They are identical for all distributions.

Expected information is calculated using the second derivative of the log-likelihood at the true $\theta$, while observed uses the sample.

Expected information is the expectation of the negative Hessian of the log-likelihood over the sample space; Observed is evaluated at the MLE.

Observed information is always larger than Expected information.