Picture a mature fox, venturing into the wild, ready to conquer the challenges it encounters. With no guidance, the fox learns through trial and error, steadily growing stronger and more capable with each passing day. Similarly, machines can adapt and improve their behavior based on feedback, using a technique called reinforcement learning. This method shows remarkable promise, particularly in game-settings. Machines trained through reinforcement learning, like those mastering the ancient game of GO or DOTA, can even surpass humans. These algorithms are specialized for very specific settings, so we cannot yet speak of artificial general intelligence. However, by combining reinforcement learning with other methods such as unsupervised and supervised learning,these machines may yet become more generalist agents.