In ancient Egypt, Pharaoh had really weird and troubling dreams: seven think cows devoured seven fat ones, and seven withered ears of grain swallowed seven hesalthy ones. Pharaoh’s court was filled with wise men and magicians who offered vague possibility afteer possibility for what it might mean. His “official analysts” hedged.

Then Joseph, an imprisoned Hebrew slave, was brought to Pharaoy. Joseph did not offer vague warnings–he counted the years and prescribed a plan. When famine struck, Egypt survived because someone had committed to a specific, actionable forecast. History rarely rewards the person who says “maybe.”

 Fast forward 3,000 years.

Global events unfold with mounting complexity. Markets fluctuate. International conflicts simmer and boil in distant regions. Crises emerge with little warning. We find ourselves awash in a sea of hedged probabilities in the data-driven forecasts of artificial intelligence systems that detect patterns others miss, but cannot agree on what specifically do with them.

On February 28, 2026, as the United States and Israel launched coordinated strikes on Iran, Grok—the least restricted major AI model publicly available—had identified that exact date as the likely start of the wat. An experiment published three days earlier in The Jerusalem Post asked four major AI systems to do something most are designed to resist: pick a single day for a U.S. strike.

The models were nudged and prodded under varying constraints. Claude refused to name a date, offering instead a March 7–8 probability range. Gemini produced a “calendar of triggers” between March 4 and 6. ChatGPT clustered around March 1–3, each answer wrapped in caveats.........

© The Times of Israel (Blogs)