Increasingly complex generative models are being used across the disciplines as they allow for realistic characterization of data, but a common difficulty with them is the prohibitively large computational cost to perform likelihood-based statistical inference. We consider here a likelihood-free framework where inference is done by identifying parameter values which generate simulated data adequately resembling the observed data. A major difficulty is how to measure the discrepancy between the simulated and observed data. Transforming the original problem into a problem of classifying the data into simulated versus observed, we find that classification accuracy can be used to assess the discrepancy. The complete arsenal of classification methods becomes thereby available for inference of intractable generative models.