Just thought Id bring to the table this bit of information.
It is believed taht a truth and logic are easily agreeable, but an illogical truth is not agreeable with logic. So a machine built around the sole purpose of logic will only perform to the most logical sequence, and can only be outhought by a machine of similair logic. (I think that makes sense, at least not a ramble yet...) The other way to trick a computer, is to flood the logic mind of the machine with the same data repeatedly, causing a loop and possible crash.
Example: A robot designed to believe it loves ham sandwhiches can be built using a simple logic table. If the sandwhich is ham, I love it. If their is a sandwhich, look to see if it is ham. Collect what I love. (simplified cpu thought). Now lets say you give it a device to fufill its task, a scooper if you will; and when it scoops up the sandwhich we return it to a standing and upright position. The only trouble is that when this last task is performed, the scoop fails to work properly (possibly due to the proper codes to prevnt this, or other manual stuff) and the sandwhich ends back on the floor. Because it is now in a position without what it loves, it begins the search for a sandwhich (possibly just looking down for it) and decides it loves it again, and repeats the process. The emotion produced is boredom and satisfaction, (for the observers and the robot).
I believe this is one example of how a CPU can 'feel'. A simple 1 in place of a 0 can simulate a sensation that causes the mind to stop its previous function/objective/desire. It should be noted, these are similar words to Douglas Adams, but not quite.