ChatGPT returns gibberish for hours, users delight in the chaos

OpenAI's ChatGPT had its first serious meltdown Tuesday, returning incoherent and nonsensical responses to user prompts for hours. And like most celebrity pop stars who go off the rails, social media users could not get enough of its downward spiral.

Observations about ChatGPT's unexpected public performance, some called a 'peculiar poetry in motion,' were first reported by users around 3 p.m. PST.

#ChatGPTgonewild

“GPT-4 is returning rubbish data,” one user wrote in the OpenAI Developer Forum. Another labeled the chatbot’s responses– which at times included random emojis and song-like lyrics – “garbage words after every answer.”

Some users said the artificially generated bot was even returning prose in both Spanish and English, otherwise known as ‘Spanglish.’

“Every prompt I now give it returns something of a real answer and then it just digresses into nonsense and some kind of weird slang usually followed with a few emojis as well. Even asking it “what’s up with chatgpt today” and the discussion is called ‘Problemas Comunes’ (?),” Richardagreen said about ChatGPT's full-on Britney moment.

X user @seanw_m posted multiple screenshots showing paragraphs of strange and unintelligible responses coming from the beloved AI chatbot.

The initial X post received over 20 million views and over 2.7K comments by Wednesday morning. “Chatgpt is apparently going off the rails right now and no one can explain why,” it said.

Fun aside, AI experts warn this kind of ‘malfunction’ can only be expected in such novel technology, and signals a warning to companies they must learn to walk with AI before they can run.

“This unpredictable behavior from a leading Generative AI model is a stark reminder to businesses integrating AI that the technology is still new to the market and problems are likely,” said Chris Eastham, Partner and AI specialist at Fieldfisher law firm in London.

Any publicity is good publicity?

ChatGPT’s handler, em I mean creator, the Microsoft-backed OpenAI, tried to control the damage by making an official statement addressing the bizarre behavior on the community developer site soon after. At about 3:40 p.m. PST, Tuesday, OpenAI announced it was “investigating reports of unexpected responses from ChatGPT.”

Like any good crisis management team, barely seven minutes later OpenAI said “the issue had been identified and is being remediated now.”

To satisfy those who complained the fix wasn’t rolling out fast enough, OpenAI simply stated, it was “continuing to monitor the situation.”

OpenAI announced it had resolved the bug by 8:14 a.m. PST, Wednesday – yet, did not reveal exactly what the “issue” was that caused the ChatGPT global calamity.

Eastham points out that ChatGPT users, especially those adopting large language models (LLM) to support business operations, “need to accept that anomalies are part and parcel of using AI applications.”

“It remains critical to put careful controls such as human oversight in place to avoid issues like this disrupting business operations and impacting customers,” he said.

ChatGPT's 'gobbledegook'

A public crash-and-burn moment typically evokes all kinds of fodder; some users welcomed the wild responses, others made fun at ChatGPT’s expense, and some even sympathized with its plight.

“Just saying, these are really interesting and useful outputs I’m getting and it would be nice to see a feature where we could Toggle button for more unexpected responses,” said forum user moleculardrugs.

“I’m in approval of these responses and they seem to be a nice addition (if they were ever added as such) to ChatGPT,” they said.

X user @promisebender seemed to agree and posted, “ I’ve come around. OAI should make this a permanent feature."

The comments on Reddit were no less snarky. “About halfway down the second response, GPT wants to play footsies?" Ar92ldm noted along with a screenshot of ChatGPT's self-proclaimed "goobledegook."

Of course, there are always some in it for the doom and gloom.

“What if it just became sentient and this is how the world ends,” pondered Reddit user l_think_therefore_I.

The artificially intelligent chatbot was launched by OpnAI and its CEO, Sam Altman, in November 2022. ChatGPT gets approximately 1.5 billion visits per month, according to the most recent numbers by Exploding Topics.