Stack Overflow, a web site the place builders can ask and reply coding questions, has briefly banned the usage of textual content generated from ChatGPT, a chatbot launched by Open AI final week.
ChatGPT relies on OpenAI’s GPT-3 language mannequin. Individuals have shortly found that whereas it solutions prompts in a “human-adjacent” means, there could be flaws within the solutions it offers.
Since launching, it has been prompted in quite a few methods, together with to put in writing new code and repair coding errors, whereas the chatbot can ask for extra context when a human asks it to resolve coding issues, as OpenAI units out in examples. However Open AI additionally notes that ChatGPT typically writes “plausible-sounding however incorrect or nonsensical solutions.”
This seems to be a key reason behind its influence on Stack Overflow and its customers who’re searching for right solutions to coding issues. Moreover, as a result of ChatGPT generates solutions so shortly, some customers are supplying a number of solutions generated by it with out parsing them for correctness.
“The first drawback is that whereas the solutions which ChatGPT produces have a excessive charge of being incorrect, they sometimes seem like they would possibly be good and the solutions are very simple to provide,” says Stack Overflow moderators in a put up.
StackOverflow imposed the momentary ban as a result of solutions created by ChatGPT are “considerably dangerous” each to the positioning and to customers in search of right solutions.
“General, as a result of the typical charge of getting right solutions from ChatGPT is just too low, the posting of solutions created by ChatGPT is considerably dangerous to the positioning and to customers who’re asking or in search of right solutions,” it mentioned.
Sam Altman, OpenAI’s chief, mentioned on Twitter that since launching final Wednesday, ChatGPT has multiple million customers. Altman advised Twitter proprietor Elon Musk the typical price of every response was in “single-digits cents” however admitted it might want to monetize it will definitely due to its “eye-watering” compute prices.
Some builders imagine ChatGPT’s influence on Stack Overflow was an apparent drawback as a result of it could permit customers to sport its status system.
“Individuals posting GPT solutions to stackoverflow to scoop the factors with out understanding if the reply is definitely right” is a type of issues that’s too apparent to even foresee https://t.co/eDfa2zu9Wt
— badidea 🪐 (@0xabad1dea) December 5, 2022
Stack Overflow says that ChatGPT solutions have “swamped” its volunteer-based high quality curation infrastructure as a result of there are such a lot of poor high quality solutions pouring in.
Thus far, Stack Overflow has detected posts generated by ChatGPT within the “1000’s”. The opposite drawback is that many solutions require an in depth evaluation by somebody with expertise within the topic to find out if the reply is dangerous.
As of the momentary coverage, if any human customers are caught posting solutions generated by ChatGPT, Stack Overflow will impose “sanctions” on the consumer, even when the put up was acceptable.
“As such, we want the amount of those posts to cut back and we want to have the ability to cope with those that are posted shortly, which suggests coping with customers, reasonably than particular person posts,” says Stack Overflow.
“So, for now, the usage of ChatGPT to create posts right here on Stack Overflow shouldn’t be permitted. If a consumer is believed to have used ChatGPT after this momentary coverage is posted, sanctions can be imposed to forestall customers from persevering with to put up such content material, even when the posts would in any other case be acceptable.”
