The Internet is surprised with ChatGPT, an OpenAI text generation model based on GPT-3.5 that is perceived, nothing more and nothing less, as a tool that could become the definitive “Google Killer”. Despite its apparent advantages, not everyone embraces ChatGPT with the same enthusiasm. Stack Overflow just banned it.
Stack Overflow is a very valuable tool for professional developers. Each month, more than 100 million people ask questions and answers about a wide range of programming topics. Now, sin embargo, many answers are not being generated by humans, only from the latest OpenAI AI model and, unfortunately, some of them are incorrect.
Stack Overflow says goodbye to ChatGPT
According to a statement from the platform owned by Prosus, the main problem with ChatGPT is that, in general, the incorrect answers that he gives are plausible, es decir, parecen ser buenas respuestas. In this sense, Stack Overflow has reached the conclusion that this mechanism “is substantially harmful for the site and for the users”.
They also point out that many users with experience or the will to verify the information are using ChatGPT to create answers quickly. Certainly, ChatGPT is much faster to elaborate a response than a human, but they do not always have the same quality. Product of this, from the volume of responses has grown substantially putting in jaque from the work of the moderators.
Although algorithms are playing an increasingly prominent role in content curation and moderation systems, in Stack Overflow there are tasks that are mainly carried out for volunteers with a certain level of knowledge que, después de una lectura detenida, determine if the answers are valid. The volunteers, at this moment, are not enough with the volume of responses they are receiving.
As The Verge points out, the phenomenon of ChatGPT puts an additional concern on the table. If so far the large-scale language models have not generated large problems, many fear that the pattern of generating an avalanche of content, with a percentage of false but plausible data, can be replicated on other platforms, putting their systems to the test moderation
In the case of Stack Overflow, as mentioned above, it has just prohibited the use of ChatGPT. If the platform detects that a user has published answers using this experimental tool of OpenAI after the publication of this new content policy, they say that will impose sanctions, including if the publication has an “acceptable” explanation.
It should be noted that Stack Overflow has not closed the door definitively to ChatGPT. The platform indicates that the prohibition has a temporary character. Before making a final decision, they say they will consult with the community. It is time to wait to know what the position of the users will be and finally ChatGPT will have a place within the platform.
Images: Arif Riyanto
In Xataka: Ya hemos visto cómo una IA aging and rejuvenating. Disney has gone one step further