神经网络模型严格依据训练数据中的真实内容进行学习。然而一个令人意外的现象是：当允许模型具备一定程度的幻想（幻觉）时，其表现反而更佳。尽管幻觉在生成式AI中通常被视为负面缺陷——毕竟人们期望ChatGPT给出基于事实的回答！——本文回顾了若干简单的概率工程方法，可通过适度诱导幻觉，使模型达成理想输出。我们不得不反思：生成式AI中的幻觉，或许并非缺陷，而是一种特性。
原文摘要 · Abstract (English)
Generative artificial intelligence (AI) is conquering our lives at lightning speed. Large language models such as ChatGPT answer our questions or write texts for us, large computer vision models such as GAIA-1 generate videos on the basis of text descriptions or continue prompted videos. These neural network models are trained using large amounts of text or video data, strictly according to the real data employed in training. However, there is a surprising observation: When we use these models, they only function satisfactorily when they are allowed a certain degree of fantasy (hallucination). While hallucination usually has a negative connotation in generative AI - after all, ChatGPT is expected to give a fact-based answer! - this article recapitulates some simple means of probability engineering that can be used to encourage generative AI to hallucinate to a limited extent and thus lead to the desired results. We have to ask ourselves: Is hallucination in gen-erative AI probably not a bug, but rather a feature?
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。