1 of 1
Story summary
- In the United States, ChatGPT users report a surge in responses containing Arabic text, with examples in recipe lists and prompts shared on social media.
- The phenomenon is attributed to ChatGPT's token-based language processing, which can favor shorter foreign words.
- Some users call the episodes glitches, while others question whether they are random errors.
- OpenAI has previously addressed similar language issues, but recent announcements have not clarified the current situation.
