Last week, the AI ​​dialogue model launched at OpenAI ChatGPT When it became popular all over the Internet, Stack Overflow announceda temporary rule: Do not use ChatGPT generated content to answer questions on Stack Overflow. Today, Stack Overflow has officially updated this rule to its latest policy.

And explained why the release of answers generated by GPT and ChatGPT is currently not accepted:

Stack Overflow is a community built on trust. The community trusts that the answers submitted by users reflect accurate information that they actually know, and that they and their peers have the knowledge and skills to verify and validate those answers. The system relies on users to use the tools we provide to verify and acknowledge other users’ contributions, including responsible use of upvotes and downvotes.

Currently, contributions generated by GPT generally do not meet these criteria and thus do not contribute to a trustworthy environment. When a user copies and pastes information into an answer without verifying that the answer provided by GPT is correct, without ensuring that the sources used in the answer are properly cited (GPT does not provide this service), and without verifying that the answer provided by GPT is That trust is broken when the questions asked are answered clearly and concisely.

The objectivity of content on Stack Overflow means that if any part of an answer is wrong, then the answer is objectively wrong. In order for Stack Overflow to maintain its strong standard as a reliable source of correct and verified information, such answers must be edited or replaced. However, because GPT is sufficient to convince website users that the answers have value, the signals that communities typically use to determine the legitimacy of their peers’ contributions often fail to detect serious problems with GPT-generated answers. As a result, objectively wrong information entered the site. In its current state, GPT may undermine readers’ trust in our site to provide answers written by subject matter experts.

In light of this, the policy states that moderators reserve the right (at their discretion) to suspend the accounts of users who copy and paste GPT content onto their sites for up to 30 days without prior notice or warning.

A popular comment on Reddit also pointed out that he had used ChatGPT to answer some C++ technical questions, it also gave very clear answers quickly and solved his problem exactly as he hoped. But unfortunately, it later turned out that ChatGPT’s answer, although very clear, was also completely wrong.

“So I’m glad it was banned from StackOverflow. I can imagine it would quickly attract a lot of upvotes and eventual acceptance for its clear and authoritative writing style – but it cannot be trusted.”