Looks to me like a way to generate models using artificial intelligence. The problem with these training algorithms is that you can never be sure what they're picking up. For example, there was a study where they were trying to train a computer to recognise the gender of a human face, they trained the computer and then set it off, it went wrong and nobody understood why. After some investigation it was realised that the computer was picking up on some detail in the background on the training images. I'm sure the algorithms are better now, but the lesson remains the same: can you trust your models if you're not quite sure what the computer is basing them on?
With regards to uncertainty, I was expecting some kind of quantitative measure of the uncertainty of geophysical models of the subsurface but I didn't see any of that.
