I like your sample above, I think it implements the spirit of MVVM. Just for clarification, though, the ViewModel code and the Model code should not be in the same source file as the actual Code Behind. In fact, I would argue that they should not be in the same project.
Here is MVVM as I understand it:
M - the Model is the data returned from the Business Layer (BL). This should be lightweight, containing read-only data. The Model classes are dumb and do not contain Update, Write, or Delete logic, and are generated by the BL as the result of requests, commands, actions, etc. The Model classes have no knowledge of the presentation needs of the consuming application, so they can be used by any manner of application. To really take advantage of this reusability, we want the Model classes to be independent of the UI project.
VM - the ViewModel contains the communications layer: it issues the requests to the BL and processes the results in a manner fit for presentation. Like the example above, it also receives the Model and reformats it for the particular presentation needs. Think of this as a "Binding Class". In the example above, the data is simply being moved from one object to the next, but the ViewModel would be responsible for such things as exposing a "FullName" type property or adding leading zeroes to a ZipCode. It is correct that the Binding Class is the one to implement INotifyPropertyChanged. And again, for reusability I would probably extract this layer into its own project as well. This would allow you to experiment with various UI options with no plumbing changes.
V - the View is bound to the Binding class object created in the VM. The View is super dumb: it knows nothing of the BL or the VM. Data can be bound in both directions, but the VM handles errors, validation, etc. Any data synchronization operations are handled by passing requests back the the BL, and again processing the results.
It would depend on the type of application, but it seems heavy handed to constantly check the Model to see if it has changed. Pretend you were connecting to a BL that built the Business Object(BO) from a DAL, which is connecting to a DB. In this scenario, you would be constantly recreating the BO, which I'm sure would be a performance killer. You could implement a checkout system on the BL that listened for notifications, or have a method for comparing last known time of change to actual, or you could cache the BO on the BL. Just some ideas.
Also, I said above that the Model should be Lightweight. There are heavyweight options, like CSLA, but I'm not sure how well they fit into the MVVM idea.
I don't mean to pass myself off as an expert, I have only been studying these ideas so far while designing our new software's architecture. I'd love to read some discussion about this topic.