Writing scientific papers with AI such as ChatGPT is prohibited at international conferences, but editing and polishing your own sentences is OK

The International Conference on Machine Learning (ICML), one of the international conferences on machine learning, has announced a policy of 'prohibiting the writing of scientific papers using AI such as ChatGPT .' Opinions among AI researchers are divided over this policy.

ICML 2023
https://icml.cc/Conferences/2023/llm-policy

ChatGPT and AI language tools banned by AI conference for writing papers - The Verge
https://www.theverge.com/2023/1/5/23540291/chatgpt-ai-writing-tool-banned-writing-academic-icml-paper

Natural language processing AI like ChatGPT has become so easy for anyone to use that it has caused confusion in many organizations, and some sites like ICML have even banned AI artifacts. In December 2022, the coding Q&A site Stack Overflow banned submission of ChatGPT-generated answers.

Conversation AI 'ChatGPT' answer posting is temporarily banned on Stack Overflow, a coding Q & A site-GIGAZINE

Also, the New York City Education Bureau has banned access to ChatGPT from school computers and networks.

New York City bans students and teachers from using ChatGPT - GIGAZINE

Under such circumstances, ICML says that although the rise of publicly accessible natural language processing models such as ChatGPT means exciting development, ``Who owns the output using such AI?'' 'Is the text and images I created new, or are they just derivatives of existing works?'

“These and many other questions will gradually be answered as AI at scale becomes more widely adopted. The ICML said in a statement that the AI ban will be re-evaluated in 2024.

However, what ICML prohibits is ``using text that is ``completely'' generated by AI in papers, and ``authors can edit or refine their own sentences. ” is not prohibited. This is because there are many cases where the text of the paper is checked using a tool that corrects the grammar with AI at the time of writing the article. In addition, we will not check papers by AI detection software, and will only investigate those that have been pointed out by other scholars.

“Unlike grammar correction tools, ChatGPT is not designed to adjust the structure or language of already written texts, but rather to create novel texts,” said Deb Raj, an AI research fellow at the Mozilla Foundation. It specializes in generating it, and it may cause a big change in the content of the sentence, 'he said, supporting the policy of ICML

On the other hand, Joav Goldberg, a professor at the University of Bar-Ilan in Israel, said, ``There is clearly an unconscious bias in peer review that favors papers written in plain English. favors native English speakers, as many non-natives seem to see a level playing field by using AI like ChatGPT to express their ideas. AI like ChatGPT not only helps researchers save time, but it may also facilitate communication with colleagues, ”he said, showing a positive attitude toward using ChatGPT. increase.

However, Professor Goldberg said, ``I think it is possible for scholars to use AI to generate a whole paper, but there is little incentive to actually do it. Even if a fake paper manages to pass peer review, Wrong descriptions will be the scholar's fault and will haunt the scholar's career all the time,' he said, saying that there is no merit in generating papers with ChatGPT.