In addition to conventional industrial tasks, robots are increasingly used in daily service activities for different purposes. Accordingly, in the last decades, robotics-based education has been involved by a rapid development. This paper presents a challenge between a human and a cobot consisting in a ball-throwing task aiming to a target. A Reinforcement Learning (RL) algorithm was exploited in virtual simulation to train the robotic arm and optimizing throwing parameters such as direction and release speed. Outcomes demonstrate how RL can be exploited to train robots to perform complex tasks, improving their performance through iterative feedback on throwing accuracy. The educational aim of this activity is to inform participants about various aspects of robotics and artificial intelligence through an interactive experience and hands-on approach. After detailing the physical set-up and the technical implementation of the training process, the paper focuses on the educational impact of the proposed activity during informative events and its potential extensions in various fields.