PDF is typically directly generated these days. There's no need to generate it from PostScript first.
The main difference between the two is that PDF is a declarative definition of the page whereas PostScript is a turing complete language. You can write a PostScript ray-tracer which will execute on the printer itself something which would never be possible in PDF. On the flip side, PDF is far more efficient as an electronic document format.
No, PDFs don't have to be generated from PS at all. PDF has it's own well-defined spec. Valid PDF files can be generated in many ways besides from PS files. It's even possible to write a valid PDF file by hand if you're feeling masochistic.
Not exactly. A PDF is generated from PostScript, as noted in the article:
"A PDF file is a static representation of the result of executing the PostScript file on a single occasion."
A PDF is a more compact version of a PostScript file - we could call it compiled PostScript, to use a term many on this site are familiar with. Just as a compiler has many output options, distilling PostScript can result in many different file sizes and levels of quality. (Check out the Distiller output settings.)