A man asked ChatGPT to count to one million and was shocked when the chatbot refused the request.
In a viral clip, the man can be heard making the simple request, but the chatbot seems to skirt around the question, refusing to comply.
This has led many to wonder why AI bots won’t perform the task and one YouTube video may have the answer.
Uploaded to social media by 7news, the clip explains: “Why can’t ChatGPT count to one million? More specifically, why won’t ChatGPT count to a million? It seems nonsensical at first, but then the assertiveness really gets stronger and stronger until eventually the chatbot seems to almost force a change in conversation, and for all the videos online, it seems that nobody’s really been able to get ChatGPT to count to one million.
Advert
“And it should be a simple task for something like artificial intelligence, so can ChatGPT count to one million? No.”
The YouTube video continues: “ChatGPT is more constrained by its own architecture than its actual ‘intelligence’. The models are designed to seek interaction and are motivated to keep the conversation flowing.
“A typical response when using the voice mode for ChatGPT is rarely longer than 15 seconds and if you were to count from one to one million at like two digits a second, it would take nearly six days, and for this reason ChatGPT can’t and ChatGPT won’t count to a million.
“It’s a good reminder that the capabilities of these AI agents at times can seem limitless but they will always be restricted by the motivations that were programmed into them at the start.
“The more and more we’re seeing this where tasks that should be simple for AIs end up hitting all of these awkward kind of roadblocks on the way to a solution, and if the guardrails were taken off then technically these LLMs would be able to do these impossible tasks.”
Many people took to the YouTube comment section to share their own reactions to the video, with one user writing: “It not that OpenAI’s: ChatGPT can’t but it won’t now, unless and until it’s fully capable.”
Another said: “Why doesn’t chatgpt explain exactly what you’ve said when ask them to count, instead of stalling and wheezing its way out of the task?”
And a third person added: “It’s obvious that it could. But it’s a waste of bandwidth and computational power and electricity. Especially if everyone did it.”
