Announcements
We ıntegrate ınformatıon ın lıfe

  • DOLAR
    %-0,01
  • EURO
    %0,09
  • ALTIN
    %-1,22
  • BIST
    %0,92
Why Is ChatGPT Obsessed With Goblins?

Why Is ChatGPT Obsessed With Goblins?

OpenAI explained why ChatGPT keeps mentioning goblins. Here is the behind the scenes of this interesting situation in the artificial intelligence model.

ChatGPT users had been noticing for a while that the chatbot was unexpectedly talking about goblins, trolls, and similar mythological creatures. OpenAI shared the behind-the-scenes and reason for this interesting situation with an official statement.

The basis of this situation lies in a special personality mode developed for ChatGPT, which is now deprecated. The company satisfied users’ curiosity by detailing the reasons for this strange behavior.

The Origin of Goblin Obsession

OpenAI offers professional, efficient or toned personality modes for users to choose from. It was published for a while and especially nerdythat is, the personality mode defined as geek showed an extreme tendency towards these mythological creatures. The company stated that two-thirds of all uses of the word goblin during the period when this mode was active were due to this specific personality style.

During the development process, engineers unknowingly assigned high rewards to creature-related metaphors when creating this personality type. This led to an approximately 40-fold increase in the use of the word goblin between the GPT-5.2 and GPT-5.4 models. The geek personality mode referenced not only goblins but also other creatures such as ogres, trolls, and gremlins much more frequently than expected.

Restrictions with GPT-5.5

Similar behavior continued during the testing phase, as OpenAI started training the GPT-5.5 model before pinpointing the source of this issue.

In this process, it was observed that not only goblins, but also words such as raccoon and pigeon were used more than necessary. To correct this situation, the company added special restrictions to the system commands of the new model.

With the new update, the GPT-5.5 model has been instructed to use these words only when absolutely relevant. Engineers recognize that widely disseminated models can develop such unexpected behavioral features.

Fortunately, this has been recorded as a harmless quirk that does not pose any security risk to users. Have you ever come across such strange expressions or goblin references while using ChatGPT, or do you think this is over now?

Social Media Share:

TOGETHER FOR A LOOK

Can you share with us your comment?