Training bots to laugh is comically difficult

It might be the robot that laughs Seen as ominous, but would it seem any less strange if he laughed at the right time during the conversation? This is a theory that scientists tested recently. A group of researchers from Kyoto University in Japan has made fun of a laughing robot named Erica that works on an AI that focuses on conversations.

Since laughter is a natural part of human dialogue, they thought it might be helpful to know how people respond to chat bots that they can also share a chuckle with. Their findings were published last week in the journal Frontiers in robotics and artificial intelligence.

AI is good at logic but laughs? Not much. To start, the researchers realized that there are different reasons why people laugh — and that makes things more complicated. To make it easier for the AI ​​system, they generalized laughter into two categories: shared social laughter, when the AI ​​laughs in response to human laughter, and one-to-one laughter, when a bot laughs in response to a topic or laughs while talking.

The researchers trained an AI model on how and when to laugh by allowing it to engage in some form of speed dating with male college students. Erica was remotely operated by an actress who spoke into a microphone and controlled physical movements such as head and other gestures.

The conversations lasted 10 to 15 minutes, and data was taken from 82 conversations. The researchers recorded conversations with microphones and cameras, annotating them based on when the social laughs and the solitary human laughs occurred, and how those laughs differed. This data was then used to train the AI ​​system to teach it when to laugh and what kind of laughter to use. They then applied their shared laughter algorithm to existing conversational programs and asked 130 volunteers to listen to it and rate how well the robot simulated empathy, understanding, and similarity to a human.

Overall, the researchers note that in situations where shared laughter is appropriate, Erica and her algorithm did a good job of convincing people that she cares about what is being said. But it has some shortcomings and limitations. Erica was good at responding to laughter, but she didn’t know when to laugh on her own. In their discussion the researchers wrote that the reason for this is that learning to respond to a prompt is more simple than understanding the real reason why the content in a conversation is funny.

Whether or not Erica gets human humor is just one part of a big project that robotics engineers are doing: giving robots social skills. Since 2017Scientists were working on how to make a robot Laugh convincingly (big tech companies like MicrosoftAnd the IBMAnd the dead also interested in this). A month ago, Italian engineers debuted a robot capable of sommelier Have a short conversation (Unfortunately, it has been delayed for the foreseeable future due to privacy concerns.) The idea is to give the robots facial expressions, body language and speech and the ability to do so understand Interacting with people’s demands will make them more interactive and better at daily interactions.

But in the end, there can be a slippery slope of seemingly natural social exchange to strange valley-Scenario. There are also ethical concerns with robots as well believable. However, there are practical reasons to continue working in this field: making talking robots less creepy and more accessible by giving them human-like features, experts believe, would be especially useful for their one day integration into healthcare, hospitalityor other Servicesoriented industries.

Leave a Comment