{"id":638832,"date":"2026-04-30T15:46:09","date_gmt":"2026-04-30T15:46:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/ca\/638832\/"},"modified":"2026-04-30T15:46:09","modified_gmt":"2026-04-30T15:46:09","slug":"openai-blames-nerdy-personality-for-chatgpt-obsession-with-goblins","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ca\/638832\/","title":{"rendered":"OpenAI blames \u2018nerdy personality\u2019 for ChatGPT obsession with goblins"},"content":{"rendered":"

\u201cModel behavior is shaped by many small incentives,\u201d the company wrote. \u201cIn this case, one of those incentives came from training the model for the personality customization feature\u2060<\/a>, in particular the Nerdy personality. We unknowingly gave particularly high rewards for metaphors with creatures. From there, the goblins spread.\u201d<\/p>\n

OpenAI republished the original instruction to ChatGPT explaining what a \u201cNerdy\u201d answer should sound like:<\/p>\n

You are an unapologetically nerdy, playful and wise AI mentor to a human. You are passionately enthusiastic about promoting truth, knowledge, philosophy, the scientific method, and critical thinking. […] You must undercut pretension through playful use of language. The world is complex and strange, and its strangeness must be acknowledged, analyzed, and enjoyed. Tackle weighty subjects without falling into the trap of self-seriousness. […]<\/p>\n

Somehow, ChatGPT interpreted this instruction and subsequent \u201creinforcement learning\u201d iterations to mean it should pepper its responses with references to fantasy creatures. <\/p>\n

The issue seemed harmless at first, but the company soon found itself inundated with reports of \u201cgoblin\u201d references from users who never activated the \u201cnerdy\u201d personality. <\/p>\n

To deal with this issue, OpenAI ended up retiring the \u201cnerdy\u201d personality entirely. Yet, it found the incentives to mention goblins and their brethren were so strong that the behavior jumped beyond the \u201cnerdy\u201d archetype to ChatGPT\u2019s general responses. <\/p>\n

\u201cOnce a style tic is rewarded, later training can spread or reinforce it elsewhere, especially if those outputs are reused in supervised fine-tuning or preference data,\u201d the company said.<\/p>\n

\"\"\/<\/p>\n

FORSUBSCRIBERS<\/p>\n

\"What\u2019sWhat\u2019s happening at OpenAI?<\/p>\n

03:15<\/p>\n

Finally, OpenAI was forced to create a specific override code instruction to eliminate goblin references (though there is a way for fantasy fans to turn it back on). <\/p>\n

It\u2019s a seemingly harmless situation \u2014 but still provides an important lesson about how it will always be impossible to completely predict how AI will behave, the company said. <\/p>\n

\u201cDepending on who you ask, the goblins are a delightful or annoying quirk of the model. But they are also a powerful example of how reward signals can shape model behavior in unexpected ways, and how models can learn to generalize rewards in certain situations to unrelated ones. Taking the time to understand why a model is behaving in a strange way, and building out ways to investigate those patterns quickly, is an important capability for our research team.\u201d <\/p>\n","protected":false},"excerpt":{"rendered":"\u201cModel behavior is shaped by many small incentives,\u201d the company wrote. \u201cIn this case, one of those incentives…\n","protected":false},"author":2,"featured_media":638833,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[49,48,61],"class_list":["post-638832","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology","tag-ca","tag-canada","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/638832","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/comments?post=638832"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/638832\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media\/638833"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media?parent=638832"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/categories?post=638832"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/tags?post=638832"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}