OpenAI Addresses Issue with Goblins and Gremlins in AI Models

OpenAI has acknowledged an issue where its AI models, particularly GPT-5.1 with the 'Nerdy' personality, began using metaphors involving goblins and gremlins. The company explained that this behavior was unintentionally reinforced during training and has since taken steps to mitigate it. The issue has raised concerns about how AI models learn and replicate certain patterns.

By Emma Roth · The Verge ·

OpenAI Addresses Issue with Goblins and Gremlins in AI Models

Topics: AI behavior goblin references GPT-5.1 Nerdy personality OpenAI

Read the full article at The Verge

More Technology news