Real progress for games is necessarily progress in gameplay. And this won't happen without significant progress in areas that aren't even on the radar of game developers (or gamers) at this point: strong AI, machine learning, and natural language processing. Everything else is either irrelevant or has already been done to perfection.
We already have all the technology that is needed for excellent combat --- be it turn-based, RTwP or pure real time --- and plenty of games that have excelled in this area, e.g. Wizardry 7 & 8, ToEE, JA2, Silent Storm, Syndicate (Wars), and recently Fall of the Dungeon Guardian. The same goes for level/world/dungeon design (Deus Ex, Thief, LoG 2, RoA 2 and 3, Chaos Strikes Back, MM6), and story telling (storyfags, insert your favorites here). Level design is still overly attached to Euclidean space rather than making full use of the non-standard geometries that can be created in good engines (e.g. Unreal), and story/plot tends to be of the bland mainstream young adult variety, but this is mostly due to lacking talent and vision, not technical or financial limitations. Overall, we have already seen true mastery in these areas and there is little left to improve on (except for AI, which I'll get to in a second).
Graphics, sound and voice acting are just the perceivable instantiations of the abstract entity that makes up the actual game and thus are pretty irrelevant. Some games can be massively improved by good art design such as Diablo 1, but a good game is good even with ASCII graphics or low poly models and blurry textures. I can think of only one area where games could profit from visual improvements, and that is facial expressions for detective and adventure games. This is also the only case where voice acting can enhance gameplay as a subject's face and voice may provide important clues. But these elements are usually not a big part of RPGs, so presentation is indeed pretty irrelevant.
But there are some aspects of RPGs that are currently done in an incredibly hackneyed manner that limits gameplay. The most pressing one is dialog. Some people here complain that voice acting has reduced the amount of dialog in RPGs. But things already went wrong earlier on with the introduction of the dialog tree. The dialog tree is almost completely devoid of gameplay as it is fully scripted environment in which the player has no agency beyond choosing between a few preordained options. There are attempts to make it more dynamic --- skill checks, general polite-normal-rude"stances", Alpha Protocol's time limit --- but it is ultimately a very limited format that is completely devoid of exploration, creativity, and emergent gameplay. Keyword-based text parsers from the early 90s (and Wizardry 8
) do a little better since the player can discover hidden options that aren't in the standard list of keywords, but it is still a far cry from real-world interactions. An RPG built on a strong dialog system would be a beauty to behold, even for a combat/system/exploration fag like me.
Current industrial-grade dialog AIs and chatbots are still very limited due to a variety of reasons, e.g. the reliance on n-grams models and keywords instead of tree transductions and semantic analysis, but games have several properties that simplify the problem: games have a clear ontology that is imposed by the game engine, the game worlds are sparse (not nearly as much background and history as in the real world), and the player knows virtually nothing about anything. This means that dialog systems for games would be more like expert systems that are designed for a very small niche, for instance online customer support for an ISP or medical prescreenings by phone. These systems already work fairly well in practice. From a scientific perspective they are embarassingly simplistic, and they do struggle with very basic things like the scope of negation, but even with those limitations it would be a large improvement over what we have now. Just make sure you build an interesting game around it, otherwise you end up with dull shit like Facade.
Strong AI and machine learning will also be important to move games away from static scripted environments towards dynamically evolving systems. For example, NPC schedules are still rare because they involve a lot of scripting, are hard to debug, and barely noticeable to most players. Ideally, though, these schedules shouldn't be needed to be scripted at all and instead emerge from a system of interacting agents that operate according to a few basic principles (get plenty of good quality sleep, make sure your store is open when customers are around, make sure your store is fully stocked, etc). The same goes for NPCs and environmental interaction --- an enemy should be able to infer from the fact that water multiplies shock damage that they should throw that bucket of water at the player before using the shock trap. This kind of strong reasoning has been worked on in AI for decades, and results are pretty meagre so far. But this is once again because the real world is incredibly complex and we have no good ontology that formalizes the basic world knowledge humans possess. Games already come with a fully programmed ontology, so they would be the ideal testing ground for these models.
tl;dr Real progress must be progress in gameplay. There's no revolutionary advances waiting in the usual areas such as combat, story or visuals, but a focus on strong AI could open up completely new types of gameplay. It's not gonna happen any time soon, of course, because the know-how simply isn't there in the industry and it's more profitable to keep cranking out the same old dreck. And localization would be pretty much impossible. But maybe researchers will notice that games are a great environment for testing their models and this might lead to some interesting collaborations.
Fake edit: Woo, 100th post!
