The Arimoto algorithm computes the Gallager function maxQ E0(ρ, Q) for a given channel P (y / x) and parameter ρ, by means of alternating maximization. Along the way, it generates a sequence of input distributions Q1(x), Q2(x),⋯, that converges to the maximizing input Q∗(x). We propose a stochastic interpretation for the Arimoto algorithm. We show that for a random (i.i.d.) codebook with a distribution Qk(x), the next distribution Qk+1(x) in the Arimoto algorithm is equal to the type (Q′) of the feasible transmitted codeword that maximizes the conditional Gallager exponent (conditioned on a specific transmitted codeword type Q′). This interpretation is a first step toward finding a stochastic mechanism for on-line channel input adaptation.