Have you ever had a really awkward conversation with a stranger? There are few things more frustrating to me in life than people who refuse to hold a dialogue with you because they can't spare the time or effort. In games, I find that this happens all the time, though the issue at hand is different. Instead of a "store employee doesn't care enough about human interaction to look up from her magazine" problem, it's a limitation of the storytelling ability of games.
In games, especially RPGs and adventures, you'll encounter a large number of people on your quest. Unfortunately for the game developers, this means that they have to create a lot of arbitrary filler dialogue, often with disappointing results:
Townsperson: "Looking for Master Swordsmith? He lives at the northwest corner of the village."
Townsperson: "Looking for Master Swordsmith? He lives at the northwest corner of the village"
Storeclerk: "Sorry, we're closed."
Old Man: "Sorry, I don't know the way through the mountains."
Old Man: "Why do you talk so much? Leave me alone! [attacks]"
The problem is that it's just too much work to write multiple dialogues for every random, no-named character in a game. The end result is usually one of the above, with the character either becoming a useless drone after one exchange or mindlessly attacking you to eliminate any further opportunities for conversation.
Games have tried different approaches to this with mixed results. Final Fantasy XII is one recent example, wherein you can only talk to about 1/3 of the people in a town, who have speech bubble icons above their heads to indicate that they are chatty people. Another, more complicated attempt was Facade. This conversation-based game focused entirely on tackling this game storytelling problem. The solution was to record thousands of pre-recorded, vague responses from the two characters with whom you interact in the game, and the game chooses which to use based on what you type into your dialogue box. The project took 5 years and was a success in some rights, but it ultimately was just a huge, mega-efforted hack.
It doesn't seem like we'll have anything approaching good dialogue generation for a very long time, but think of the possibilities. Hopefully, AI will reach a point where you can speak something into your headset to an AI-controlled character, who will then respond accordingly in dynamically-generated audio. Although, once we reach that point, the AI will be sophisticated enough to conquer the world, destroy mankind, and maybe, with a lot of hard work, make a good Superman game.