Togelius: Automatic Game Design

An interpretation of Raph Koster's "Theory of Fun for Fame Design", or of Juergen Schmidhuber's curiosity principle wrote:
"A good game is one that is not winnable by a novice player, but which the player can learn to play better and better over time, and eventually win; it has a smooth learning curve."

Not sure I agree with this, certainly not for what I call a good game, though perhaps that definition is so obscure, capricious and mercurial that it would elude anyone."

I think the definition above is more of "a game that is not too easy, nor frustratingly hard" - focused only on the difficultly, not whether it is actually any fun. Of course measuring this is a bit more tricky, involving the usual method of employing playtesters. This is why I think making a fun game, as the title mentions is beyond anything automated, however recursive it may be - feedback is needed from a human at some point. After all, it takes legions of designers and developers to come up with some right stinkers - what chance does a machine have of lucking out with an awesome game?