This article explores the origin of large mass ratios among quarks and leptons, proposing that these ratios arise naturally from fundamental couplings that are assumed to be of the same order of magnitude but otherwise random. The authors suggest that the bare left- and right-handed components of quarks have different values for some almost conserved quantum number, which determines the order of magnitude of their masses. They predict that the generalized Cabibbo mixing angles connecting different quark types are proportional to the square root of the corresponding quark mass ratios. However, they find it challenging to explain why CP violation is so well conserved. The article discusses the renormalization group equations for Yukawa couplings and concludes that large mass ratios are not naturally generated without special selection rules or quantum numbers. They propose a model based on a single almost conserved Abelian charge R, where the mass matrix is factorized with random complex coefficients. This model naturally generates large quark mass ratios and predicts the order of magnitude of the weak coupling matrix elements. However, it conflicts with the observed smallness of CP violation in K-L decay, suggesting that the basic philosophy of naturality or random couplings may need to be reconsidered.This article explores the origin of large mass ratios among quarks and leptons, proposing that these ratios arise naturally from fundamental couplings that are assumed to be of the same order of magnitude but otherwise random. The authors suggest that the bare left- and right-handed components of quarks have different values for some almost conserved quantum number, which determines the order of magnitude of their masses. They predict that the generalized Cabibbo mixing angles connecting different quark types are proportional to the square root of the corresponding quark mass ratios. However, they find it challenging to explain why CP violation is so well conserved. The article discusses the renormalization group equations for Yukawa couplings and concludes that large mass ratios are not naturally generated without special selection rules or quantum numbers. They propose a model based on a single almost conserved Abelian charge R, where the mass matrix is factorized with random complex coefficients. This model naturally generates large quark mass ratios and predicts the order of magnitude of the weak coupling matrix elements. However, it conflicts with the observed smallness of CP violation in K-L decay, suggesting that the basic philosophy of naturality or random couplings may need to be reconsidered.