They sound like interesting tools, especially since there are some sorts of problems where small amounts of unpredictable error can be helpful, like repetitive Monte Carlo-style simulations.

But it also seems like an intimidatingly large area to start learning about.