The image above displays the result over 38 generations of machine learning agents. The agents start off at generation 1 by moving randomly. The genetic path of the agents that get the closest to the target are passed down to future generations. Agents that hit a wall are disregarded. It can be seen that over time, the path taken by the agents become more optimized. This is because in each generation, the best agents are kept. It is therefore intuitive that the path becomes more efficient, and I find it quite satisfying to watch them get better through unsupervised learning.