The rise of AI language models has sparked concerns that the Internet will be flooded with automated spam. While such fears have not materialized yet, early signs indicate that bots powered by tools like ChatGPT are generating fake reviews and low-quality filler content. A shibboleth for machine learning spam is the phrase “as an AI language model,” which is often used by ChatGPT as a disclaimer. The phrase has become a telltale sign of machine learning spam, revealing the presence of bots that generate inflammatory content to boost engagement with certain audiences.
Examples of bots generating opinions about high-profile figures and gossip about trending crypto influencers and publications have been observed on Twitter. Some messages even appear to be criticism of the bot’s operator, who may have been asking the system to generate negative and hateful tweets. The phrase “as an AI language model” can be searched on any site with user reviews or a comment section, exposing the presence of bots.
Amazon is one site where the phrase shows up in fake user reviews. In some cases, the system used to generate fake reviews is transparent about its deception, stating that it hasn't used the product personally but can give it a five-star rating based on its features and customer reviews. Elsewhere on Amazon, the phrase appears in real reviews. One of them is presented below.
Variants of the phrase emerge in various contexts, such as on the website of a Finnish electronics store that used AI to translate English-language products into Finnish. On an influencer marketing agency's website, the phrase appears in a blog post title. It even shows up in a directory of malls in Qatar and in a user profile on the freelancer platform Upwork.
In conclusion, while these examples demonstrate the use of AI to generate spam and low-quality text, the extent of the practice and its impact on online ecosystems remain uncertain. The text that can't be easily detected is the real challenge, as software to detect AI-generated text is unreal and may be mathematically impossible. Obsession over machine learning fakery is so widespread that real people now fall under the slander of being AI. Detecting such fakes may become increasingly difficult in the future, making it challenging to combat AI-generated spam.
You can read our previous news such as Google’s chatbot Bard which can assist programers to write and debug code.
Also, we wrote about Snap’s My AI chatbot that is now available absolutely free.
Besides, our previous report reveals Musk’s thoughts to build TruthGPT.