A recent interaction with ChatGPT has shocked the entire world, including meme lord and CEO of Tesla, xAI, Neuralink and SpaceX, Elon Musk. Based on the new GPT-5 model from OpenAI, the new version of ChatGPT is dealing with an inability to answer questions far different from what it used to do on the older models. Instead of hallucinating and coming up with incorrect answers, ChatGPT has now learnt to say ‘No’.
An isolated incident has seen ChatGPT 5 admitting that it didn’t know about something that was asked in a query. The remarkable moment, which saw the GPT-5 model declare “I don’t know” after 34 seconds of thinking, has been hailed as a significant leap forward for AI and its accountability.
In an era where humans are fighting against AI-driven misinformation, this development sheds a light of hope for fighting misinformation.
ChatGPT says ‘I don’t know’
It all began when a ChatGPT user, Kol Tregaskes, shared a screenshot showing the AI’s surprising response to a query. ChatGPT with the GPT-5 model underneath has given an honest and humble reply, “I don’t know – and I can’t reliably find out.” In stark contrast to the common hallucinations that have plagued earlier AI versions, this new way of avoiding incorrect answers seems to keep it transparent.
Since it was something unseen, AI enthusiasts couldn’t hold themselves from showering praises on the Internet and social media. Chief among the praisers was Elon Musk, who has been busy promoting the newest feature additions to the Grok-4 chatbot and engaging in an online feud with OpenAI CEO Sam Altman for getting his ChatGPT bot favoured over other AI tasks on the Apple App Store.
In a response to the original post, Musk said, “impressive”, thus lauding the AI chatbot for its ability to amidst mistakes instead of diverting the user to an incorrect response.
Note that OpenAI has apparently reduced the chances of incorrect responses to a mere 10 per cent in this latest version of GPT-5.
ChatGPT head warns of relying on the AI bot
Despite these advancements, OpenAI’s head of ChatGPT, Nick Turley, has warned users against relying on the AI model as their primary source of information. Turley says that until the model is “provably more reliable than a human expert on all domains,” it should be used as a source for a second opinion only, not a definitive authority.
A good example of this was seen in the recent case involving a Spanish content creator who missed out on her flight to Puerto Rico after ChatGPT failed to clarify the visa document requirements.
