Oops, thought I could trust “reasoning” AI (Gemini 3 Pro) for such a simple seeming question. Had it redo the estimate taking your comment into account, and it came up with 1m assuming N(90,15) globally, which still felt wrong, so I had it redo the estimate using country-level data, and it ended up with 7.5m with 6.1m in East Asia, 1.1m in the West, and .3m in RoW. This assumed N(105,15) for East Asia (so not quite using country-level data), which Opus and GPT point out might be an overestimate due to China being a bit lower than this. Had them redo the EA estimate using country-level data and they came up with 4.5m and 5.5m for EA (using N(103,15) and N(104,15) for China) respectively.
This is actually a significant update for my mental world model, as I didn’t previously realize that China had more than half of the world’s population of IQ>145 people.
Oops, thought I could trust “reasoning” AI (Gemini 3 Pro) for such a simple seeming question. Had it redo the estimate taking your comment into account, and it came up with 1m assuming N(90,15) globally, which still felt wrong, so I had it redo the estimate using country-level data, and it ended up with 7.5m with 6.1m in East Asia, 1.1m in the West, and .3m in RoW. This assumed N(105,15) for East Asia (so not quite using country-level data), which Opus and GPT point out might be an overestimate due to China being a bit lower than this. Had them redo the EA estimate using country-level data and they came up with 4.5m and 5.5m for EA (using N(103,15) and N(104,15) for China) respectively.
This is actually a significant update for my mental world model, as I didn’t previously realize that China had more than half of the world’s population of IQ>145 people.