I would be curious as to whether the graphs would look dramatically different if, instead of taking the UN population projections which assume mean-reversion (very unclear if / how soon that'll happen) it instead assumes that the birth rates stay at their current levels in each country.
I would like to see the datasets and methodologies used to predict global demographics by the UN. As with all things, geographic scale matters. It’s perhaps easier to analyze spatial and demographic data of a city like Padua than it is to predict the economic and demographic future of Brazil. What is needed is general “cases” and “frameworks” that can analyze trends and movements as they metastasize and grow.
I wonder, what happens in this model to countries that can replace ~half their workers with 120+ IQ workers, which is where the latest publicly available leading edge AI models are? With 140+ IQ? (Next December's models.)
(Sure, AIs currently have problems with maximum time on task and with memory/learning from experience. But max task time is up from ~15 minutes to ~420 minutes over the course of 2025 and there does not appear to be a slowdown in improvement yet, and memory is a hot research topic with no apparent show-stoppers. This time next year we can expect max task duration to be ~200 hours, an intense month for a human.)
Super interesting!
I would be curious as to whether the graphs would look dramatically different if, instead of taking the UN population projections which assume mean-reversion (very unclear if / how soon that'll happen) it instead assumes that the birth rates stay at their current levels in each country.
central asia seems to be a very underrated region with a lot of potential
I would like to see the datasets and methodologies used to predict global demographics by the UN. As with all things, geographic scale matters. It’s perhaps easier to analyze spatial and demographic data of a city like Padua than it is to predict the economic and demographic future of Brazil. What is needed is general “cases” and “frameworks” that can analyze trends and movements as they metastasize and grow.
Thaks for all that work, Sebastian!
I wonder, what happens in this model to countries that can replace ~half their workers with 120+ IQ workers, which is where the latest publicly available leading edge AI models are? With 140+ IQ? (Next December's models.)
(Sure, AIs currently have problems with maximum time on task and with memory/learning from experience. But max task time is up from ~15 minutes to ~420 minutes over the course of 2025 and there does not appear to be a slowdown in improvement yet, and memory is a hot research topic with no apparent show-stoppers. This time next year we can expect max task duration to be ~200 hours, an intense month for a human.)