For example: the GTK file picker dialog.
There's no reason why you can't innovate along the way, but it's pretty clear to me that if Linux is ever to make any inroads into non-geek circles, it'll need a familiar interface.
i see a lot of "make Windows looks like Mac OS X" from google or even "make Linux look like Mac OS X"
i believe I've never seen the reverse (OSX->xp or even OSX->linux)
to me linux is copying the wrong model, her role model should be osx
OS X has the most beautiful aesthetic of any computer I know. But it's not just the looks that matter. It's the functionality. You don't have a taskbar and an in-window menu bar. You have a dock and a top menu. And that's the best system I've ever worked with. You can't copy that over to Windows or Linux effectively, either.
If a Linux distro was smart, it would emulate OS X rather than Windows, because fundamentally, the design is more efficient and more usable. As long as they copy Windows they're at a disadvantage.
Only the absolute minority of people make their Windows or Linux machines to look like Mac OS X either - you're looking at a minority of enthusiasts using the fringes of operating system functionality and you're pointing to that as definitive proof that something is better. To me, that's outright flawed, especially when there are core software limitations that prevent you from imitating the Windows or GNOME UIs in Mac OS X.
You can have a top-menu and a dock in Linux - there's pre-built tools that let you do that (e.g. https://wiki.ubuntu.com/global_menu, activate the MacStyle menu setting in KDE, combine them with a dock app like Avant Window Navigator). Only an small minority of people use them and it's not very well supported because of it, but it's certainly possible. The Linux desktop is incredibly configurable.
Note that I haven't actually stated my personal opinion on the matter at all so far. I think you're suffering from a case of self-confirmation bias.
You can have that in Linux, but it's one thing for it to be possible and another for it to be the out-of-the-box default. It's the problem with choice: it's confusing and it takes time and other programs don't configure to the same standards.
My personal opinion is that for every advantage that OS X has via its menubar and dock layout it has an equal disadvantage, so both styles are equal.
For example, for the menus at the top gives you the advantage of Fitts' law but makes accessing menus while multitasking and switching between applications and using menus more difficult. Moving from looking at the top of the screen to the window is a further distance, which may make viewing slower. You do get the advantage of a more intuitive MDI applications though.
The dock is better for frequently used applications, but it uses more screen space, the locations of applications in the dock shift as you use it and it uses dramatically more space than the quick-launch bar. It also stylistically makes maximisation look rather 'off'.
If you solely rely on keyboard navigation, Windows-style is probably better - you can easily load an application without the mouse (especially in Vista), plus you primarily use alt+tab/cmd+tab which means that the menu is largely wasted space. If you use many applications, the Start Menu is likely quicker than using Finder, especially on Vista with its built-in search. Of course, there's Quicksilver and their ilk but that's not standard and similar applications are available cross-platform.
Finally, if you believe they're approximately equal, if you're developing a new OS from scratch with no existing users and you have to choose between the two it's probably better to lean to lean towards a Windows-style layout as more people are familiar with it to narrow the learning curve, which would make choosing Windows-style windows as the default a reasonable decision.
I'm certainly not saying this was the thought process at all - let's not forget Mac OS X was years from release when the GNOME and KDE projects were started and they were building on top of the paradigms of existing X applications - but it certainly shows that if you were making the decision today choosing the Windows model would still be a reasonable one, and that's excluding the factors involved in changing like alienating the current user-base.
I wonder how this sentiment would apply to linux DEs. Consider the design decisions which would be made if the designers of gnome and kde thought about what would make them most productive instead of thinking about what will get people to switch from windows to linux. Perhaps, we will get a lot less dumbing-down of interfaces and an end to blind adherence to the standard windows UI and finally see some genuine innovation in this field again.
This is not to say that all DEs are stuck in a cycle of continuously copying windows. FVWM (my wm of choice) splits completely. Similarly, from an eye-candy perspective, beryl and compiz bring new things to the table. Still, beryl and compiz are relatively small projects and the vast majority of coding effort goes into make nice GUIs for systems configuration and things like emulating MS Office with OOo.