I disagree. It may be "smart" from a technical point of view, but not of a users point of view. There is very little correlation between the image size and what image is the most important, or most relevant, to the user. On top of that, the thumbnail often crops the image in the middle, making it almost unrecognizable. Don't see anything "smart" about this. Choosing the first image whose width and height are both greater than that used for borders would be "smart", as well a give the user control over which image is used for the thumbnail.