people compared GPT-5.4's solution of erdos #1196 to alphago's move 37 but i think a tighter analogy is to alphago's unusual (at the time) p…

people compared GPT-5.4's solution of erdos #1196 to alphago's move 37 but i think a tighter analogy is to alphago's unusual (at the time) preferences for the 3-3 opening and early 3-3 invasion against the 4-4 opening. human go professionals managed to completely miss alpha's continuation of the 3-3 invasion and had dismissed the usual continuation as inefficient - professionals would never have played an early 3-3 invasion before, it would have been dismissed as a beginner move - even though alphago's idea in retrospect was relatively easily understood once introduced. the space of possibility is large and it really is just possible to miss a good idea like this! https://www.eurogofed.org/?id=127 Terence Tao's takeaway is that GPT didn't have any grand idea, but human researcher culture has just… missed the basin where this problem is almost trivial. GPT, being nonhuman, reliably solves it in under an hour. In a way, this is even more humbling. https://erdosproblems.com/f…

Source: Emad Mostaque (X) | 2026-04-21