A new robot in the US has just learned on its own to write “hello” in Greek and in nine other very different languages like English, Japanese, Chinese, Hindi, Urdu etc. It is also able to sketch a copy of Leonardo da Vinci’s “Mona Lisa”.
This is a robotics creation at Rhode Island University of Brown, headed by the Japanese origin, Atsimobu Kotani.
The robot has a special algorithm that allows it to replicate -with a pen holding it with its robotic hand- various images or words, just by looking at them.
The algorithm uses two deep learning networks that analyze the images of a word or a sketch and conclude what are the best “strokes” to recreate the phrase or sketch. Unlike other robots that need to be programmed and fed with relevant information in advance, the new robot learns on its own how to write something it has never seen before.
So, if it sees somewhere the phrase “hello” in Greek or in another language, the robot can write it too. As Kotani said,
“It’s hard for people to see if the phrase is written by the robot or a human hand.”
The same applies for the sketches.
The creators of the robot will present it in May at the International Conference on Robotics and Automation.
Its accuracy in character replication from different languages reaches 93% (more than a six-year-old child as demonstrated by comparative tests) and can be further improved.
Researchers envision that in the future such robots will communicate better with people at work, leaving “handwritten” sticker notes, sketching diagrams etc.