http://www.youtube.com/watch?v=5hfYJsQAhl0
Welcome to video games. We push buttons and things die. If you're holding your breath waiting for that to change I think you might be disappointed.
Well, that may be, but there are a lot of people in this thread that want to see "living, breathing" or "immersive" worlds. To me, there's nothing more immersion breaking than walking into a dungeon or a castle with a few of my friends and seeing a bunch of monsters just standing around waiting to be killed. If the primary means of advancement is the death of an endless stream of generic enemies for arbitrary gains, then the rest is artificial. It doesn't matter if you instance the dungeon or make more qualitatively similar dungeons so people don't get bored. The structure is artificial, because it's just a different system for the same, tired bread crumbs.
If you want to have real player interaction, you need to give players the option to have substantively different roles, including non-combat roles that can be situationally more powerful than combat roles. Then you can actually use the geography of your world as a tool of gameplay, rather than making the difference between a large world and a small world the amount of travel time between the only meaningful activities (combat).
For example, let's say you want to get the tooth of a certain lion so you can give it to some local chieftain. This lion doesn't just sit in the same area waiting to be killed. He paths around a large area. So you need a tracker or a hunter to find him. You can build gameplay mechanics around this activity. However, when you find him, you need to have a plan to kill him, so you can poison him, or fight him, or whatever, and get the tooth. If what's required actually needs something more than "combat power," than you can get outside the box of min-maxing combat power and avoid the holy trinity or underpowered class can't get a group problems.
Example 2: you want to get a certain jewel from the lockbox of a castle, so you can sell it to buy dragonhide for your bowstring, or make a consumable that increases your power level or whatever. However, the castle is full of guards and the doors or locked, so you can develop a plan to break in. You can either bring enough guys and a siege engine (player crafted) to kill everyone in a giant battle, or you can have a thief steal a key, or you can levitate onto a tower at night, or whatever. You can design a game that allows players to use non-combat roles in useful ways that still contribute to the same goals.
That said, as long as combat is the only rewarding experience (and by this I mean you are directly awarded loot, experience, character advancement from combat only), then your ability to interact with other players will only be through the same, played out mechanics. To be honest, I don't think that it requires any breakthrough in enemy AI or world design to make a game like this. Developers just need to borrow from other genres like thief/assassin games, puzzle games, crafting games enough to make a world where direct combat with 1-2 monsters if you're solo and 3-5 if you're in a group isn't the only avenue for meaningful interaction and meaningful advancement.
Look at EVE. The most successful players aren't required to endlessly kill rat after rat to build a badass character, and more than any other game, there are distinct roles for players in the process of empire building.