Random Sampling of a Custom Probability Distribution

I am trying to make a point-cloud representation of implicit 3D graphs. However, I can’t seem to find a way to reliably and quickly do random sampling on an implicit 3D function (So, treating the function as a custom probability distribution then doing the equivalent of normaldist(x).random()). Does anyone have any ideas?

Interesting it’s not treated as a list, but a slight adjustment assuming x is a list:

normaldist(x.random())

which will choose a random element from the list x and define normaldist for that element.

Here’s a sample graph.