Hey guys, first of it's very long time to come to Mapster, and I'm the creator of GetHypered, joint 6th place in Rock the Cabinet 2015.
Since I'm so busy in these days, recently I heard that Google's deep-learning API will be applied to SC2's melee, but I'm wondering whether that can be used&modified to Arcade maps or not.
I'm really focused about it because in two reasons. First of the API will be provided in Cplusplus, and that leads me to the thought that Arcade creators can implement their own algorithms or add a new features etc. Second, is, sounds little selfish, in Versus games like my arcade(GetHypered) the Versus AI mode is quite important. If DeepMind can be used in Arcade, If it can be, I'm willing to implement Versus AI mode, to satisfy single-players who still loves my game.
"Researchers will also have full flexibility and control to create their own tasks using the existing StarCraft II editing tools."
I found this quote from an article. There isn't much info on this yet. But it seems to suggest they plan on allowing you to be able to use some of their tools at some point.
Deepmind will be used solely offline to start. However, since it is an deep learning system, you could set it up and specify the "goal" of your arcade map, and thus have it learn and play it. However, that is massively oversimplifying the problem.
As for the C++, the API is NOT in C++, it is being done in a very system independent way. The actual communication is done via Protocol Buffers (https://developers.google.com/protocol-buffers/), which allow almost any language to have a compatible library. The actual communication will probably be done over a local websocket.
Do note that this is not meant to provide you with an exceptional AI player for your games. This is a research project to evolve AI systems to handle situations in real time that involve imperfect information, forcing the AI to either store previous knowledge and also use interpolation/extrapolation in a non predetermined scenario.
What the AI probably can help with for Arcade developers is balancing, because if you correctly set the goal, the AI may discover edge/corner/unique cases that could break your games balance, for it may consider a much wider range of tactics or combos that humans discard out of the normal social processes (bias, peer pressure, etc.).
Hey guys, first of it's very long time to come to Mapster, and I'm the creator of GetHypered, joint 6th place in Rock the Cabinet 2015.
Since I'm so busy in these days, recently I heard that Google's deep-learning API will be applied to SC2's melee, but I'm wondering whether that can be used&modified to Arcade maps or not.
I'm really focused about it because in two reasons. First of the API will be provided in Cplusplus, and that leads me to the thought that Arcade creators can implement their own algorithms or add a new features etc. Second, is, sounds little selfish, in Versus games like my arcade(GetHypered) the Versus AI mode is quite important. If DeepMind can be used in Arcade, If it can be, I'm willing to implement Versus AI mode, to satisfy single-players who still loves my game.
Anyone knows about it? Thanks.
"Researchers will also have full flexibility and control to create their own tasks using the existing StarCraft II editing tools."
I found this quote from an article. There isn't much info on this yet. But it seems to suggest they plan on allowing you to be able to use some of their tools at some point.
New to the Editor? Need a tutorial? Click Here
Want data assets? Click Here
@killduk: Go
Deepmind will be used solely offline to start. However, since it is an deep learning system, you could set it up and specify the "goal" of your arcade map, and thus have it learn and play it. However, that is massively oversimplifying the problem.
As for the C++, the API is NOT in C++, it is being done in a very system independent way. The actual communication is done via Protocol Buffers (https://developers.google.com/protocol-buffers/), which allow almost any language to have a compatible library. The actual communication will probably be done over a local websocket.
Do note that this is not meant to provide you with an exceptional AI player for your games. This is a research project to evolve AI systems to handle situations in real time that involve imperfect information, forcing the AI to either store previous knowledge and also use interpolation/extrapolation in a non predetermined scenario.
What the AI probably can help with for Arcade developers is balancing, because if you correctly set the goal, the AI may discover edge/corner/unique cases that could break your games balance, for it may consider a much wider range of tactics or combos that humans discard out of the normal social processes (bias, peer pressure, etc.).