Let's play Minecraft/Doom/StarCraft! Game Agent AI for Udacity Artificial Intelligence Nanodegree Term 2!
38 Comments
G
Gonzalo Barrientos
9 years ago
Totally agree with this idea.
K
Kevin Avignon
9 years ago
Featured
We were talking about this in the game agent ai channel; we'd like to implement a really smart bot that would learn by experimenting and exploring to play with street fighter games and have a flexible data model to be able to appropriately work when changing environment such as moving from SF2 to SF3 alpha for instance. To make it easier on us, because CV with creating a proper data model with Q-learning seems extremely heavy, it'd be cool to do a mod directly in Minecraft ! Experimenting with game AI is something that I'm fond of as I am right now working on a game AI for a RPG prototype. I would really like to have see either Spec rec swapped with Game AI or simply adding Game AI as a new possibility for students.
A
Allan Visochek
9 years ago
great idea!
F
Francis
9 years ago
love the idea
P
Pengyu
9 years ago
Want to know something new for AIND term 2. :)
A
Abhinav Goyal
9 years ago
I got my feet wet in the realm of Q-Learning with the MLND and would love to take it a notch (or two) higher with the AIND.
A
Anonymous
9 years ago
Yesssss! q learning ftw!
M
Madhava Jay
9 years ago
Featured
I love Doom / Quake and StarCraft! Whats not to love. Given that Facebook has provided Broodwar as an AI platform, and Google with Blizzard are looking to provide SC2 it would be great to do this in the AIND!
T
Taqiyeddine Abderrahim
9 years ago
It sounds challenging, I am totally agree ;)
A
Anonymous
9 years ago
Good idea~~
A
Angel Martinez-Tenor
9 years ago
Working with OpenAI Gym in AIND. Great idea!
K
Krishna Kumar
9 years ago
Let's make this happen. This would seriously be the killer specialization for AIND.
N
Nash Taylor
9 years ago
Featured
If I got the opportunity to apply reinforcement learning to a game playing agent, it would be the most exciting project I've ever done in my life.
Totally agree with this idea.
We were talking about this in the game agent ai channel; we'd like to implement a really smart bot that would learn by experimenting and exploring to play with street fighter games and have a flexible data model to be able to appropriately work when changing environment such as moving from SF2 to SF3 alpha for instance. To make it easier on us, because CV with creating a proper data model with Q-learning seems extremely heavy, it'd be cool to do a mod directly in Minecraft ! Experimenting with game AI is something that I'm fond of as I am right now working on a game AI for a RPG prototype. I would really like to have see either Spec rec swapped with Game AI or simply adding Game AI as a new possibility for students.
great idea!
love the idea
Want to know something new for AIND term 2. :)
I got my feet wet in the realm of Q-Learning with the MLND and would love to take it a notch (or two) higher with the AIND.
Yesssss! q learning ftw!
I love Doom / Quake and StarCraft! Whats not to love. Given that Facebook has provided Broodwar as an AI platform, and Google with Blizzard are looking to provide SC2 it would be great to do this in the AIND!
It sounds challenging, I am totally agree ;)
Good idea~~
Working with OpenAI Gym in AIND. Great idea!
Let's make this happen. This would seriously be the killer specialization for AIND.
If I got the opportunity to apply reinforcement learning to a game playing agent, it would be the most exciting project I've ever done in my life.
I support
I support this. it needs to happen.
I support this
I highly support this.
This needs to happen!!