Summary
- A viral new TikTok trend sees users attempting to persuade ChatGPT to count to one million.
- Try as users might, OpenAI’s flagship chatbot refuses to list off numbers.
- In all likelihood, this is a hardwired decision made by OpenAI to limit bandwidth, data use, and server overload.
A new viral trend has made its way onto the internet across short-form video platforms like TikTok and Reels: attempting to coerce ChatGPT to count to one million in sequential order. With the OpenAI-made chatbot refusing to comply with each and every attempt, hilarity inevitably ensues.
Whether in Voice mode or in written text mode, ChatGPT runs circles around users’ requests to count upward to one million, often counting for several seconds before concluding with “and so on,” “and on we go,” or “I’ll spare you the time.” This is despite the chatbot’s insistence that it’d be possible for it to complete the count if given enough time and patience.
Curious whether ChatGPT would be willing to entertain my own lofty counting request, I pulled up the app on my iPhone and gave it a shot. I attempted a number of different phrasings for my request, and despite putting in some elbow grease, the large language model-based (LLM-based) chatbot simply refused to oblige.
In fact, not only did the chatbot shut down my incessant pleads, but it also began to gaslight me. In no time at all, it suggested that I’d feel a sense of overwhelm with so many lines of text on screen, that I’d never even be physically capable of seeing the entire list if I tried, and that my device would outright crash during the process. Sounds like a challenge, if you ask me.
- OS
-
iOS, Android, macOS, Windows
- App
-
Free, with in-app purchases
- Price
-
$20 per month
- Free trial
-
No
The flagship chatbot and large language model (LLM) service from OpenAI, which can answer complex queries and leverage generative AI skill sets.
ChatGPT is opinionated when it comes to big numbers
I simply couldn’t find a way to trick ChatGPT into counting to one million
Back in April, OpenAI’s CEO Sam Altman confirmed on social media platform X that the user habit of saying ‘please’ and ‘thank you’ to its ChatGPT service was costing tens of millions of dollars in electrical fees. This raised very real questions as to whether there is a moral obligation to speak to AI chatbots in a polite manner, or whether there is actually a moral obligation to do the opposite to save on energy.
While the company hasn’t come out publicly with any acknowledgment of the up-and-coming ‘count to one million’ trend, I reckon it has much to do with similar cost and energy-saving considerations. To count to one million — whether in written or audio form — would surely put a strain on the electrical grid, to say nothing of server costs and potential bandwidth blockages.
For what it’s worth, I replicated my (admittedly unscientific) experiment across several other LLM chatbots on the market, including Google Gemini, Microsoft Copilot, and Meta AI. Across the board, each and every AI service left me high and dry, unable to complete my request.
Interestingly, Copilot suggests that counting to one million at a rate of one number per second would take 11 days, 13 hours, 46 minutes, and 49 seconds. This is in stark contrast to ChatGPT’s questionable claim of the act requiring ‘months.’ With Copilot relying on the same underlying tech as ChatGPT, I’m not sure what caused the latter to come to this head-scratching conclusion.
