Is the brain a predictive machine? Is prediction its only function or one among many? Predictive coding theories argue that all the brain is doing is inference, and the way it is doing it is through Bayesian inference (i.e. predicting what will be the input to the brain in a Bayes-optimal way) and active inference (i.e. the brain acts through a body with the intention of ensuring the predictability of the environment or the same, reduce the free energy [2]). Nonetheless, Bayesian inference is a computationally expensive process for which there's no explanation on how a brain built with neurons and neuromodulators could perform such process. Previous attempts at defining a neuronal mechanism for performing Bayesian inference failed at defining how a model of the world was originally formed [1]. We reformulate the previous attempts to provide a neuronal model that performs Bayesian inference without previous assumptions about the world except for normality. This model becomes a biologically plausible learning mechanism that uses unsupervised local learning rules to learn the statistics of an input and stop learning when the prediction error can't be further reduced.