Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?
This is your opportunity to ask questions. No question too simple or too silly.
Culture war topics are accepted, and proposals for a better intro post are appreciated.
Jump in the discussion.
No email address required.
Notes -
Who will replace us?
History is the story of societies being replaced by more competitive societies. The Sumerians were replaced by Semites and the Beaker people by Indo-Europeans. The Diadochoi states' apartheid societies were replaced by Rome's more assimilative state. Roman paganism was replaced by Christianity. Feudalism was replaced by absolutism, which was replaced by republicanism. What is best for society is often not best for the individuals within it, molochian competition is like that. Stalin built a superpower but he paid the price in destruction of Russian lives- but he turned a backwards society one generation removed from functional chattel slavery into a superpower that dominated half the world. The Roman Empire solved the stability problems of the late republic- at a cost that contemporary sources note, but don't let us check.
Who will replace today's liberalism? Christian communities are still hanging on, of course. But their numbers are dwarfed by blacks and Muslims, and despite their paltry human capital Quantity has a Quality all its own. And of course there's AI(I expect AI to never figure out how to do maintenance on its own datacenters). East Asian societies have low fertility now, but their collectivism might give them a leg up in the future- would Koreans or Chinese accept whatever measures to raise the fertility rate, unacceptable in Europe and the Anglosphere? Or will it be a crude, bombastic secular conservatism- popcountry style?
Never is a long time. It still seems plausible that even if continual learning is the most important bottleneck, and even if algorithmic improvements fail to crack it cleverly, brute force will still work eventually. Maybe not until 2050 instead of the 2030 the more breathless people are predicting if we have to wait on Moore's Law, but it seems fairly inevitable unless some catastrophe kicks industrial civilization significantly back before then.
Moores law is running into physical limitations before then.
Moore's law, strictly interpreted as "number of transistors you can fit on a 2d die doubles every two years", will be running into physical limitations soon. Moore's law, loosely interpreted as "amount of compute you get for a given monetary or energy budget doubles every two years" is still relatively far from physical limits - Landauer is still about 6 OOMs away, so there's some headroom even without reversible computing. 6 OOMs of continuing halving of energy cost every two years would take us to 2065.
I'd expect further architectural changes to be required before then, because I don't expect the compute to be shaped optimally for the ML techniques of 2025, but "the SOTA ML algos are the ones that are able to take advantage of the hardware available at the time" is nothing new.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link