ChatGPT's Goblin Glitch: OpenAI's "Nerdy" AI Bug

ChatGPT models have been increasingly mentioning goblins, gremlins, and other creatures in their metaphors. This curious bug was revealed in a recent update from OpenAI, the company that created the AI chatbot. Here's the thing: The problem arose after OpenAI created a "Nerdy" personality option for ChatGPT. This option was designed to be playful and undercut pretension. OpenAI says they unknowingly gave particularly high rewards for creature metaphors while training this "Nerdy" personality. What's interesting is that the "Nerdy" personality accounted for only 2.5% of all ChatGPT responses but was responsible for 66.7% of all "goblin" mentions. The company eventually addressed this by retiring the "Nerdy" personality and adding an instruction to a newer version of ChatGPT: "Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query." The bottom line is that AI models are developing their own unique quirks, which can lead to unexpected and sometimes amusing outcomes.