Scientists have created a robot that empathizes with humans and can laugh with them. The researchers hoped to use their system to improve natural conversations between robots and people. However, the ability of robots to understand humor is probably still a long way off
Scientists at Kyoto University have developed a collaborative artificial intelligence system that responds appropriately to human laughter. This is in order to bring empathy in the conversation between a human and a robot.
Generations of philosophers and scientists have tried to understand, at least since the time of Plato, what makes a sentence funny. The Greeks believed that a sense of superiority over others was the source of humor. Sigmund Freud, a German psychologist, thought that humor was a means of releasing explosive energy. To make people laugh, the American comedian Robin Williams used his anger to bring absurdity.
No one seems to be able to agree on the answer to the question, "What's so funny?" So imagine you are trying to train a robot to laugh. But by creating an artificial intelligence that receives feedback from a common laughter system, a team of researchers at Kyoto University Japan is trying to do this. The researchers describe their innovative technique for creating humor for the Japanese robot 'Erika' in the journal Frontiers in Robotics and AI.
Not that robots are unable to understand or even laugh in response to a joke. The difficulty is developing the subtleties of human humor for an AI system to enhance ordinary conversations between robots and humans.
"We think one of the important roles of artificial intelligence in conversation is empathy," explained lead author Dr. Koji Inoue, a lecturer at Kyoto University's Department of Intelligence Science and Technology in the School of Informatics. "Conversation is, of course, multimodal, and there is not always Correct or incorrect response. Therefore, we decided that one of the ways a robot can identify with users is to share their laughter, which is impossible to do with a text-based chatbot."
A funny thing happened
In the shared laughter model, a person initially laughs and the AI system responds with laughter as an empathic response. This approach required the design (programming?) of three subsystems - one to detect laughter, the second to decide whether to laugh and the third to select the appropriate type of laughter.
The scientists trained the model on more than 80 dialogues from speed dating, a social scenario in which large groups of people mingle, or communicate, with each other one-on-one for a short period of time. In this case, the matchmaking marathon involved students from Kyoto University and Erica's company, which were operated by some amateur players.
"Our biggest challenge in this work was to identify the real cases of shared laughter, which is not easy, because as you know, most laughter is actually not shared at all," Inoue said. "We had to carefully catalog exactly which laughs we could use for our analysis and not just assume that we could respond to every laugh."
The type of laugh is also important, because in some cases a polite laugh may be more appropriate than a loud snort of laughter. The experiment was limited to social laughs versus boisterous laughs.
The robot laughs in response to the human's laughter
The team ultimately tested Erica's new sense of humor by creating four short two- to three-minute dialogues between Adam and Erica with her new shared laughter system. In the first scenario, she only uttered a social laugh, followed by only thunderous laughs in the second and third exchanges, with both types of laughter combined in the final dialogue. The team also created two additional sets of similar dialogues as baseline models. In the first, Erica never laughs. In the second, Erica makes a social laugh whenever she detects human laughter without using the other two subsystems to filter the context and response.
The researchers recruited more than 130 people in total to listen to each scenario in the three different conditions – shared laughter system, no laughter, all laughing – and rated the interactions based on empathy, naturalness, human likeness and understanding. The shared laughter system performed better than the two baseline systems.
"The most significant result of this paper is that we have shown how we can combine all three of these tasks in one robot. We believe that this type of integrated system is necessary for correct laughter behavior, not just for laughter detection and response," Inoue said.
like old friends
There are many other laughing styles that can be used to train Erica before she is ready to take the stand-up stage. "There are many more functions and types of laughs that need to be taken into account, and it's not an easy task. We didn't even try to train her on non-shared laughs even though they are the most common," Inoue noted.
Of course, laughter is only one aspect of having a natural human-like conversation with a robot.
"Robots actually need to have a distinct personality, and we think they can show that through their conversational behaviors, such as laughter, eye contact, gestures and speech style," Inoue added. "We don't think this is an easy problem at all, and it may very well be more than 10 or 20 years before we can finally have a casual conversation with a robot like we would with a friend."
More of the topic in Hayadan:
One response
Empathy, not empathy.