Smells like reinforcement learning in real life. Master sets ups the task environment, collects samples from students, picks the best and maybe even augments them. Students watch the master and learn... and the cycle continues.

And then the master becomes a grandmaster (unless entropy explosion occurs).