Researchers in Japan have created a robotic device called “Erica,” which is trained to detect laughter during speech and then decide whether to laugh in response and what kind of laughter is best.
The bots’ conversations and laughter were monitored by volunteers, who determined that their responses showed empathy and human-likeness.
Lead author Dr. “We believe that one of the important functions of speech AI is empathy,” said Koji Inoue.
“So we decided that one of the ways the bot could empathize with users is to share their laughter, which you can’t do with a text chatbot,” added Inoue.
The goal of the research, published in Frontiers in Robotics and AI, is to develop an artificial intelligence that can understand the nuances of human humor. it does not reproduce natural speech.
The researchers initially developed a “shared laughter” model that allows the robot to respond to human laughter as an empathetic gesture.
The types of laughter that Erica can choose from are “social laughter”, a polite laugh to fill the conversation with a lack of humor, or a “happy laugh” for funny situations.
To teach the AI to use the shared laugh model effectively, the researchers collected data by sending a speed dating bot, and Erica, led by four remote amateur actors, had more than 80 conversations with college students, and then the laughter that emerges during the conversations can be used solo, socially, or playfully.
This was then used to train Erica’s neural network on how to determine when to laugh and what kind of laugh to use.
Dr Inoue said: “The biggest challenge we faced in this study was to identify the true states of shared laughter, which is not easy because you know that most laughter is never shared. It can respond to any laugh.
Erica is limited in social laughter and fun, and there are many techniques you must practice before you can have natural, human-like conversation.
“There are many other functions and types of laughter to consider, and it’s not an easy task,” says Inoue.
Source: Daily Mail
Source: Arabic RT
