—
A small group of AI researchers is interested in developing an AI that can identify our laughter and generate its own amusing stuff. Artificial intelligence can read maps, diagnose malignancies more quickly than humans and play games. But can AI develop corny jokes?
This humorous topic was tested by a study team led by Harvard Business School postdoctoral scholar Michael H. Yeomans. He used the following joke and 32 others in a new study to see if people or artificial intelligence (AI) could better predict which jokes other people find amusing.
When a lawyer unlocked the door to his BMW, a car came along and slammed into it, completely ripping it off. When the cops arrived, the lawyer was furious about the damage to his prized BMW.
He wailed, “Officer, look what they’ve done to my Beeeeemer!”Â
The officer said, “You lawyers are so materialistic. You make me ill!” “You were so preoccupied with your silly BMW that you didn’t even realize your left arm had been severed!”
“Oh, my God,” the lawyer exclaimed, finally spotting the bloodied left shoulder where his arm had been. “Where has my Rolex gone?”Â
Do you think your friends—well, maybe those who aren’t lawyers—would find that joke amusing?
Why use AI for making jokes?
For the time being, though, linguistic comedy is a people thing. Jon can work in blue, with a segment on robot dating that includes cryptic texts, encrypted messaging, and the eggplant emoji—but only because a human has created and programmed a setlist for it. Finding a means to train robots to be amusing on their own would be a tremendous development, with the potential to drastically alter how we interact with the technology around us. To comprehend a person’s sense of humor, you must first grasp what they enjoy, how they think, and how they view the world. An AI that comprehends this is capable of much more than making jokes.
The subject is increasingly pertinent now, as more companies use computer-based recommendation technologies to assist customers in making decisions. The findings of Yeomans shed light on the challenges that AI technology will have to overcome to win over skeptical consumers.
75 couples, including spouses and close friends, were enlisted by the team. Seventy-one percent of the participants had known each other for more than five years.
Is AI suitable for humor?
According to one view of humor, the degree to which we find something amusing corresponds to how far a corny joke deviates from the listener’s unconscious anticipation. This one is used as a case study by Thomas Winters, a doctorate student in artificial intelligence at Katholieke Universiteit Leuven in Belgium: In a tank, there are two fish. “You operate the weapons. I’ll drive,” one says to the other.
The participants first assessed jokes on a scale of “very funny” to “not at all funny.” They then predicted their partners’ rates for eight more jokes after seeing their spouses’ ratings for four of them.
Meanwhile, a computer algorithm performed experiments to arrive at its own conclusions. The computer had no method of analyzing the jokes’ language, and it didn’t follow a model that indicated what made a joke humorous. Instead, it used “collaborative filtering” algorithms to determine which sample jokes were statistically similar to each test joke based on the participants’ previous joke likes.
Computers make fantastic suggestions, but are people willing to listen?
Businesses are putting a lot of money into complex computer algorithms that use prior consumer behavior to forecast people’s tastes and offer other things they would like, such as movies, books, clothes, and food
The big global data and business analytics market is predicted to grow by 12% this year to $189 billion and by another 45 percent to $274 billion by 2022. Netflix, for example, was so confident in machine recommendations that in 2009 it offered a $1 million prize to anyone who could create a system that could enhance prediction accuracy by just 10%.
The Bottom line
With this in mind, businesses should examine how to persuade customers to value AI-based recommended corny jokes. One possibility, according to Yeomans, is to give the machine “human-like traits.” People may be more accepting of an airline algorithm’s output if it stops briefly to search for flights, providing the impression that the machine is thinking. The more a company can explain how these systems work, the more likely customers will believe and accept it. And in today’s digital world, that’s no laughing matter.
—
This content is brought to you by Shahbaz Ahmed.
Shutterstock
