OpenAI was in the spotlight this week with its incredibly impressive Sora text to video toolbut it appears that the allure of AI-generated video has led to the popular chatbot being sidelined, and it is now spiraling out of control.
Yes, ChatGPT has gone mad – or, more accurately, has briefly gone mad for a brief period sometime in the last 48 hours. Users have reported a large number of confusing and even threatening responses from the bot; Some saw it getting stuck in a loop of repetitive nonsensical text, while others were subjected to made-up words and strange monologues in broken Spanish. One user even stated that when asked about an encryption issue, ChatGPT responded with a puzzling statement that ended with the claim that it was “in the room” with them.
Naturally, I immediately checked the free version of ChatGPT and it now seems to be behaving again. It’s unclear at this time whether the issue was just with the paid GPT-4 model or with the free version as well, but OpenAI has acknowledged the issue and said that the “issue has been identified” and that its team is “continuing to monitor the situation . ”. However, it did not explain ChatGPT’s latest tantrum.
This isn’t the first time – and it won’t be the last time
ChatGPT has had a lot of problems in the past – when I tried to break it last year it said some pretty hilarious stuff – but this one seems to have been a bit more widespread and problematic than the chatbot nonsense of the past.
It’s a pertinent reminder that AI tools are generally not infallible. We saw it recently Air Canada was forced to issue a refund after its AI-powered chatbot invented its own policy, and it seems likely that we will only see more of these strange problems as AI is implemented into the various facets of our society. While these current ChatGPT issues are relatively harmless, there is a chance that real problems will arise. The Air Canada case feels disturbingly like a sign of things to come, and could set a real precedent for human moderation requirements when AI is deployed in business environments.
As for exactly Why ChatGPT had this little episode, speculation is currently rife. This is a completely different issue user complaints about a ‘dumber’ chatbot late last year, and some paying users of GPT-4 have suggested that this may be related to the ‘temperature’ of the bot.
To be clear, that’s not a literal term: When discussing chatbots, temperature refers to the degree of focus and creative control the AI exerts on the text it produces. A low temperature gives you direct, factual answers with little to no character behind them; a high temperature makes the bot come out of the closet and can result in more creative – and possibly weirder – responses.
Whatever the cause, it’s good to see that OpenAI seems to have a grip on ChatGPT again. These types of ‘chatbot hallucinations’ look bad for the company, given its status at the forefront of AI research, and threaten to undermine users’ confidence in the product. After all, who wants to use a chatbot that claims to live in your walls?