That is kinda what I mean. Sure automatation of content for most things has been done. Diablo 3 for example uses it by pulling from a set of tiles to put together each dungeon/map. However, the only way that it will ever grow is if the developers or artists create more parameters or tilesets for the automation to take hold of. WIth the use of a neural network, after a a given amount of generations, that would no longer need to happen. Since the network itself is crossbreeding the tiles sets(including art assets, sound effets, the limits would be the paramaters really) that already existed in a way and create completly new content.
Granted most content would be kinda of wonkey at first since it is of course learning, but the public would not see that unless it was released. A good example of this is a network that is reading the database of hearthstone cards. The first few generations it had some really out of the way cards, but after so many generations a lot of the cards that had started popping up are solid cards that could be completly viable in the game today.
But as far as I am aware, I do not think that many developers are using neural networks for content creation. Most are done in more "simplistic" ways.
That is kinda what I mean. Sure automatation of content for most things has been done. Diablo 3 for example uses it by pulling from a set of tiles to put together each dungeon/map. However, the only way that it will ever grow is if the developers or artists create more parameters or tilesets for the automation to take hold of. WIth the use of a neural network, after a a given amount of generations, that would no longer need to happen. Since the network itself is crossbreeding the tiles sets(including art assets, sound effets, the limits would be the paramaters really) that already existed in a way and create completly new content.
Granted most content would be kinda of wonkey at first since it is of course learning, but the public would not see that unless it was released. A good example of this is a network that is reading the database of hearthstone cards. The first few generations it had some really out of the way cards, but after so many generations a lot of the cards that had started popping up are solid cards that could be completly viable in the game today.
But as far as I am aware, I do not think that many developers are using neural networks for content creation. Most are done in more "simplistic" ways.
Link to the hearthstone story: http://www.reddit.com/comments/3cyi15/