Why humor is the frontier of artificial intelligence

Why humor is the frontier of artificial intelligence

This week my audience greatly appreciated the Motherboard post “Joke-Telling Robots Are the Final Frontier of Artificial Intelligence“, so I decided to spend a few words on humor and artificial intelligence.

Why allowing an artificial intelligence to joke is important? I like the honest way Bloomberg asks for the same question in a post which is now 4 years old, but is still a good read: Can a computer be taught to be funny? It doesn’t seem nearly as important an endeavor as getting computers to identify malignant tumors or prevent airplanes from crashing, but being able to model humor is a key problem in attempting to model human thought.

It’s important because there’s much more than humor and jokes. As Motherboard explains “Some specialists even see humor as the final frontier for artificial intelligence, because it requires mastery of sophisticated functions like self-awareness, empathy, spontaneity, and linguistic subtlety.” Yes they are adjectives, but they explain there’s something intangible behind this effort that is hard to codify. Another relevant post, by iq.intel, adds a couple of pieces to this puzzle: “According to experts like David Gelernter, professor of computer science at Yale, a machine must understand the full range and nuance of human emotion before it can be deemed capable of creative thought.” and “Taylor believes that any joke-telling computer needs to have a linguistic foundation.“. It seems to me fair to say that solving the “humor” task requires a complex multidisciplinary approach.

Is humor so complex? I spot on the web a fantastic example on this topic, provided by the eresearch.org blog, which I quote below because it poses all the questions in the right way.

On a recent trip to Australia, comedy writer, David Misch observed two manta rays engaged in—shall we say—extracurricular activities. With perfect comedic timing, he quipped “Hey! It’s fifty shades of ray!” The joke led his friend, a former computer programmer interested in artificial intelligence, to think about whether a computer could ever be programmed to make that joke—not merely be programmed to repeat it, but truly generate it were it exposed to the same circumstances that David Misch was.

In the end, it was determined that in order for an artificially intelligent computer to make that joke, it would need to be able to perform numerous, instant calculations. It would need to be able to connect the two very different topics of manta ray intercourse and human S and M, then it would need to be able to access the entirety of pop culture references to human S and M ultimately settling on Fifty Shades of Grey. Then it would need the ability to appreciate the pun, understand the rhyme of “ray” and “grey,” and gauge the audience’s ability to get the joke. Finally, artificial intelligence would need to do all of this in a blink of an eye to achieve good comedic timing (the joke wouldn’t have been funny five minutes later).

Share this post, the Hub needs your help to grow

This answers fairly well to my question. Generating humor is complex. It’s much more than a linguistic task. Natural Language processing is just a small piece of the story. Computational power and Machine Learning are clearly not enough. Somebody proposed the solution to feed a software with all the comedy, humor and jokes resources available on the web and let the machine learn what humor is. A part the role of the cultural and geography factors (I can laugh at something that is not even recognized as humor by an Asian or an African person and viceversa), we have simply failed to codify humor up to now. It’s again Motherboard the one using the best and simple words: “Despite our best efforts to explain the mysterious evolution and prominence of humor across every human culture, the core mechanisms behind it still remain elusive.” In other words we don’t even have a definition of what we want the machine to learn.

Why humor is the frontier of artificial intelligenceThere are several tests in this direction anyway. One that I like has been reported by the Newsweek: “Researchers have trained an artificial intelligence algorithm to understand and predict visual humor, representing a major development towards creating “common sense” machines. The study was limited to images created using a clip art program containing human and animal models that can be placed around objects like tables and chairs. Humans were used to judge whether pictures generated were funny or not, which allowed the researchers to generate an algorithm that understood how a specific object category contributed to humor.” This evidence is important, but if possible, adds another piece of complexity to the story, the visual side of it. It’s relatively easy to laugh when you see something funny (for example at the television), it’s much harder when you just can listen (for example at the radio).

Scientists are working hard on this topic and if you want to go more in deep with technical stuff I suggest to give a look at the Papers from the 2012 AAAI Fall Symposium, not surprisingly titled, Artificial Intelligence of Humor.

While they are working I tried to answer a very last question, do we really need an artificial intelligence to master humor? Machines should be tools to replace or support us where we have a need. I need some help to manage healthcare, energy management, perform routine or dangerous tasks, but honestly, I would never ever replace my Groucho Marx with any computer generated joke. Bloomberg post again offers a professional answer: “But Mazlack argues that even simple humor-detection capabilities would have their uses. One he offers is a sort of humor screen that would detect unintentional puns or jokes in memos or e-mail, saving the writer from embarrassment—this sort of humor-proofing would be especially useful, he says, for people writing in a non-native tongue.

In the end, it’s clear that having an artificial intelligence with humor, is only a step toward the creation of artificial intelligence. It’s not about the joke, it’s about the path to artificial intelligence.

Newsletter: because there’s more than robot humor here!

The Futurist Hub Newsletter is the greatest thing after the Big Bang. Once per month, only the news, free of spam. And with a free ebook as a bonus.

2 Comments

  1. Thanks for your post on the next Holy Grail of artificial intelligence: teaching a computer how to be funny. From what I’ve read, a computer with a sense of humor is a more human computer, and could be worth billions of dollars.

    How about a robotic home companion that a lonely senior citizen could carry on amusing conversations with? Don’t you think you could sell a few of those?

    You say that the core mechanisms behind humor still remain elusive, but you clearly outline some of those mechanisms in your post, when you talk about the manta ray joke.

    One of those mechanisms is looking up pop culture references to human S and M. Another is recognizing that “ray” rhymes with “grey.” Another is performing those tasks in the blink of an eye. I don’t think those mechanisms are elusive at all. In fact, all of those mechanisms are merely linguistic tasks that have already been performed by a computer. That computer is Watson, which beat human champions at Jeopardy. If Watson can do that, why couldn’t Watson write that manta ray joke?

    The process of putting words together in a way that makes most people laugh is not as mysterious as you think. Generating humor is not as complex as you make it out to be. There are step-by-step algorithms for comedy. In fact, I wrote a book about them. It’s called “Comedy Writing for Late-Night TV.”

    Get my book and you’ll see that the manta ray joke was written using what I call Punch Line Maker #4. Here’s a link to the book on Amazon:

    http://www.amazon.com/Comedy-Writing-Late-Night-Monologue-Short-Form/dp/0615953891/ref=sr_1_4?ie=UTF8&qid=1459106752&sr=8-4&keywords=comedy+writing

    Reply
    • Thanks Joe, that’s a great answer and I think your book could be a starting point for the developers of AI humor, who are probably fighting with some algorythms which are not always flexible. I think that if we will be able to teach humor “rules” to machines, they will anyway need a vast knowledge of the reality around them and of the specific context of that moment, before they succeed in making us smile or laugh. So it’s a long road and, as I said in the post, we don’t even need a robot to excel in this task…

      Reply

Leave a Comment