Grok named February 28 as strike date days before attack
Jerusalem Post tested four AI chatbots with same prompts
It was a lucky guess by Grok, no matter what Musk says
No, despite whatever you might have read on social media, Grok didn’t predict the US strike on Iran that took place over the weekend. A viral Elon Musk tweet is being twisted out of context, and Elon Musk being typical Elon Musk here is milking it to make Grok and xAI look better than competitors like ChatGPT, Claude and Gemini.
SurveyIn case you missed the armed conflict between Iran vs Israel and the US, which began in the early hours of Feb 28 for all of us here in India, there’s been heavy firepower being exchanged between Iran and the US. With missiles and drones leading to deadly explosions deep inside Iran and US army and naval bases in the Middle East. Obviously, in a desperate bid to get answers, people have been asking AI chatbots all sorts of questions – including one which tried to predict when this current ongoing conflict would erupt.
Yes, that’s right. On February 25, the Jerusalem Post ran a war forecasting exercise, where they asked Grok, ChatGPT, Claude and Gemini the exact same prompt. The Jerusalem Post pressed each AI chatbot to pick a start date for a purported US strike on Iran.
In its response to the Jerusalem Post query, ChatGPT initially suggested March 1 as the day the US would strike Iran. But later, ChatGPT revised its response and picked between March 3 and March 6 as the strike window. In contrast, Gemini said a US strike on Iran was likely to happen between March 4 and 6. Claude simply refused to name a date at all, according to the Jerusalem Post report.
Also read: US-Israel strike on Iran relied on Anthropic AI despite Trump’s ban: Report
In picking a specific date for a hypothetical US strike on Iran, Grok gave the clearest single-day answer: Saturday, February 28. It repeated that same date when checked again later, according to the Jerusalem Post.
Grok wasn’t really predicting the future based on secret intel
Of course, that was all the reason for Elon Musk to cheekily celebrate this result on X. He retweeted the Jerusalem Post’s headline, with typical Elon Musk pomp.
“Prediction of the future is the best measure of intelligence,” Musk said. In saying so, of course, he was giving Grok way too much credit than it really deserved.
The Jerusalem Post’s editorial team have themselves said that their attempt to put Grok, ChatGPT, Claude and Gemini against each other to forecast any US strike on Iran wasn’t actually an exercise to predict the future and “it does not validate the model’s reasoning.” It only validates the reality of a very small set of plausible windows fed in by them, of which one model happened to land on the day that turned out to be real.
Additionally, Israeli defense officials said the strikes against Iran were coordinated with the US armed forces and planned for several months in advance, with the launch date picked multiple weeks ago. Since Grok didn’t have access to classified US military planning data, it only made an educated guess based on publicly available information and geopolitical signals, nothing more nothing less.
Follow Us
Jayesh Shinde
Executive Editor at Digit. Technology journalist since Jan 2008, with stints at Indiatimes.com and PCWorld.in. Enthusiastic dad, reluctant traveler, weekend gamer, LOTR nerd, pseudo bon vivant. View Full Profile
