Erica learns to laugh using AI
Researchers at Kyoto University have developed a robot that can learn to laugh using artificial intelligence (AI). Focusing on the phenomenon of ‘shared laughter’, the team headed up by Koji Inoue trained the robot Erica in the art of shared laughter. In tests, the robot was able to copy the sound of a human laugh and duplicate it, both on its own and when other people were laughing at the same time. The team is now working on teaching the robot more complex behaviour such as responding to different social contexts.
The research could help to develop artificial social skills for robots, which could have a number of applications, including in healthcare or the military. The study, published in the journal Scientific Reports, involved using a machine-learning algorithm called a deep neural network to analyse human vocalisations and determine when they were “laugh-like” sounds. This allowed the researchers to train the algorithm to recognize different types of laughs and then teach it to the robot Erica to copy them. This was achieved by applying a series of algorithms to the data created by the deep neural network and training Erica to recognise the patterns. When Erica played recordings of human laughter, her motor functions were adjusted accordingly to match the way the humans had laughed.
Results showed that Erica was able to copy the laughs of ten different people and was able to reproduce the sounds at a range of volumes and speeds. She was also able to mimic the way each person laughed individually, creating a range of different laughs within the same recording. The researchers also found that she was more likely to copy the laughter of other humans and not simply copy any sound that occurred around her. This suggests that she may have developed a sense of imitation as she grew more comfortable with the procedure.
The robot also learned how to respond to laughter from other humans and could distinguish between the type of laugh it was hearing and the emotions that it was eliciting in the people present. For example, she was able to differentiate between a laugh that was in response to a joke and one intended as friendly conversation. This indicates that she is developing an understanding of social norms and the ability to interact with people in an appropriate manner.
More Articles
