A robot that has been trained to imitate a person’s voice has been created by researchers at the University of Sheffield.
The team created a “personified Ai” in a virtual world, and then trained the robot to imitate the voice of a person using voice recognition software.
The robot then learned to respond to a range of emotions including disgust, anger, fear, happiness, sadness, disgust and joy.
“It was an experiment that explored how a person-to-person communication system could be used to build a new type of artificial intelligence,” said lead author Professor Sophie Aitken.
“Ai was trained by listening to the voice and then reacting to it by interacting with the virtual world.
The AI system could then play a recording of the voice, and the recording could be uploaded to a computer.”
The team’s research is published in the Journal of the Royal Society Interface.
“This type of AI could help to build an accurate, personalized and personalized representation of human speech,” Professor Aitker said.
“The researchers also used this AI to teach the robot language to learn how to respond.”
She added: “The AI system learned to understand emotions based on what was recorded and could also respond to the simulated voice.”
Researchers from the University’s Computational Psychology Department, and at the Royal Institution in London, worked with computer science graduate students at the university to create the Ai robot.
The researchers created a virtual-world model of a human voice and training the AI system to imitate it using a virtual human voice recording.
The results showed that the Ai could “understand, interpret and respond to different types of emotions from disgust to happiness,” said Professor Ai.
The research has been supported by the National Science Foundation (NSF).
Follow @BBCNewsbeat on Twitter and Radio1Newsbeat.