Видео
click to show
click to show
🤖 AI cracked a math problem that stood untouched since 1946 and the scary part is how ordinary the system was.
OpenAI says its general-purpose reasoning model disproved a major conjecture from legendary mathematician Paul Erdős, solving the long-running “planar unit distance problem.”
For decades, mathematicians believed the best way to maximize one-unit distances between points was basically some variation of a square grid. The intuition looked obvious. Turns out it was wrong.
The model discovered an entirely new infinite construction that beats the old assumptions with a polynomial improvement and external mathematicians verified the proof.
What makes this different from past AI math wins is that this wasn’t a specialized theorem-proving machine built for one task. It was a general reasoning model, and its performance improved simply by giving it more inference-time compute, more time to “think.”
Even more interesting: the breakthrough required connecting simple geometry to deep algebraic number theory, including concepts like infinite class field towers and Golod–Shafarevich theory.
That suggests AI may become less about replacing human intuition and more about exploring conceptual paths humans never thought to check.
Source.
@aipost 🏴
5.6K ·