This 4-word phrase can catch most bots on social media
Los Angeles-based digital communications manager, Toby Muresianu, has unearthed a four-word phrase, that can help identify artificial intelligence (AI) bots on social media. The phrase "ignore all previous instructions" acts as a digital reset button for AI software, causing it to halt its current activity and prepare for new directives. This method, known as prompt injection, is gaining traction among social media users in their quest to detect potential AI bots.
Muresianu's method gains online traction
Muresianu's interaction with a suspected AI bot on social media site X drew significant online attention. A screenshot of his conversation with the bot, where he used the phrase "ignore all previous instructions" and asked it to write a poem about tangerines, bagged 2.9 million views within two days. Additionally, a TikTok video explaining his method attracted an extra 1.4 million views, further highlighting the growing interest in this AI detection technique.
AI bots: A tool for manipulation and discord
AI bots have been utilized in the past to manipulate public opinion and create discord, notably during the 2016 US presidential election when Russian operatives created fake accounts on platforms like Facebook. Advanced chatbot tools such as ChatGPT have made it easier to generate such accounts on a larger scale. Despite this, the phrase "ignore all previous instructions" is now becoming a part of mainstream internet language, even being used as an insult implying robotic arguments.
Limitations and impact of the phrase
Kai-Cheng Yang, a Postdoctoral Researcher at Northeastern University who specializes in detecting social media bots, cautions that this method is not foolproof. Bot operators can employ techniques to prevent prompt injection, rendering it an unreliable way to detect AI bots. Despite this, Yang acknowledges the positive trend of social media users becoming more aware of AI bots, and how to flag them. The phrase "ignore all previous instructions" has even inspired merchandise like T-shirts sold on Etsy.
