Lost Yangtze Sea is a hyper-text fiction built up by Generative AI tools. It is an experiment about using fake generated materials to challenge the real world. The aim of this work is to see whether one can mix up the real world facts with AI generated materials.

Based on a real archaeological research finding — that there is fossil evidence in the Yangtze River region of an ancient sea that existed 400 million years ago — ChatGPT and Midjourney have generated more new organisms and new archaeological discoveries in this ancient sea.

A fictitious archaeology conference website, also created by ChatGPT, has published these “cutting edge research findings”. Traces of these “new discoveries” can be found on various social media platforms such as Wiki, Quora, Zhihu, Twitter, and so on. See the conference here.

These fake topics, fake articles, and fake academic conferences were shared with top scholars and the public to see how they‘d react—and yes, we apologized afterward.

Then, we got serious about fixing the mess we made, and move this project into a further research. Using GPT again, we crafted a workflow that can real-time check the authenticity of everyday text, and we implemented this as a chatbot on platforms like Telegram. Our goal is to step into reality and help people figure out what’s real and what’s not in what they come across daily. See this further research: Post-Truth Validator.