Tech mogul Elon Musk stated that "predicting the future is the best measure of intelligence," pointing to the accurate prediction by his chatbot Grok of the date of the strike targeting Iran.
Musk's remarks came in response to a user on the X platform (formerly Twitter), who said that "predicting the future is the best measure of intelligence," noting that Grok had predicted February 28th as a likely date for escalation.
He explained that this was not a mere coincidence, indicating that Grok based its prediction on an analysis of geopolitical signals, the outcomes of the Geneva talks, and real-time data to pinpoint the date precisely. He believes the platform is capable of gauging global sentiment and anticipating international trends.
A report in The Jerusalem Post compared four artificial intelligence models—Clude, Gemini, Grok, and ChatGPT—in an attempt to predict the date of a potential US strike against Iran in 2026, based on publicly available data and media reports. The aim of the experiment was to test the models' ability to analyze general indicators, not to obtain classified information or make definitive predictions.
Initially, the Claude model offered a gradual analysis, moving from rejecting the prediction to proposing a possible scenario of a limited strike in early to mid-March, later narrowing its scope to March 7 or 8, 2026. The Gemini model, on the other hand, focused its analysis first on diplomatic catalysts, then identified a timeframe between March 4 and 6, 2026, highlighting the potential for nighttime military operations.
Read more : Why did Elon Musk postpone the "Grok" update ?
The Grok model provided a clear date from the outset: February 28, 2026, linking it to the outcome of the Geneva talks, while acknowledging that various scenarios could alter the likely date.
In contrast, the ChatGPT model initially proposed March 1, 2026, then revised its estimate to March 3, based on a deadline set by Donald Trump and subsequent diplomatic and military visits and movements. The report concluded by noting that all models relied on open-source information, such as Reuters reports, and that the differences between dates reflect the different analytical weights of each model, stressing that the experiment aimed to assess the ability of artificial intelligence to interpret general indicators, not to predict events with certainty.
