A recent study published in the PNAS Nexus journal suggests the rise of AI models like ChatGPT may be challenging the dominance of traditional knowledge-sharing sites like Reddit and the programming forum Stack Overflow. This shift could impact the availability of freely accessible public information.
The research, conducted by Maria del Rio-Chanona and her colleagues, shows Stack Overflow saw a 25% decline in user activity within just six months of ChatGPT’s launch.
This drop was not observed on similar sites where ChatGPT access is restricted, highlighting the significant impact of the AI model’s rapid adoption. According to the study, users may be turning to AI-generated responses instead of seeking human-driven content, shifting how people obtain information online.
“LLMs are so powerful, have such a high value, and make a huge impact on the world. One begins to wonder about their future,” says Del Rio-Chanona, who is also an associate faculty member at the Complexity Science Hub (CSH). The findings raise concerns that a growing dependence on AI could reduce the number of contributions to public forums, leading to a shortage of diverse and authentic data needed to train future models. “This has quite big implications. This means there may not be enough public data to train models in the future,” she warns.
Python and JavaScript
The trend could disrupt the open web ecosystem, as AI tools like ChatGPT rely on publicly shared knowledge for training data. “Even AI models like ChatGPT are trained on human-generated content like Stack Overflow posts,” explains Johannes Wachs, a faculty member at CSH. Ironically, as AI displaces these platforms, the quality of training data may deteriorate over time.
The impact is especially pronounced in posts related to widely-used programming languages like Python and JavaScript, where activity has dropped significantly. The study suggests this shift is not limited to novices but affects users across all experience levels, indicating a broad move from public to private interactions on AI platforms.
With fewer people contributing to public platforms, AI models may eventually rely on lower-quality data, which could degrade their performance. The researchers call for a balanced approach that maintains the open exchange of knowledge while embracing AI advancements.
More from TechRadar Pro
Services Marketplace – Listings, Bookings & Reviews