# A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT

**Authors:** Yihan Cao, Siyu Li, Yixin Liu, Zhiling Yan, Yutong Dai, Philip S. Yu, Lichao Sun  
**arXiv:** [2303.04226](https://arxiv.org/abs/2303.04226)  
**Date:** 2023-03-07  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.04226v1/

## Key points

- AIGC (Artificial Intelligence Generated Content) has been rapidly advancing in recent years
- Techniques such as ChatGPT, DALL-E-2, and Codex have gained significant attention from society
- AIGC involves the creation of digital content through AI models, including images, music, and natural language
- The goal of AIGC is to make the content creation process more efficient and accessible by extracting intent information from instructions provided by humans and generating content according to this knowledge
- Large-scale models have become increasingly important in AIGC as they provide better intent extraction and improved generation results
- With the growth of data and model size, the distribution that the model can learn becomes more comprehensive and closer to reality, leading to more realistic and high-quality content generation
- This comprehensive survey provides a detailed review of the history of generative models in AIGC, including basic components and recent advances in both unimodal interaction (text and image generation) and multimodal interaction (cross-application between modalities)
- Existing open problems in AIGC include improving diversity in generated content, addressing ethical concerns related to bias in training data, and developing methods for evaluating generated content quality
- Potential benefits of AIGC for various industries such as entertainment, advertising, education healthcare are highlighted
- Continued research into improving these techniques while addressing ethical concerns is emphasized

## Summary

The field of Artificial Intelligence Generated Content (AIGC) has been rapidly advancing in recent years, with techniques such as ChatGPT, DALL-E-2, and Codex gaining significant attention from society. AIGC involves the creation of digital content through AI models, including images, music, and natural language. The goal of AIGC is to make the content creation process more efficient and accessible by extracting intent information from instructions provided by humans and generating content according to this knowledge. Large-scale models have become increasingly important in AIGC as they provide better intent extraction and improved generation results. With the growth of data and model size, the distribution that the model can learn becomes more comprehensive and closer to reality, leading to more realistic and high-quality content generation. This comprehensive survey provides a detailed review of the history of generative models in AIGC, including basic components and recent advances in both unimodal interaction (text and image generation) and multimodal interaction (cross-application between modalities). The authors discuss existing open problems in AIGC such as improving diversity in generated content, addressing ethical concerns related to bias in training data, and developing methods for evaluating generated content quality. Overall, this survey highlights the potential benefits of AIGC for various industries such as entertainment, advertising, education healthcare and more. It also emphasizes the need for continued research into improving these techniques while addressing ethical concerns to ensure responsible use.