@jdw@mathstodon.xyz i really think the marketing is too strong, and a lot of people, myself included, started making these non-obvious leaps as given. im not even saying we won’t get there, but im saying taking it for granted that we will is a really big leap. i wouldnt say that such leaps were needed for predicting the problem solving we saw so far, if i may use the very fact that you predicted it to argue that. in the same way id consider changing my mind on the size of these leaps if i saw that people were speculating that we found most of the big/revolutionary ideas. now, my whole argument basically hinges on the fact that the ai would either be very slow, or would be unable to to rediscover the mathematics of today, say, if its starting point was only basic addition, or precalculus mathematics, or some other mathematically underdeveloped wrt today’s standards period in mathematics’ history. this i didn’t think through all the way (hence the occasional sprinkled question marks in brackets in my comment). (this is perhaps a silly thought experiment, given that we wouldn’t have the mathematics to build the machine in the first place. but am i allowed to pull “alien tech of equivalent capabilities” card?) @maxpool@mathstodon.xyz @tao@mathstodon.xyz