Inspired by explaining machine learning ideas in children’s books, we developed an alternative approach to introduce supervised, unsupervised, and reinforcement learning using a block-based programming language in combination with the benefits of educational robotics. Instead of using blocks as high-end APIs to access AI cloud services or to reproduce the machine learning algorithms, we use blocks as a means to put the learner “in the algorithm’s shoes”. We adapt the training of neural networks, Q-learning, and k-means algorithms in a design and format suitable for children and equip the student with the hands-on tools for playful experimentation. The children learn direct supervision modifying the weights in neural networks and immediately observing the effects on the simulated robot. Following the ideas of constructionism, they experience in practice how the algorithms and underlying machine learning concepts work. We conducted and evaluated this approach with children from primary, middle, and high school. The children of all age groups perceived the topics very easy to moderately hard to grasp. Younger learners noticed the direct supervision challenging, whereas Q-learning and k-means algorithms were much more accessible. The vast majority of high school children could cope with all topics without particular difficulties.