Well, the probability that the kernel of $M$ is nontrivial is bounded above by around $(3/4)^n,$ by Tao-Vu (J. Amer. Math. Soc. 20 (2007), 603-628), so there is that...