ChatGPT loses its marbles, talks gibberish, and worries one user by saying it's 'in the room'

ChatGPT appears to have misfired in spectacular fashion at times since yesterday, and although the gremlins in the works may now have been smoothed over by OpenAI, an official investigation into what caused the problems hasn't yet been concluded.

So, what did go wrong? Well, as you can see in the above post on X (formerly Twitter), the AI started talking complete nonsense, sprinkling in Spanglish, and generally acting highly erratically. There are quite a few reports of this kind of behavior, too.

This is the sort of thing we've seen from AI models before, but usually close to launch, when the products are only just out of testing, and still being honed.

ChatGPT is a veteran chatbot - well, in terms of the length of time that large language model-powered AIs have been in existence - so really, we wouldn't expect it to be having meltdowns like those in evidence across various social media forums right now.

One of the more worrying episodes here, as highlighted by The Independent, is that ChatGPT suggested it was in the room with a user.

This was a person who requested help with a coding problem, which was met by a lengthy and waffling response from ChatGPT, one that wasn't coherent and included the chatbot mentioning: "Let's keep the line as if AI in the room."

With this happening late at night (at 2am we're told), the ChatGPT user was suitably spooked, as anyone would be, frankly.

Unable to resist, there are some fake ChatGPT responses floating around, like the above one - which is actually a quote from the book 'I Have No Mouth & I Must Scream.' It's almost believable, though, given the other (real - well, presumably) responses cited on Reddit, X, and elsewhere.

There's clearly something going on, and it's not good, as OpenAI acknowledged the bug in ChatGPT yesterday, and said it was investigating reports of 'unexpected responses' from its chatbot.

We were quickly told that the issue was identified and a fix was being applied - that happened within 10 minutes of flagging up that OpenAI was investigating, in fact. The latest status report says that the company is continuing to monitor the situation, but doesn't actually mention that it's resolved.

Presumably we'll hear a final clarification that the bug, or whatever it was, has been fully remedied soon enough. There don't seem to be any further recent reports of ChatGPT going badly awry.

Considering Google has just launched its new AI model, Gemini - which replaces Bard, and spans multiple platforms and offerings - which is busy making a big splash, this isn't a good time for OpenAI's ChatGPT to be misbehaving.

It won't be the last time an AI goes wrong, of course, which is the worry as we move further into the future, as the stakes of what artificial intelligence is doing get increasingly loftier.