The point is that you can view things as a (approximate) form of bayesian inference, in order to think differently about how they work, and what they're doing.

The point is that you can view things as a (approximate) form of bayesian inference, in order to think differently about how they work, and what they're doing. Another common example is ridge regression being expressed as a bayesian regression (where the "ridge" part is due to the prior).