Laughing in Binary Code
6 9 2018
Laughing in Binary Code

"Hey Siri, tell me a joke"

There will be plenty of laughter at this year's Festivaletteratura. There are a huge number of events, the work of writers and artists who have cultivated this passion almost in secret, dedicated to one of literature's most misunderstood and unfairly criticised genres.


America, 14 February 2011: during an episode of Jeopardy!, Brad Rutter, from the show's old guard, was competing against the famous contestant Ken Jennings. Nothing strange, so far. At least until the audience realised that the contest would include a third contestant: IBM's supercomputer, Watson, in the form of a bright screen, who, against all odds, defeated the two human contests.

(caricamento...)

Marco Malvaldi's most recent book, Per ridere aggiungere acqua, starts off precisely with this anecdote. The inspiration came from Luciano de Crescenzo, who, responding to the question about if we could ever really consider a computer to be intelligent, responded: "I'll start considering my computer as smart when it starts laughing at my jokes. When it will understand that if I say, "damn, I want to kill Bob," that I don't really want to kill him.

Well, computers never laugh. They can’t understand jokes, because they require too many different layers of understanding of a language. Humour takes more than a simple a yes or no logic.

(caricamento...)

“Here,” Malvaldi says, “computational linguists enter our story. They work in order to teach computers to be as complex as a human”. In doing so, they have found a formula for humor.

A joke is made of surprise, incompatibility (between the joke and its final plot twist) and absurdity. Buster Keaton falling from a building straight into a manhole isn’t real. It’s absurd. It’s funny.

“Humor has always been a way to understand and empathize with others. When two people laugh at the same joke, their brain works in a similar way, and they both recognize it. It creates a connection.”

Is it working then? Have linguists found a way to teach artificial intelligences how to laugh? Not at all. And they’re far from the goal, but they’ll keep on trying. The end result should be funny.