In recent months, the robots have invited themselves on stage. For example, we saw an astonishing machine of two arms playing the cello with the Malmö symphonic orchestra in Sweden, in October 2024, or a humanoid disguised handling the Keytar (a synthesizer which is worn like a guitar), during a festival in China, last July, as can be seen in the video below. In short, robotic creativity is progressing, but with results … more than ones, retraces the American online media Futurism.
It was in Switzerland, at the Institut Dalle Molle studies on artificial intelligence (located in Lugano and affiliated at the University of Italian Switzerland), that the Drummer Robot project was born, under the leadership of the researcher in mechanics Asad Ali Shahid and his colleagues. Their idea was to test the ability of a humanoid robot to learn a creative task, far from the simple utility actions. The battery seemed to them the ideal support: demanding, physical and requiring fine coordination. Their work was pre -published in July on the specialized platform Arxiv.
“Not quite my tempo”
Thanks to an automatic learning algorithm applied to the Humanoid SURITE G1 model, the robot tried to reproduce some great rock classics. On the program: Nirvana, The Police, Good Jovi or AC/DC. On paper, the simulation displayed an impressive 90%precision rate. In fact, the result is more like a chaotic rehearsal where each measurement arrives late.
Internet users obviously did not fail to react to this more than mediocre performance, some wondering even if a drummer in the flesh had been consulted during the experience, as this simulation seems unable to keep up.
Despite everything, the researchers highlight notable progress. The robot has learned to anticipate its strikes, to make arms crossings and even redistribute its baguettes in the middle of the game. So many typically human behaviors, even if in the ear, improvisation remains catastrophic.
For Asad Ali Shahid and his colleagues researchers, the adventure is just beginning. The next step is to transfer the algorithm to a real physical robot and teach it to improvise, by adjusting his game according to music. In other words, give it a form of “sense of rhythm” close to that of human musicians.
Even if no jazz festival is likely to start this drummer robot, the experiment opens wider tracks. Teaching precise timing could be used in many other fields, from physical rehabilitation to sports training.
