Tekamthi
Emperor
- Joined
- Aug 12, 2016
- Messages
- 1,692
My point is we might consider building the API necessary that some future implementation of machine learning could make use of it, when this might be more accessible to the at home user -- this project has been ongoing for close to a decade, say it is still active for another decade, it's conceivable we might have more consumer level hardware to run machine learning algorithms locally at some point...I wouldn't hope for that. AI is computionally expensive to run and graphics sells more. It's also very hard to train AI for games that complex. We have much simpler games in which AI models are a rarity. It's possible, though.
The machine learning piece need not come from this community at all, so long as the game state were sufficiently readable and manipulatable via code. We build the infrastructure now, perhaps some comp sci major takes it on in the next 5 years as a term project -- we already understand what this would require, just not how it can be practically implemented (currently) once the DLL side of things is achieved. The vision here would be somewhat of 'build it and they will come' type deal, but I imagine at some point, someone will make the attempt, if the codebase supports it. Civ franchise is particularly popular amongst the crowd that gets into this kind of field, I think. Being the developer behind the number one civ opponent AI would be quite the resume item for new dev, perhaps...
Anyway maybe thinking too long term here, but something to at least consider IMO