Bifalgorithm post

For today's machine learning interview, make Minecraft's mob smarter. To start, I build a simple mob that focuses on two things, the player's distance and the mob's health. And that outputs one of two things, attack or flee. It's a good start, but it's just one set rule. If you have the same inputs, you get the same outputs. Won't players just exploit that rule? Good point. So then we switch to a decision tree. So if the player's close and the health is low, we can decide to flee or to chase. Okay, but a tree has no memory. If you got what you did a second ago. That's true. So we switched to reinforcement learning. So we make the mob an agent and reward it if it lands a hit or it survives and through thousands of fights, the mob learns and we don't have to hard code anything. And what do you get with that? So that gets us a zombie that flanks from around the darkness instead of going straight from the front. So how did you learn all of this exactly? I learn machine learning every day on Axon. The links in my bio.

For today's machine learning interview, make Minecraft's mob smarter. To start, I build a simple mob that focuses on two things, the player's distance and the…