The problem that I have with all the new GUIs that are coming out it seems like it's all just change for the sake of change.
For this to be correct would suggest that what we already had was just fine and having to help people interact with previous-gen UIs on a regular basis tells me that it very seriously isn't. Windows XP, OSX, and Gnome 2 (not to exclude others I have less experience supporting) all have a slew of bizarre behaviors and ungrokable configuration layouts. If you want a very solid example of why these systems all fail your average user, take file management and the number of icons which exist on most peoples desktops. There is seemingly no end of problems for HCI designers to tackle.
Another issue which has changed a lot is a fundamental change in input method which has finally become prominent for many users: touch-screens. One of the great challenges of a touch screen is that even though people complained about Macs only having a single mouse button, touch-screens have to fake mouse buttons using tricks of touch duration and force, which are very poor. This makes several paradigms in design more difficult, and in particular it is difficult to do the click-drag which is at the center of a great many activities performed previously. When you see the changes in UIs such as Unity, Gnome 3, and the various mobile GUIs you can see efforts in this area.
So, while I think I have a bit of a handle on some of the problems trying to be fixed, I do have gripes about how they are trying to fix them:
Sorted grids of icons are a terrible way of organizing data so that it can be found. Yes, I understand that you can fit more icons on the screen if you do a grid, but it means that I have to do far more mental acrobatics to find anything once I have more than a few icons, so the only real usage case (wanting to display many elements) is the worst when actually using it. I have found a very strong correlation between people who are effective computer users and those that change their file browser settings to a proper list view (proper meaning not what Windows calls 'list', but what Windows calls 'details').
Unsearchable, fixed categorization is just maddening. Helping people override Gnome 2's program menu is something I have done far too many times to be reasonable. If I am someone's Windows computer and go into Control Panel and the user hasn't switched to 'classic view' tells me that the person using the computer doesn't come in here much. Recently I wanted to disable the Windows 7 edge of screen window docking feature and found it in the 'Ease of Access' control panel under 'Make My Mouse Easier to Use' and to top it off, to disable a feature I had to add a check to a box, which is totally out of paradigm. Making a series of choices among fewer options may seem like a good idea to reduce confusion, but it all falls apart if users can't relate to the organizational structure you have invented and if a user wants to invent his or her own structure it should be easy to do.
Automatically generated lists of 'recent' or 'frequently used' resources (like applications and documents) are unreliable and highly problematic. I frequently hear the cries of users who become accustomed to accessing a resource via these lists and then freak out when they don't know any other way of finding that resource if it disappears from the list for some reason. If people aren't accessing resources from a reliable source which only changes when they intend it to, then this will be a problem.
The OSX dock is a terrible idea and it is creeping into other UIs where it is even worse. The idea is definitely engaging in that a user who wants to open a frequently used application should just click the launcher for that application whether it is open or not, so just 'pin' a reference to the application manager. The problem comes when we want to then separate the act of initiating a new instance of an application or selecting a currently running instance. Now we have to have hovering sub-windows or alternate clicks.
I could go on, but I hope you see that there is a lot to complain about classic UI design which really should be addressed, so it isn't simply for the sake of change that these people are proposing changes even if I think that the changes they are proposing are ill-informed and sometimes are trying to solve problems which they have invented without a personally explainable cause.