Researchers at the University of Cincinnati are trying to program computers to recognize a joke. So far, they've only gotten to the kindergarten level, by teaching computers to recognize words that sound the same (but are spelled differently) or words that sound somewhat similar and can be taken more than one way. In a word, puns. The example in the article reminds me of my daughter's first joke. She approached me carrying a broom and said, "Shh! Be quiet! I'm trying to SWEEP!" I laughed at that all day, but mainly because she was two years old and had never successfully told a joke before.
Joke recognition software could be very useful to someone like me. I'm always searching the internet for humor, but funny stories are often not labeled with the words "humor", "funny", or "joke". The program in this project allows a computer to recognize a joke, but it still cannot discriminate between a funny joke and a dud. That's fine, if you're a kindergarten humorist, but it won't lead to an improvement in joke-telling when all you receive is positive feedback (insert mechanical voice: "That is a joke. Ha ha ha"). They hope to expand the program's repertoire eventually, and learn more about human reactions to humor by replicating them in a computer. It seems like an uphill battle to me. The human brain has an almost unlimited capacity for obscure connections, which many people never use. You can tell a lot about a person by whether they understand and appreciate very subtle humor. If he "gets it", he shows a certain level of intelligence. If he gets it and still doesn't crack a smile, he may be a snob or just too serious for my tastes. Studying human reactions to humor is a complex process that most of us do without thinking.
What will this lead to? Will we eventually have household robots who not only do chores, but laugh at our lame attempts at humor? That's what we have children for!