A funny thing happened on the way to the computer
by Peter W. McOwan, Queen Mary University of London
(from the archive)
Laugh and the world laughs with you they say, but what if you’re a computer. Can a computer have a ‘sense of humour’?
Computer generated jokes can do more than give us a laugh. Human language in jokes can often be ambiguous: words can have two meanings. For example the word ‘bore’ can mean a person who is uninteresting or could be to do with drilling … and if spoken it could be about a male pig. It’s often this slip between the meaning of words that makes jokes work (work that joke out for yourself). To be able to understand how human based humour works, and build a computer program that can make us laugh will give us a better understanding of how the human mind works … and human minds are never boring.
Many researchers believe that jokes come from the unexpected. As humans we have a brain that can try to ‘predict the future’, for example when catching a fast ball our brains have a simple learned mathematical model of the physics so we can predict where the ball will be and catch it. Similarly in stories we have a feel for where it should be going, and when the story takes an unexpected turn, we often find this funny. The shaggy dog story is an example; it’s a long series of parts of a story that build our expectations, only to have the end prove us wrong. We laugh (or groan) when the unexpected twist occurs. It’s like the ball suddenly doing three loop-the-loops then stopping in mid-air. It’s not what we expect. It’s against the rules and we see that as funny.
Some artificial intelligence researchers who are interested in understanding how language works look at jokes as a way to understand how we use language. Graham Richie was one early such researcher, and funnily enough he presented his work at an April Fools’ Day Workshop on Computational Humour. Richie looked at puns: simple gags that work by a play on words, and created a computer program called JAPE that generates jokes.
How do we know if the computer has a sense of humour? Well how would we know a human comic had a sense of humour? We’d get them to tell a joke. Now suppose that we had a test where we had a set of jokes, some made by humans and some by computers, and suppose we couldn’t tell the difference? If you can’t tell which is computer generated and which is human generated then the argument goes that the computer program must, in some way, have captured the human ability. This is called a Turing Test after the computer scientist Alan Turing. The original idea was to use it as a test for intelligence but we can use the same idea as a test for an ability to be funny too.
So let’s finish with a joke (and test). Which of the following is a joke created by a computer program following Richie’s theory of puns, and which is a human’s attempt? Will humans or machines have the last laugh on this test?
Have your vote: which of these two jokes do you think was written by a computer and which by a human.
1) What’s fast and wiry?
… An aircraft hanger!
2) What’s green and bounces?
… A spring cabbage!
Make your choice before scrolling down to find the answer.
More on …
- Computer science jokes written (as far as we know) by humans [ON TLC]
- Natural Language Processing [PORTAL]
Related Magazines …
- Issue 16 – Clean up your language
- Issue 14 – Alan Turing: the genius who gave us the future
- Issue 3 – Computer Science and Entertainment
This blog is funded through EPSRC grant EP/W033615/1.
Could you tell which of the two jokes was written by a human’s and which by a computer?
Lots of cs4fn readers voted over several years and the voting went:
- 58 % votes cast believed the aircraft hanger joke is computer generated
- 42 % votes cast believed the spring cabbage joke is computer generated
In fact …
- The aircraft hanger joke was the work of a computer.
- The spring cabbage joke was the human generated cracker.
If the voters were doing no better than guessing then the votes would be about 50-50: no better than tossing a coin to decide. Then the computer was doing as well at being funny as the human. A vote share of 58-42 suggests (on the basis of this one joke only) that the computer is getting there, but perhaps doesn’t quite have as good a sense of humour as the human who invented the spring cabbage joke. A real test would use lots more jokes, of course. If doing a real experiment it would also be important that they were not only generated by the human/computer but selected by them too (or possibly selected at random from ones they each picked out as their best). By using ones we selected our sense of humour could be getting in the way of a fair test.