A recent paper called 'world models' has gotten really popular in the machine learning community. They trained an AI to play a racing game by having it learn inside of its own simulated dream environment. Meaning, the AI learned a model of what the game world was like, then was able to generate a game world that was roughly similar to what it learned and train inside of that. A simulation inside of a simulation. I'll explain how their model was structured both theoretically and programmatically in this video.
Code for this video:
https://github.com/llSourcell/world_models
Please Subscribe! And like. And comment. That's what keeps me going.
Want more education? Connect with me here:
Twitter: https://twitter.com/sirajraval
Facebook: https://www.facebook.com/sirajology
instagram: https://www.instagram.com/sirajraval
More learning resources:
https://worldmodels.github.io/
https://arxiv.org/abs/1803.10122
https://www.reddit.com/r/MachineLearning/comments/87nqbz/r_world_models/
https://medium.com/applied-data-science/how-to-build-your-own-world-model-using-python-and-keras-64fb388ba459
https://news.ycombinator.com/item?id=16860247
https://towardsdatascience.com/world-models-in-tensorflow-episode-1-2b3c217ebc8f
Join us in the Wizards Slack channel:
http://wizards.herokuapp.com/
Sign up for the next course at The School of AI:
https://www.theschool.ai
And please support me on Patreon:
https://www.patreon.com/user?u=3191693
