P: (Masking) (AI) generated depth map for depth masks

Create a depth mask option and make lens blur available in all masks. This would have two great advantage: first, it would allow all the regular mask adjustments to be made based on depth of field. For example, you could warm up the in focus area. Inverting the mask would allow you to cool down the out of focus area. Second, adding a lens blur tool to all masks would be handy to blur out a vignette or to add a lens tilt effect using a linear gradient.

We have three standard mask types, Subject, Sky and Background.  I would like to see one more.  Not sure what to call it but for sake of this request, lest's call it Out Of Focus (OOF) or maybe Blur.  In the Lens Blur panel, the algorithm does a great job of isolating OOF areas and even has provision to adjust it.  I would like to be able to create an "OOF" mask so that these areas can be teated differently e.g. darken and soften to add focus on the subject.  This would have a similar effect to using "background", only, I think, better.  It will include close OOF areas as well. 

On the face of it, it seem a simple task since Lens Blur is already there, but I appreciate things are rarely that simple.   

I love this idea! I stumbled across it while searching for ideas on extracting the depth map for masking to increase the lighting on individuals while making the background darker. 

I like this idea too and would use it in much the same way as described. 

Thanks for the info. 
My first thought was to add a blur slider to Local Adjustments with Bokeh Control that could be zone-controlled via an AI-calculated depth mask. This is a different interpretation. 

Sorry, let me clarify, by saying "different things to different layers of mountains" i mean adjust exposure/color temperature/highlights. The usual local adjustments.

I missed this feature when editing a landscape shot with several "layers" of mountain ranges behind each other. Sometimes I want to do different things to different layers of mountains. Here is an example where AI foreground/background detection didn't select anything:

Other applications are that it might just make selecting things easier. When the AI object detection/subject/background code fails me and one is back to manual masking, quickly removing certain depth ranges might help and would reduce the pencil precision required when masking.

Does that sound reasonable or am I holding it wrong?

While I don't disagree with your request, can you describe when and how you would use it?