Today, we are living in an era of technological advancement that blurs the lines between science fiction and reality. From cyborgs to humanoid robots - the replica of human beings, our world is constantly expanding and growing. We humans have innate abilities such as facial expressions, ethics, humor, and much more. But for humanoid robots, which are considered to be the exact replicas of human beings, having human-like abilities is quite difficult. But, there are no limitations on what humans can make machines do.
A team at Kyoto University has implemented an AI shared-laughter model on the humanoid robot named Erica. This model, trained to respond naturally to real-time jokes, enhances Erica's human-likeness, allowing her not only to resemble but also to interact with humans naturally.
Fig: Erica
Ref: https://robots.ieee.org/robots/erica/?gallery=photo1
In human behavior, humans decide what kind of laugh is appropriate and what is not based on the situation, the laughter can be polite or mirthful.
Fig: Shared laughter by the spoken dialogue system
Ref: https://www.frontiersin.org/articles/10.3389/frobt.2022.933261/full#B26\
In the shared laughter approach, the user laughs (initial laugh), and then the system responds with an appropriate empathetic laugh (response laugh) if needed otherwise it remains silent. The model was trained on the data from more than 80 speed-dating dialogues between male university students and Erica.
The proposed system consists of three modules:
1. Initial Laugh Detection
2. Shared Laughter Prediction
3. Laughter Type
Firstly, the model detects if the user has laughed (Initial Laugh Detection), and then it will decide if the system should laugh or not (Shared Laughter Prediction). Next, the model will determine the type of laugh it should have as a response (Laughter Type).
- Initial Laugh Detection is implemented using the Recurrent Neural Network (RNN) with the F1 score (performance metrics) of 82.6%.
- Shared Laughter Prediction and Laughter Type also scored higher than a random model.
- Erica's sense of humor has been tested by 130 volunteers who rated the shared-laughter model most favorably for empathy, friendliness, and understanding.
Even though scientists are trying hard to make robots replicate humans and can make humanoid robots more effective over time. And, as technological innovation is reshaping the whole world, perhaps one day we will be able to have a large community of robots in each country, state, or street.
But, I am still skeptical that humanoid robots can ever be able to understand us because they don't understand our exact emotions. And talking about the sense of humor, it's innate in humans but for humanoid robots, we need to train them on a large humorous dataset.
To learn more about robots' sense of humor, you can read the paper by Koji Inoue, Divesh Lala, and Tatsuya Kawahara here. Let me know your thoughts on the robot's empathetic laugh in the comment section.
Interesting that we are in an era where human can replicate ourself. No wonder what we can do in near decades
ReplyDelete