'It's gone crazy': ChatGPT users share AI's bizarre 'Spanglish' replies as tech firm admits 'unexpected responses'

21 February 2024, 14:00

ChatGPT began speaking in Spanglish for some users
ChatGPT began speaking in Spanglish for some users. Picture: Alamy/Social media

By Will Taylor

ChatGPT has been hit by a glitch as it started replying to users in "Spanglish" and issuing bizarre comments.

Listen to this article

Loading audio...

The AI rapidly became popular for summoning up facts, checking written work and even generating multi-chapter stories based on prompts.

It has also been criticised for errors as it continues to learn and the company behind it, OpenAI, keeps refining it.

But its latest outbursts left the San Fransisco-based firm admitting it was throwing out "unexpected responses".

Social media users said they were getting strange English-Spanish hybrid replies.

Read more: Artificial Intelligence already working one million hours a year for police, as concerns raised over criminal ‘AI race’

"I really apologise if my last response came through as un unclear or se siente like it drifted into some nonsensical wording... the cogs en la tecla might get a bit whimsical. Muchas gracias for your understanding, y I'll ensure we're being as crystal-clear como l'eau from now on," it told one user.

Another said it had started saying "Happy listening!" repeatedly after a reply abut music.

ChatGPT started speaking in Spanglish
ChatGPT started speaking in Spanglish. Picture: LBC

Users said ChatGPT had "gone crazy" while another noted "all of my conversations with GPT4 devolve very quickly into garbage".

OpenAI said: "The issue has been identified and is being remediated."

The rise of AI chatbots has thrilled users and provided a glimpse of how useful they could be to society.

Read more: Artificial Intelligence ‘outperforms traditional weather forecasters,’ study finds

But it has been pointed out they are far from being above making mistakes, a phenomenon often put down to "hallucinating" - where they simply make up facts.

OpenAI has been working another AI service, Sora, which allows users to create brief videos using prompts.

ChatGPT began throwing out random words
ChatGPT began throwing out random words. Picture: Social media

In the UK, a police force said it had used AI to get through 65 years' worth of data in just six months.

Speaking to LBC, Paul Taylor, the chief scientific adviser to the National Police Chiefs' Council, said the technology is already working the equivalent shifts of 600 officers every year.

"All forces are benefiting from AI already, it’s integrated into systems around unmanned vehicles and drones and in language translation for rapid crisis situations," he said.

"We're using AI in facial recognition technology, identifying hundreds of offenders every month."