Can AI predict the winner of a nobel prize in physics or chemistry with 85% accuracy a decade in advance ?

The question asks whether it’s possible to name a plausible future Nobel laureate in physics or chemistry—and be right 85 times out of 100—before any public announcement, fully ten years in advance. Researchers have explored data-driven techniques to spot tomorrow’s breakthroughs, but the limits of such forecasts remain hotly debated.

Background

Predictive efforts rely on mining citation networks, research trends, and past award patterns to identify scientists whose work is accumulating unusually strong and fast-growing influence. Early studies reported modest early-warning signals—accuracies of only 40–60 % when projecting major prizes five or more years ahead—raising questions about scalability to a full decade. Work that claims 85 % accuracy often depends on small, hand-curated datasets or post-hoc labeling rather than rigorous out-of-sample validation over multiple decades of Nobel history. Independent assessments emphasize the role of unquantifiable committee judgments and idiosyncratic factors in the Royal Swedish Academy of Sciences’ selection process, arguing that deterministic prediction at such strict thresholds remains infeasible.