they can't even draw text, or count fingers without multiple attempts
it's because the depth of the logic required to do this systematic, logical stuff is too deep for the current level of compression they are using, if they do like you do with jpeg, and put it to 95% the size will go up dramatically of the models but they will be able to do this kind of systematic stuff
i know what i'm looking at... the human brain can compress this information far more effectively than their algorithms can which means that even if they could reach the memory storage power efficiency of a brain they are still gonna need millions or more times as much memory to do the same thing