Robots Are Developing A Sense Of Humor And Learning To Laugh At Our Jokes
By Nicole Rodrigues, 19 Sep 2022
Researchers at Kyoto University have developed a robot, named ‘Erica’, who is here to save all budding comedians from being met with silence after a bad joke.
Erica utilizes an AI learning software that studies human laughter, examining everything from a chuckle to loud eruptions of amusement. The study, published in Frontiers in Robotics and AI, delves deep into the subtle nuances of laughter humans have become so accustomed to in everyday conversation.
Dr Koji Inoue, the lead author of the study, notes that empathy during conversation is necessary to make talking to an AI feel more real. It isn’t just about programming bots to deliver the right answer, but for them to attach feelings and emotions behind how they speak as well. And the perfect way to do this is by equipping them with the ability to respond with the right type of laughter at the right time.
This shared-laughter model was developed by having the AI analyze if the person was laughing, whether it should do the same back, and the appropriate type of laughter it should give in return.
Training data from over 80 sample dialogues of speed dating was fed to the system so it could learn how to act in front of a large group or a one-on-one setting.
According to the study, identifying and cataloging each type of laughter proved to be a challenge as the subtle distinction between giggling and nervous snickering is something that is internalized as one experiences life. It is incredibly hard to teach to young children as they begin to navigate social cues, let alone a computer system.
Yet Erica prevails, as the scientists were able to cultivate her own humor in four different forms. In the first instance, she was able to let out a social laugh when spoken to. In the second and third instances, she chortled cheerfully, while in the last test she combined both types.
From this, 130 people were asked to rate her likeness to natural conversation against two sets of baseline dialogues that were formed. In these controlled dialogues, Erica would either not laugh at all or do so without examining the context of the conversation.
There was an overwhelming response that showed Erica performed better than the baseline dialogues.
So will Erica be replacing the funny guy of the group soon? Not exactly. For better or worse, the researchers believe we could still be decades away from talking to a robot like we would a friend.
[via The Guardian and E&T, cover image via Dimitry Ðayer / Adobe Stock]