Stack Overflow announced a temporary ban on answers generated by AI NLP model ChatGPT. The organization finds the average quality of ChatGPT-generated answers unacceptable.
Unfortunately, the accuracy of the average ChatGPT-generated answer is low. The answers can be harmful to the credibility of Stack Overflow and the developers that use the website to solve problems.
Some ChatGPT-generated answers are very credible. The texts can be quickly and easily generated. According to the website’s moderators, users who post ChatGPT-generated answers often don’t bother fact-checking their content. The high volume of AI-generated answers leaves volunteers with little to no time to review the accuracy of texts.
In addition to placing a ban on ChatGPT-generated answers, Stack Overflow will allow moderators to impose sanctions. Moderators that suspect a user of violating the platform’s policies can issue penalties — even if the user’s answers are correct. The specifics of these sanctions were not disclosed.
In the short term, Stack Overflow will assess what policies are needed to regulate ChatGPT and equivalent tools. These policies will be developed in collaboration with the platform’s administrators and users.