With the rise of easily accessible large language models (LLMs), the problem of spam on the internet has intensified. Anyone familiar with online activities over the past few years will have noticed an increase in spam and automated content.
In response, Reddit has created tools that utilize LLMs to tackle the spam issue, notably stemming from similar models. While it seems rather ironic, the reality is that platforms now must employ advanced technology to combat these challenges. Reddit reports that it blocks approximately 23 million spam views daily and identifies around 25,000 new spam posts and comments every day.
Though social media networks have been implementing automated tools to manage spam for many years, Reddit asserts that its enhanced tools are now achieving better results.
According to a blog post from Reddit, “We make use of LLMs to identify the intricate, coordinated patterns of fake interactions and artificial excitement that previous systems often overlooked.” The company states that, between January and March, it managed to decrease user exposure to spam by 20% compared to the preceding three months.
Various platforms, including YouTube, Meta, and Instagram, permit users to share AI-generated content as long as they label it, whereas TikTok goes further by allowing users to adjust their preferences regarding the amount of AI-generated content they encounter.
If platforms can more swiftly identify AI-generated content, they also have the capacity to flag harmful material, such as hate speech, more effectively. However, industry experts consistently emphasize the necessity of combining AI moderation with human oversight for optimal results.
