This blog post indicates no, based on a number of good arguments.
It is really true that Silverlight 2.0, out of the box, has a lot of holes or gaps that must be filled in or worked around.
Of course my shameless plug is for CSLA .NET for Silverlight, which fills in most of those gaps. We did encounter the issues listed, and many others. Ultimately CSLA .NET for Silverlight is the result of around 9-10 man-months of effort. That's 9-10 man-months of time finding, researching and solving issues around business application development in Silverlight 2.0.
So while I agree that just picking up Silverlight and running with it will require some serious effort (as compared to WPF, Windows Forms or Web Forms), I don't think that's the whole story. The reality is that frameworks and tools will come into existence (or already have in the case of CSLA .NET) that bring Silverlight's productivity at least up to the level of WPF.
Then the argument becomes that WPF (and Silverlight) lack solid tooling for building the UI. When compared to the designers available for Windows Forms and Web Forms that's absolutely true. However, if you look under the covers, the amount of code and effort required to actually build an app with WPF/Silverlight is absolutely less than Windows Forms or Web Forms - at least when using CSLA .NET.
I say this, because the exact same business objects and data access work in all those environments. So the business and data code is a wash - it is the same in all cases.
But the UI code for WPF/Silverlight is a tiny fraction of the code required for Windows Forms or Web Forms. XAML does require a learning curve, but once you get over that curve it is far more productive, even without good designer tooling.
Yes, I'm totally sold on the Silverlight/WPF/XAML UI future. Yes, there's a learning curve, and yes there's some initial pain. But once you get past that initial hurdle it is sooooo nice. You won't find me going back!