@masayoshisson It really only has a real world impact if new algorithims are found and encryption breaks or AI R&D speeds up. which would be the P=NP problem. But yes, mathematicians are now more analogous to historians.
@scaling01 I was expecting to see more concrete evidence out of OpenAI at this point tbh and the data for Anthropic is suspect..they clearly have the higher capability models , but not by much, and you also pay for it in token spend.
@treswritesstuff Wait so is every artist who is signed to a label an industry plant ? Distribution channels are unfortunately necessary to sell most products in the US lol
@GregHBurnham is this because you don’t think we can create good enough synthetic data and can’t self play on fuzzy stuff yet? And in the future we will be able to?
@Gegam245074 If we are plateauing at coding capabilities, is this signal that intelligence is topping off and gains elshwhere are recent RL priorities paying off ?