With the rapid spread of large-scale modelling tools, the cost of bulk manufacturing waste content and robotic interaction continues to decline. Reddit states that the Platform has used large models to identify more hidden anomalies and synergetic behaviour in order to reduce user access to garbage information.
Platform disclosure of intercept data
Reddit writes in a blog post that the platform currently intercepts about 23 million garbage content visits every day and finds about 25,000 new garbage posts and comments. These tools are used primarily to identify false interactions and man-made patterns that are more difficult to detect in old systems.
According to the company, between January and March of this year, there was a 20 per cent decrease in user exposure to waste content compared to the previous three months. This indicates that the large model has been used for governance scenarios such as content clearance and abnormal behaviour identification.
AI content increases the burden on the platform
Waste information is not a new issue, but the generation AI lowers the content production threshold and enhances the capacity to disguise. For social platforms, it is more difficult for traditional rule systems to cover such decentralized, repetitive and synergistic elements.
Reddit ' s approach reflects a more direct change: AI has, on the one hand, amplified content governance pressures and, on the other hand, has become a tool for the Platform to address such issues. The platform is using similar technologies to counter content pollution driven by the same technologies.
Auto-censorship still requires manual cooperation.
TechCrunch mentions that YouTube, Meta and Instagram allow users to publish AI-generated content, but usually require an apostille; TikTok also provides options for users to adjust how much AI-generated content they want to see.
It was also noted that faster identification of AI-generated content meant that the platform could also more quickly mark information about violations such as hate speech. However, the Platform ' s governance researchers have long argued that automated audits still need to be combined with manual audits to improve accuracy and performance.