By 2070, it will become possible and financially feasible to build APS (Advanced, Planning, Strategically-aware) systems (65%)
For what it’s worth, my estimate for APS by 2035 is ~95%. I think there are a lot of gradually accelerating processes underway affecting AI R&D which have positive feedback cycles. Trying to estimate how fast AI R&D will lead to a given capability threshold without taking positive feedback loops into account is almost certainly going to result in a huge underestimate of the speed.
For what it’s worth, my estimate for APS by 2035 is ~95%. I think there are a lot of gradually accelerating processes underway affecting AI R&D which have positive feedback cycles. Trying to estimate how fast AI R&D will lead to a given capability threshold without taking positive feedback loops into account is almost certainly going to result in a huge underestimate of the speed.